paper_id
stringlengths 9
16
| version
stringclasses 26
values | yymm
stringclasses 311
values | created
timestamp[s] | title
stringlengths 6
335
| secondary_subfield
sequencelengths 1
8
| abstract
stringlengths 25
3.93k
| primary_subfield
stringclasses 124
values | field
stringclasses 20
values | fulltext
stringlengths 0
2.84M
|
|---|---|---|---|---|---|---|---|---|---|
1805.00656 | 1 | 1805 | 2018-05-02T07:45:02 | Rolled-up self-assembly of compact magnetic inductors, transformers and resonators | [
"physics.app-ph"
] | Three-dimensional self-assembly of lithographically patterned ultrathin films opens a path to manufacture microelectronic architectures with functionalities and integration schemes not accessible by conventional two-dimensional technologies. Among other microelectronic components, inductances, transformers, antennas and resonators often rely on three-dimensional configurations and interactions with electromagnetic fields requiring exponential fabrication efforts when downscaled to the micrometer range. Here, the controlled self-assembly of functional structures is demonstrated. By rolling-up ultrathin films into cylindrically shaped microelectronic devices we realized electromagnetic resonators, inductive and mutually coupled coils. Electrical performance of these devices is improved purely by transformation of a planar into a cylindrical geometry. This is accompanied by an overall downscaling of the device footprint area by more than 50 times. Application of compact self-assembled microstructures has significant impact on electronics, reducing size, fabrication efforts, and offering a wealth of new features in devices by 3D shaping. | physics.app-ph | physics | Rolled-up self-assembly of compact magnetic inductors, transformers and resonators.
Dmitriy D. Karnaushenko#, Daniil Karnaushenko#,*, Hans-Joachim Grafe, Vladislav Kataev,
Bernd Büchner, Oliver G. Schmidt
Dr. D. D. Karnaushenko, Dr. D. Karnaushenko, Prof. Dr. O. G. Schmidt
Institute for Integrative Nanosciences, Institute for Solid State and Materials Research Dresden
(IFW Dresden), 01069 Dresden, Germany
E-Mail: [email protected]
Dr. H.-J. Grafe, Dr. V. Kataev, Prof. Dr. B. Büchner
Institute for Solid state Research, Institute for Solid State and Materials Research Dresden (IFW
Dresden), 01069 Dresden, Germany
Prof. Dr. O. G. Schmidt
Material Systems for Nanoelectronics, Chemnitz University of Technology, 09107 Chemnitz,
Germany
Center for Advancing Electronics Dresden, Dresden University of Technology, 01062 Dresden,
Germany
#Authors contributed equally to this work.
*Corresponding author.
Keywords: strain engineering, self-assembly, coil, transformer, shapeable electronics.
Three-dimensional self-assembly of lithographically patterned ultrathin films opens a path to manufacture
microelectronic architectures with functionalities and integration schemes not accessible by conventional
two-dimensional technologies. Among other microelectronic components, inductances, transformers,
antennas and resonators often rely on three-dimensional configurations and interactions with
electromagnetic fields requiring exponential fabrication efforts when downscaled to the micrometer range.
Here, the controlled self-assembly of functional structures is demonstrated. By rolling-up ultrathin films
into cylindrically shaped microelectronic devices we realized electromagnetic resonators, inductive and
mutually coupled coils. Electrical performance of these devices is improved purely by transformation of a
planar into a cylindrical geometry. This is accompanied by an overall downscaling of the device footprint
area by more than 50 times. Application of compact self-assembled microstructures has significant impact
on electronics, reducing size, fabrication efforts, and offering a wealth of new features in devices by 3D
shaping.
Strain-driven self-assembly of thin films into three dimensional geometries has become a powerful
technique to produce a plethora of differently shaped microarchitectures.[1,2] In combination with
established micropatterning techniques, highly parallel processing of ultra-compact three-dimensional
devices on a single chip has been demonstrated.[3,4] Potential application areas of such mass-produced
components are manifold ranging from basic microelectronic components and sensors to systems and
robotics both on and off the chip.[1,5–7]
A special and particularly interesting category of micro-origami is the strain-driven roll-up of patterned
layer stacks into microtubular devices.[8,9] Such devices have been employed for fabrication of capacitors[10],
inductors[11], transistors[12], sensors[13], and microbots[14], to only name a few. Rolled-up nanotechnology
has recently seen significant progress driven by cleverly designed layer stacks and material choices. The
introduction of a polymer based platform in combination with advanced functional materials and device
layouts has led to fully integrated biomimetic circuitry[15], microtubular GMI sensors[16] and impedance
matched antenna arrays.[17] However, the complexities associated with the self-assembly of microorigami
devices and overall process compatibility remains challenging requiring the right choice of materials and
establishing predictable and viable design rules. For instance, reproducibility of the geometry is one of the
key challenges, which affects functionality of a final device and must be precisely controlled in order to
achieve high yield and reliability of electronic systems.[15,17]
Here, we demonstrate advanced roll-up assemblies for a new category of devices. We create inductive and
inductively coupled systems for applications
in
transformers and ultra-compact high quality
electromagnetic resonators for nuclear magnetic (NMR) and electron spin (ESR) resonance characterization
set-ups. Normally, fabrication of these devices require numerous fabrication steps[18,19] in a conventional
planar microelectronic process.[20,21] In contrast, application of the parallel self-assembly strategy reduces
fabrication efforts and footprint area[11], and enhances performance up to an order in magnitude if compared
to their planar counterparts. The geometry of the assembled rolled-up stack shown in Fig 1 a contains
polymeric and metallic parts that finally reveal magnetic inductive coils. The initial planar stack consists
of four patterned, functional, shapeable, ultrathin, polymeric layers with a rectangular geometry having
length (L) and width (W) of 17 mm and 5.5 mm respectively. The layers contain the sacrificial layer (SL),
hydrogel layer (HG) and two rigid reinforcing layers made out of polyimide (PI) material (Fig. 1 b). Each
of the two rigid layers was patterned using a specially designed structure required for a correct and
reproducible self-assembly of the overall architecture into a cylinder. Normally, structures with such an
aspect ratio and geometry do not assemble properly into tubular shapes, rather fold from sides and transform
into polygonal geometries.[22,23] To avoid such problems a number of bracket structures and crack
propagating edges (CPEs) were implemented on both sides of the rigid layer in order to guide the self-
assembly process (Fig 1. b) in a desired direction during the batch wafer (100 x 100 x 1 mm3) fabrication
process. The optimal spacing of ~200 µm between every couple of brackets was determined considering
the desired final circumference of the microtube.[17,22,23] The CPEs, introduced in the double reinforcing
layer (PI 1 and 2 with 1:1 thickness ratio), ensures a controlled rupture of the bracket structures during the
rolling process.
Figure 1. Illustration of the self-assembly of functional 3D magnetic components using shapeable ultrathin
films. (a) Self-assembled and encapsulated 3D device revealing coils and their mutually coupled
configurations. (b) Shapeable layer stack consisting of three distinct polymeric structures, namely the
sacrificial layer (SL), the hydrogel layer (HG) and the reinforcing polyimide layer (PI). The latter possesses
a two layer structure where the first layer (PI 1) is equipped with fixators and the second (PI 2) layer
develops crack propagation edges (CPEs), which helps to release the brackets during the rolling process
and guide the overall (c) self-assembly process. (d) Rolling of a straight wire leads to a Swiss Roll coil, (e)
while rolling of the tilted wire forms a helix geometry.
We applied different planar layouts of conductors (see e.g. Fig. 1 c I) on the surface of these layered
shapeable materials. The roll-up self-assembly leads to a Swiss-roll architecture (Fig. 1 c I-III, d) with a
final diameter of ~300 µm, winding thickness of 4 µm and a roll width W of 5.5 mm. Upon self-assembly
the footprint of the structure is effectively reduced from about 100 mm2 to 1.5 mm2 (Fig 1. c). The
inductance of a conductor in the assembled state depends on its orientation on the planar surface (compare
Fig. 1 c, d and e) and is maximum when the device is assembled into a zero pitch coil (Fig.1 d). The planar
design of the conductor should contain, due to the constraints imposed by the Swiss-roll geometry, a couple
of straight conductors with a perpendicular section that electrically connect both coils after the self-
assembly. This so called "𝚷" geometry allows (see Fig. 1 c) electrical feeding from the "anchor" side of
the device attached to the rigid support. We have prepared a number of such shapes (Fig. 2 a-e) made out
of Ti – Cu – Ti with a conductor cross section of 100 x 3 µm2 in a single lithographic run (the exact geometry
is shown in Sup. Fig. 1) and self-assembled into Swiss-rolls in a batch wafer scale process (Fig. 2 b).
During this process, the planar "𝚷" shaped conductors were transformed into a number of coils
corresponding to the equivalent circuit shown in Fig. 2 c. Each conductor, oriented along the rolling
direction, forms a coil, while the perpendicular conductors do not reshape during this process and just
provide an electrical connection between the left (Fig. 2 d group 1) and right (Fig. 2 e group 2) set of coils
(Fig. 1 a, 2 a). Each conductor possesses a DC resistance ranging from 1 to 10 Ohm in both the planar and
3D assembled states corresponding to the cross section and geometry of the conductors. With this resistance
value it was possible to push up to 300 mA of AC current without any thermal damage of the 3D structures.
The value of the quality factor of the inductors ranges from 0.5 to 5 demonstrating an average enhancement
in the performance of up to 2.5 times due to the rolling process, measured at 10 MHz (the spectrum is
shown in Sup. Fig. 2). It is clear from the geometry of the planar layout that the self and mutual inductances
of each planar loop should have a direct relation to the wire length and the area surrounded by the wire. For
instance, the planar conductor patch between electrodes (0) and (8) has the initial value of the planar
inductance equivalent to LPL(0)-(8) = 35.5 nH at 10 MHz. The loop between the electrodes (3) and (7)
possesses an inductance LPL(3)-(7) = 36.5 nH – a little bit larger than LPL(0)-(8) due to the longer length of the
conducting path. The measured inductance LPL(1)-(2) of the smallest loop between the electrodes (1) and (2)
has a value of only 11 nH.
Figure 2. Self-assembled micro coil structures that include inductors and their mutually coupled
configurations – transformers. (a) Three equivalent rolled up structures fabricated in a single run on the
glass substrates using (b) batch fabrication and self-assembly process. (c) Planar layout of conductors
forming a set of coils and its representative equivalent circuit. The circuit consists of (d) left and (e) right
side of coil groups. (f) Inductance spectra demonstrating substantial (3.9 times) enhancement of the self-
inductance between the electrodes (0) and (8) due to the self-assembly process.
The self-assembly into the tubular structure (Fig. 2 c) converts the overall inductance of the wire LPL(0)-(8)
into the sum of the coil inductances LRU(0)-(8) = L1 + L9 = 138.2 nH that includes the inductance of the straight
section as well. As can be recognized from the obtained characteristics (Fig. 2 f), the inductance of LRU(0)-
(8) in the assembled state is 3.9 times higher if compared to the LPL(0)-(8) inductance of its initially planar
counterpart. Similarly, two other inductances in the rolled state demonstrate increased values of their
inductances, ~2.5 times for LRU(3)-(7) = L6 + L4 + L5 + L8 = 91.4 nH and ~3.3 times for LRU(1)-(2) = L2 + L3 =
35.9 nH compared to their planar counterparts respectively at 10 MHz (the spectra are shown in Sup. Fig.
3). Here, the inductance ratio between the rolled and planar geometries is affected by the series inductance
of the straight conductor (parallel to the Swiss-roll axis) according to
𝐿𝑅𝑈+𝐿𝐶
𝐿𝑃𝐿+𝐿𝐶
, where LC, LPL and LRU are the
inductances of the straight conductor, the same conductor in the planar and rolled states respectively. The
inductance of the straight conductor is obviously not altered due to the rolling process, thus strongly
affecting the inductance ratio of the structure between the electrodes (3) and (7), but not the structure
between the electrodes (1) and (2) (see Fig. 2 c). The ratio between the inductances in the rolled and planar
states provides a good merit for the self-assembly process that has to be carefully analyzed during the design
for the sake of maximum final performance of 3D assembled devices. Besides, this analysis reveals an
optimal 2D layout of conductors allowing to effectively exploit the available 3D space. In this respect, the
self-assembly of the coils demonstrates substantial enhancement of the inductance due to the geometric
transformation and shrinking of the overall footprint area by up to 60 times.
The rolled coils (Fig. 1 a, 2 a, c - e) e.g. L1 & L2, L1 & L4 & L5 and L8 & L9 are mutually coupled, due to
their close proximity. This is obvious from the circuit shown in Fig. 2 c. The coil set (L1 & L9) between the
electrodes (0) and (8) shares a mutual inductance with the coil set (L5 + L4 & L8) formed between the
electrodes (3) and (7) respectively MRU(08)-(37), which is equivalent to 18.8 nH. The mutual inductance of the
rolled structure is higher, compared to the mutual inductance MPL(08)-(37) = 12 nH of the initially planar
structure. A similar character is observed for the set of coils between the electrodes (0) and (8) coupled to
the set of coils between the electrodes (1) and (2), revealing more than twice an enhancement of the value
of the mutual inductance in the rolled state MRU(12)-(08) = 3.9 nH compared to MPL(12)-(08) = 1.5 nH of the
planar structure (the spectra are shown in Sup. Fig. 4 a-d). The mutual inductances of the planar and rolled
geometries have, however, different origins. The flux linkage in the planar loop structure is mostly
perpendicular to the surface, which vanishes in the self-assembled structure. Due to the rolling process, the
set of axial coils is formed with axially oriented flux, allowing inductive coupling of these coils. This aspect
is appealing for fabrication of axial micro transformer structures in the self-assembled state (Fig. 2 c). The
key parameter of these devices is the coupling coefficient that shows how much of the total flux is shared
among the coils. A coupling coefficient kRU(08)-(37) = 16.8% was measured for the same set of coils, which
in the rolled state of the device is half of the coupling coefficient kPL(08)-(37) = ~33.4% of the initially planar
structure (the spectra are shown in Sup. Fig. 5 a-d). However, an opposite behavior is observed for the set
of coils between electrodes (0) and (8), and the set of coils between electrodes (1) and (2). This set of
mutually coupled coils demonstrates enhancement of the coupling coefficient in the rolled state kRU(08)-
(12)=9.8% (spectra are shown in Sup. Fig. 5 b), over the low coupling coefficient of only kPL(08)-(12) = 7.6%
in the planar state. Such a difference in the behavior of the two coupled sets of coils can be understood from
the layout of their planar conductors (Fig. 2 c) and the definition of the coupling coefficient, which is
k=M/√𝐿𝑃𝐿𝑆, where M, LP and LS are the mutual and self-inductances of the primary and secondary coil
sets respectively. An increase in the self-inductance, as it occurs during roll-up, should be accompanied by
an increase of the mutual inductance in order to keep the coupling coefficient constant. This is not the case
for the partial inductance L6 (Fig. 2 c) located further away from the partial inductance L1, compared to the
positions of the partial inductances L4 and L5. Thus, the increase of the inductance L6 cannot be compensated
during the rolling by its vanishing mutual inductance with the inductance L1, therefore kRU(08)-(37) < kPL(08)-
(37). The axial proximity of coil L6 to L4 and L5 also explains the only 2.5 times enhancement of the LRU(3)-(7)
self-inductance compared to other coil sets. In this respect, coil L6 has a vanishing mutual inductance with
the inductances L4 and L5, negatively affecting the overall inductance increase between electrodes (3) and
(7). This difference clearly demonstrates the demand of accurate design rules for the planar layout by
carefully defining the position and orientation of the conductor structures directly affecting the final
characteristics of the self-assembled 3D device.
For many applications, including signaling circuits as well as power converters, a particularly interesting
configuration of mutually coupled coils is the centered-tap-secondary transformer.[24] This configuration
mainly relies on inductances L1, L4 and L5 in the self-assembled device (Fig. 2 c). The windings formed by
the inductances L6, L7 and L8 should have much less coupling due to the twice larger distance between
them. Otherwise, inductances L9 and L8 should have a strong mutual coupling that contributes to the overall
coupling between the conductors (5) - (7) and (0) - (8). The measured characteristics confirm this behavior
(the spectra are shown in Sup. Fig. 3 c, d and 4 c, d). The self-inductance value and the inductance ratio of
the set between (3) - (5) is much smaller (LRU(3)-(5) = 37.2 nH and LRU/LPL =1.9 respectively) compared to
the set between (5) - (7) (LRU(5)-(7) = 83.7 nH and LRU/LPL =2.8 respectively), which is purely due to the
geometry of the initially planar conductor containing several small sections with vanishing self- and mutual
coupling among them in the self-assembled state. Sections (3) - (5) and (5) - (7) differ by almost a factor
of two in their coupling coefficient kRU(35)-(08)=13.5% and kRU(57)-(08)=25.6% with the section (0) – (8), which
is obvious considering the relative positions (Fig. 2 c) of their conductors (the spectra are shown in Sup.
Fig. 5 c, d). Thus, it is very critical to consider the right design of the 2D layout in order to achieve
symmetric operation of such an electronic device.
Starting at about 40 MHz, the mutually coupled inductors reveal an increase and then drop in their
inductances to almost 0 H at a frequency of about 97 MHz (the spectra shown in Fig. 2 f and spectra's
shown in Sup. Fig. 3). This happens due to the resonance condition promoted by the intrinsic capacitance
of the rolled-up coils. For demonstration purposes we energized an LED connected to the secondary side
(0) - (8) of the micro transformer by powering the primary (3) - (7) side (detailed circuit in Sup. Fig. 6).
The resonance frequency of the primary and secondary coil sets was matched to the frequency of ~40 - 50
MHz with external capacitors of about 100 pF connected in parallel on either side of the transformer. At
the resonance frequency the energy can be transferred with higher efficiency[25] thus the LED can generate
light (Fig. 2 g, h) rectifying the power transferred over the mutual flux between the self-assembled micro
coils. The overall insulation among windings was measured to be more than 4 MOhm at 32 V, which was
characterized by the source measuring unit Agilent B2902. All the impedance characterizations were
performed by the two port network analyzer Agilent ENA5071 with an appropriate calibration routine.
We have applied the micro scale self-assembled coil as an NMR transducer to probe nuclear spin states in
a small volume of a model material. The self-assembled micro coil was introduced in a commercial NMR
probe forming an LC resonator. The resonance frequency of the resonator was adjusted with an external set
of capacitors to the Larmor frequency of 1H nuclear spins precessing in an external magnetic field. In this
configuration the micro coil is able to transfer energy into the material subsystem and receive the response
signal. For demonstration purposes we have integrated the tuned transducer (Fig. 3 a, b) into a commercial
NMR setup equipped with 3 T and 7 T superconducting magnets. We used glycerin in the inner opening of
the rolled up structure (Fig. 3 a) for the test measurement. Glycerin shows two hydrogen (1H) peaks at
about 3 ppm and 4.5 ppm, which are revealed by a standard solenoid copper coil possessing a diameter of
d = 0.6 mm and a length of l = 2 mm. (Fig. 3 c). In the same figure we demonstrate the first successful
NMR measurements performed with the self-assembled micro coil. The signal shows a clear signature of
1H in glycerin at the correct resonance frequency. The broader line width in the obtained signal for the
rolled up micro coil compared to the standard solenoid copper coil is expected for small scale coils and just
confirms the functionality of the transducer. Possessing a strong field strength such coils are affected by
different local magnetic susceptibilities of the coil itself and of the materials surrounding the coil. This
effect is well known[26] and increases with the downscaling of the coil size where the bulk of the sample
more closely approaches the coil windings[27].
Finally, we demonstrate the feasibility to realize high quality LC resonators without using external
capacitors just relying on an alternative planar layout of the conductor to form a parallel plate capacitor and
the coil in a single batch self-assembly process (Fig. 3 d). The resonator, in its 3D shape, was designed
using an FEM model based on the ANSYS Academic software package (Fig. 3 e) and then fabricated in
the same process as the micro coils with the only difference in the planar layout of the conductor. The most
prominent design (Fig. 3 e, f) achieved a quality factor >40000 (Fig. 3 g). In this design, the coil stripe
(Fig. 3 e right and the planar layout is shown on Sup. Fig. 1 b) is shifted away from the capacitor stripe
(Fig. 3 e left) in order to ensure spatial separation of electric and magnetic fields in the assembled state that
avoids radiation of electromagnetic waves. The self-assembled structure contains a cylindrical parallel plate
capacitor (Fig. 3 e) and the multiwinding inductor, resulting in a high quality factor resonator tank.
Figure 3. Self-assembled high quality micro coils and micro resonators applied for NMR and ESR
characterization. (a) Micro tubular self-assembled architecture possessing an opening of 330 µm suitable
for insertion of a small amount of an NMR sample. (b) Rolled up self-assembled devices integrated into a
commercial NMR probe and connected in a circuit with high-Q capacitors forming an LC resonator. (inset)
Magnified view of the device demonstrating the soldered connections directly to the printed circuit board.
(c) NMR response of the rolled coil and sampled with a glycerol shows characteristic peaks at the correct
positions, which is compared to a regular Cu wire solenoid. (d) Self-assembled high quality micro
resonators that were fabricated in a batch wafer scale process. (e) Optimal design of the 3D resonator was
simulated and (f) fabricated using shapeable ultrathin films. (g) The high-Q resonator demonstrates
variation in the resonance frequency and the quality factor among three fabricated in a single run devices.
This happen due to the slight variations in the geometry of the devices. The highest Q resonator was applied
to measure the ESR spectrum. (h) Measured response peak of the DPPH radicals in magnetic field around
196 mT and f = ~5.5 GHz. Several versions of rolled up structures (i, j) revealing various geometries of
conductors.
For testing, a simple ESR setup was built integrating the micro coil ESR resonator (Fig. 3 f). The setup was
equipped with a strong biasing electromagnet, a shimming couple of coils and small field scanning coils
supporting a very precise settling of the magnetic field between the pole shoes (2.5 µT). An additional
filter circuit and a custom designed nonmagnetic probe were implemented in order to eliminate residual
magnetic inhomogeneity's and noise coming from the power supply of the magnet, which otherwise
disturbs characterization of the high Q resonances presented in the ESR model sample. We used 2,2-
diphenyl-1-picrylhydrazyl (DPPH) as the model material, which is an important functional agent in
chemistry i.e. for ESR monitoring of the antioxidation activity of biologically relevant substances.[28] The
material was loaded into the resonator and characterized with an applied magnetic field, which was swept
around 196 mT. The central field of 196 mT corresponds to the ESR resonance of DPPH free radicals at
about 5.5 GHz which is close to the resonance peak of the chosen resonator (Fig. 3 h). We had to choose
the field and an appropriate frequency range between 5.48 GHz and 5.53 GHz measuring return loss (S11)
to find resonance characteristics of the micro resonators (Fig. 3 g). The variation in characteristics of the
resonators accounts for a slight sensitivity of the high Q device to a deviation of geometric parameters of
the rolled up structure. A subtle variation in the diameter or the winding misalignment can be dramatic for
the device performance affecting the capacitance or the inductance.[17] The resonator design is sufficiently
tolerant towards this issue (Fig. 1 a) purely relying on the planar conductor design among other less stable
structures (Fig. 3 i, j).
In conclusion, we have designed and fabricated rolled-up cylindrical micro coils, transformers and
resonators relying on shapeable polymeric ultrathin films. As the result we achieved up to 48 different
devices on a single square (100 x 100 mm2) shaped wafer. We could demonstrate a fully parallel wafer
scale process utilizing the same fabrication routine, shapeable layer stack and the 3D tubular geometry. The
self-assembly of very simple conductor structures into the Swiss-roll geometry provided means to fabricate
coils, transformers and high quality resonators. The overall process allows omitting a number of
intermediate steps, which are otherwise required in conventional 2D processing schemes. We showed for
the first time that shaping of the initially planar structure like a conductor can lead to >50 times more
compact 3D inductive coils with enhanced up to 4 times inductance, reaching >100 nH at MHz regime. The
mutual inductance among some inductors and the coupling coefficient were enhanced promoting
applications such as ultra-compact transformers. We envision applications of these magnetic self-assembled
devices in microelectronics, radio frequency, communication devices and power converters. Fully
integrated systems will employ micro scale inductances, transformers, and resonators operating as e.g.
impedance matching networks, filters or -- if combined with active electronics[15] -- fully integrated compact
DC-DC converters. Additionally, sensitive micro coil electromagnetic resonators were realized, and their
application for micro scale NMR and ESR spectroscopies was demonstrated. Application of NMR inert
materials and other measures to create a homogeneous magnetic surrounding of the micro coil will
substantially increase the resolution of the rolled-up NMR micro coils. Shimming and gradient coils can be
accommodated within the same geometry to define the magnetic field profile of the resonators. These
devices are promising if combined with sampling systems such as microfluidics or micro catheters for
realization of compact resonance spectroscopies and imaging techniques of micro scale samples.
Experimental Section
(1) Treatment of substrates:
Square shaped glasses of 100 x 100 x 1 mm3 were used as the substrates in this work (D263T eco glass,
SCHOTT AG, Mainz, Germany). Initially, all the substrates were washed in the professional washer DS
500 (STEELCO S.p.A., Riese Pio, Italy) to remove all of the organic and inorganic contaminants, presented
in a form of dust or films. Then, the surface was activated with oxygen plasma in the GIGAbatch 310M
(PVA Metrology & Plasma Solutions GmbH, Wettenberg, Germany). This ensures further chemical surface
modification with a monolayers of 3-(trimethoxysilyl) propyl methacrylate (TMSPM). For this, the glasses
were placed in the vacuum oven at 150°C for 2h together with 150 µL of TMSPM.
(2) Sacrificial layer:
The sacrificial layer was prepared from acrilic acid (AA) and hydrated LaCl3 obtained from Alfa Aesar UK
and used without further purification. Firstly, 10 g of AA was partially neutralised using sodium hydroxide
(Sigma-Aldrich Co. LLC., Germany) until the solution reaches pH = 5.5. The pH value of the solution was
monitored using the pH meter CyberScan PC510 (Eutech Instruments Pte Ltd., Singapore). Than we added
7.36 g of LaCl3 × 7 H2O to this solution, dissolved in 50 ml of DI to achieve the proportion of AA : La =
7:1. The AA : La solution was stirred for 30 min using a magnetic stirrer and during this time pH was
slowly increased to 10 by dropping of NaOH. At elevated pH lanthanum complex precipitates and collected
using a filter paper and a vacuum exicator with subsequent washing in DI water until litmus test paper
(DuoTest, Macherey-Nagel GmbH & Co. KG, Germany) shows neutral reaction. White precipitate was
collected and dried under nitrogen atmosphere resulting in a white powder of 4.68 g. As a final step, the
dried and milled material was dissolved in AA at the concentration of 33% (wt/wt), photosensitized using
2% (wt/wt) of 2-Benzyl-2-(dimethylamino)-4-morpholinobutyrophenone and 4% (wt/wt) methyl
diethanolamine (MDEA) (Sigma-Aldrich Co. LLC., Germany). Achieved solution was applied on the
substrate using spincoating at 3000 rpm for 35s with the acceleration of 500 rpm/s. After predrying at 35°C
for 5 min it was exposed with a broadband UV mask aligner MJB4 (SÜSS MicroTec SE, Garching,
Germany) for 30 sec and then developed in DI water for 30 sec. In order to remove water insoluble rests of
photo initiator a subsequent washing in propylene glycol monomethyl ether acetate (PGMEA) (Sigma-
Aldrich Co. LLC., Germany) was done for 5 sec. Finally, samples were hard baked at 220°C for 5 min
revealing high resolution structures of the sacrificial layer on the glass surface.
(3) Shapeable bilayers:
The assembly of the planer stack occurs in the bi-layer polymeric system due the swelling of the hydrogel
layer (HG) reinforced by the polyimide layer (PI) in a basic conditions pH= 9 of the rolling solution.
Preparation of the bilayer started from the synthesis of HG using poly(ethylene-alt-maleic anhydride)
(PEMA) and N-(2-hydroxyethyl)methacrylate (HEMA) (Sigma-Aldrich Co. LLC., Germany). At first 6 g
of PEMA has been dissolved in the 50 ml of N,N-Dimethylacetamide (DMAc) (Sigma-Aldrich Co. LLC.,
Germany), then to the solution 5.66 g of HEMA was added. All components were mixed thoroughly for
24 h at room temperature by roll-mixer to finish the reaction.
The PI was prepared by reaction of 12 g 4,4'-Methylenedianiline (MDA) (Sigma-Aldrich Co. LLC.,
Germany) and 19.65 g of 3,3',4,4'-Benzophenonetetracarboxylic dianhydride (BPDA) (Sigma-Aldrich Co.
LLC., Germany). Both monomers were dissolved in 20 ml DMAc and gradually over an hour reacted
together under culling conditions at 15°C, resulting in the highly viscous solution. Then a
dimethylaminoethyl methacrylate (DMAEMA) (Sigma-Aldrich Co. LLC., Germany) added in a 1:1 molar
ratio to the number of carboxylic groups in BPDA. The solution was left in the mixer for 4 h under cooling
conditions to complete the reaction. At the end HG and PI solutions were photosensitized with 4% (wt/wt)
of 2-Benzyl-2-(dimethylamino)-4-morpholinobutyrophenone (Sigma-Aldrich Co. LLC., Germany) and
additionally diluted to have thickness in the range of 500 nm and 1500 nm respectively at 3000 rpm.
The bilayer stack was formed in a sequential photoparnening process of HG and PI. Materials in amount of
3 ml were applied on the substrate through 1 µm filter using following settings: polymer pre-spinning step
was made at 500 rpm with acceleration 50 rpm/s for 10 s, main spinning is done at 3000 rpm for 30 s with
acceleration of 500 rpm/s. The final post spinning step was done with acceleration of 2500 rpm/s for 3 s
reaching 5000 rpm minimizing edge-beads. Then, polymers were prebacked at 40 °C for 5 min and 50 °C
for 10 min respectively. After the drying step substrates were exposed using MJB4 mask aligner for 1.5
min and then developed. For HG layer development was done in diethylene glycol methyl ether (DEGMEE)
(Sigma-Aldrich Chemie GmbH, Munich, Germany) for 60 sec with washing in PGMEA for 5 sec. PI layers
were developed in the solution of solvents: 40 ml of NEP (Sigma-Aldrich Chemie GmbH), 20 ml of
DEGMEE and 10 ml of ethanol (VWR International GmbH) for 90 sec with a subsequent washing in
PGMEA for 5 sec. After the development layers were hard baked at 220°C for 10min.
(4) Metal layers:
The metal layer stack was patterned via etching process. The layer stack consisting of Ti10 nm/Cu100 nm/Ti10
nm was deposited using magnetron sputtering at room temperature. For sputtering, Ar gas at partial pressure
of 10-3 mbar and the base pressure 2.3x10-6 mbar was used. Then layer of 1 um AZ5214E photoresist
(Microchemicals GmbH, Ulm, Germany) was patterned following the protocol provided by the
manufacturer. Titanium etching process was made in the solution of sodium fluoride (NaF), ammonium
peroxydisulfate and DI water in the proportion of 1 : 1 (mol/L) in 200 mL of water. Copper etching process
was done during 5 s in the solution of HCl : H2O2 : H2O in a proportion of 1:2:10 (vol./ vol./ vol.).
(5) Self-assembly rolled-up process of tubular architectures:
The 2D layouts of polymeric structures with electronics layer were self-assembled into 3D Swiss-rolls by
a selective etching of the SL in the solution of a strong chelating agent containing 15 g of sodium
diethylenetriaminepenta-acetic acid (DETPA) (Alfa Aesar, UK) in 500 ml of DI water and a subsequent
swelling of the HG. For dissolution of the chelate in water it was neutralized with sodium hydroxide in an
amount sufficient to reach pH = 5.5. Additionally 10% of benzotriazole (Sigma-Aldrich Co. LLC.,
Germany) was introduced into the solution in order to inhibit etching of Cu layer. After the SL etching
process rolling was performed in an equivalent DETPA solution, but having pH = 8.
(6) Electrical characterization:
All prepared self-assembled structures were characterized acquiring theirs S and Z parameters using the
Cascade PM-8 probe station and vector network analyzer Agilent ENA 5071 (Agilent Technologies GmbH
& Co.KG, Waldbronn, Germany) in the frequency range from 300 kHz to 100 MHz All the electrical
connections with rolled-up strucures were made using ground signal ground (GSG) CASCADE Z-probes.
The LED experiment was performed using a signal generator MG3692B (Anritsu Corporation, Kanagawa,
Japan). To cancel any adverse effect of the interaction probe system with the magnetic field, the ESR
measurements were done in an in-house made magnet, the probe and probe holder were made of non-
magnetic materials. The electromagnet with a magnetic field in the range from 0 to 600 mT and maximum
deviation less than 2.5 µT was achieved. The magnetic field was swept using an additional set of coils with
the Keithley source meter (Tektronix UK Ltd., UK). Magnetic field was measured with the MAGSYS hall
sensor (MAGSYS magnet systeme GmbH, Dortmund, Germany), where the probe sensor was closely
placed to the sample in order to eliminate any differences in the readings of magnetic field and estimation
of the Larmor precession frequency.
Acknowledgements
We thank C. Krien and I. Fiering (IFW Dresden) for the deposition of metallic thin films. Dr. Pavel
Leksin and Dr. Stephan Fuchs for the fruitful discussions about the construction of the ESR
measuring setups. The support in the development of the experimental setups from the research
technology department of the IFW Dresden and the clean room team headed by R. Engelhard (IFW
Dresden) is greatly appreciated. This work is financed in part via the DFG research grant "Aktive
verlustarme Magnetlager hoher Steifigkeit und Präzision mit integrierter Induktionsmessung und
schneller Leistungselektronik" Nr. 221322256.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
J. Rogers, Y. Huang, O. G. Schmidt, D. H. Gracias, MRS Bull. 2016, 41, 123.
J.-H. H. Na, A. A. Evans, J. Bae, M. C. Chiappelli, C. D. Santangelo, R. J. Lang, T. C. Hull, R. C.
Hayward, Adv. Mater. 2015, 27, 79.
T. G. Leong, A. M. Zarafshar, D. H. Gracias, Small 2010, 6, 792.
G. M. Whitesides, Science (80-. ). 2002, 295, 2418.
V. Magdanz, M. Guix, F. Hebenstreit, O. G. Schmidt, Adv. Mater. 2016, 28, 4084.
C. Yoon, R. Xiao, J. Park, J. Cha, T. D. Nguyen, D. H. Gracias, Smart Mater. Struct. 2014, 23,
094008.
H.-W. Huang, M. S. Sakar, A. J. Petruska, S. Pane, B. J. Nelson, Nat Commun 2016, 7, 1.
C. Cesar, B. Bufon, J. Diego, A. Espinoza, D. J. Thurmer, M. Bauer, C. Deneke, U. Zschieschang, H.
Klauk, O. G. Schmidt, Nano 2011, 11, 3727.
[9]
J. S. Vorobyova, A. B. Vorob'Ev, V. Y. Prinz, A. I. Toropov, D. K. Maude, Nano Lett. 2015, 15, 1673.
[10] R. Sharma, C. C. B. Bufon, D. Grimm, R. Sommer, A. Wollatz, J. Schadewald, D. J. Thurmer, P. F.
Siles, M. Bauer, O. G. Schmidt, Adv. Energy Mater. 2014, 4, 1301631.
[11]
X. Yu, W. Huang, M. Li, T. M. Comberiate, S. Gong, J. E. Schutt-Aine, X. Li, Sci. Rep. 2015, 5, 1.
[12] D. Grimm, C. C. Bof Bufon, C. Deneke, P. Atkinson, D. J. Thurmer, F. Schäffel, S. Gorantla, A.
Bachmatiuk, O. G. Schmidt, Nano Lett. 2013, 13, 213.
[13] M. Medina-Sánchez, B. Ibarlucea, N. Pérez, D. D. Karnaushenko, S. M. Weiz, L. Baraban, G.
Cuniberti, O. G. Schmidt, Nano Lett. 2016, 16, 4288.
[14] M. Medina-Sánchez, O. G. Schmidt, Nature 2017, 545, 406.
[15] D. Karnaushenko, N. Münzenrieder, D. D. Karnaushenko, B. Koch, A. K. Meyer, S. Baunack, L.
Petti, G. Tröster, D. Makarov, O. G. Schmidt, Adv. Mater. 2015, 27, 6797.
[16] D. Karnaushenko, D. D. Karnaushenko, D. Makarov, S. Baunack, R. Schäfer, O. G. Schmidt, Adv.
Mater. 2015, 27, 6582.
[17] D. D. Karnaushenko, D. Karnaushenko, D. Makarov, O. G. Schmidt, NPG Asia Mater. 2015, 7,
e188.
[18] M. Z. Ahmed, M. S. Bhuyan, A. K. M. T. Islam, B. Y. Majlis, Asian J. Sci. Res. 2015, 8, 237.
[19] O. G. Schmidt, C. Deneke, S. Kiravittaya, R. Songmuang, H. Heidemeyer, Y. Nakamura, R. Zapf-
Gottwick, C. Müller, N. Y. Jin-Phillipp, IEEE J. Sel. Top. Quantum Electron. 2002, 8, 1025.
O. G. Schmidt, Ch. Deneke, Method for producing a microcoil, 2002, US 2005/0118733 A1,
US7707714B2.
A. Kleiner, Rolled-up Micro-Solenoids and Micro-Transformers, 2008, WO2008148413A1.
[20] W.-H. Chen, S. Joo, S. Sayilir, R. Willmot, T.-Y. Choi, D. Kim, J. Lu, D. Peroulis, B. Jung, IEEE J. Solid-
State Circuits 2009, 44, 2711.
[21] A. Moazenzadeh, N. Spengler, R. Lausecker, A. Rezvani, M. Mayer, J. G. Korvink, U. Wallrabe, J.
Micromechanics Microengineering 2013, 23, 114020.
[22] G. Stoychev, S. Turcaud, J. W. C. Dunlop, L. Ionov, Adv. Funct. Mater. 2013, 23, 2295.
[23] G. Stoychev, N. Puretskiy, L. Ionov, Soft Matter 2011, 7, 3277.
[24] D. Fu, F. C. Lee, S. W. S. Wang, Appl. Power Electron. Conf. Expo. APEC 2010 TwentyFifth Annu.
IEEE 2010, 940.
[25]
Y. Tang, A. Khaligh, IEEE Trans. Power Electron. 2016, 31, 340.
[26] D. L. Olson, T. L. Peck, A. G. Webb, R. L. Magin, J. V. Sweedler, Science (80-. ). 1995, 270, 1967.
[27] M. E. Lacey, R. Subramanian, D. L. Olson, A. G. Webb, J. V. Sweedler, Chem. Rev. 1999, 99, 3133.
[28] Baca-, E. Baca Solis, M. Flores Alamo, J. Anal. Bioanal. Tech. 2016, 7, 1.
Supplementary information:
"Rolled-up self-assembly of compact magnetic inductors, transformers and resonators".
Supplementary Figure 1. Planar layouts applied for fabrication of self-assembled high quality
micro coils and resonators (a) containing vertical stripes that were rolled into a set of axial coils
and (b) an LC resonator structure. Horizontal lines connect left and right vertical stripes.
Supplementary Figure 2. Measured data represents the quality factor spectra of a coil set formed
between electrodes (0) and (8) within a planar and self-assembled rolled up architectures.
Particular values and their ratios are given at 10 MHz.
Supplementary Figure 3. Measured data represents the self-inductance spectra of different coil
sets within a planar and self-assembled rolled up architectures. Particular values and their ratios
are given at 10 MHz. (a) The self-inductance of coils formed between electrodes (3) and (7). (b)
The self-inductance of coils formed between electrodes (1) and (2). (c) The self-inductance of coils
formed between electrodes (5) and (7). (d) The self-inductance of coils formed between electrodes
(3) and (5).
Supplementary Figure 4. Measured data represents the mutual inductance spectra among
different coil sets within a planar and self-assembled rolled up architecture. Particular values and
their ratios are given at 10 MHz. (a) The mutual inductance of coils formed between electrodes
(37) and (08). (b) The mutual inductance of coils formed between electrodes (12) and (08). (c) The
mutual inductance of coils formed between electrodes (57) and (08). (d) The mutual inductance of
coils formed between electrodes (35) and (08).
Supplementary Figure 5. Measured data represents the coupling coefficient spectra among
different coil sets within a planar and self-assembled rolled up architecture. Particular values and
their ratios are given at 10 MHz. (a) The coupling coefficient of coils formed between electrodes
(37) and (08). (b) The coupling coefficient of coils formed between electrodes (12) and (08). (c)
The coupling coefficient of coils formed between electrodes (57) and (08). (d) The coupling
coefficient of coils formed between electrodes (35) and (08).
Supplementary Figure 6. Schematic that shows electrical connection between power generator
transformer and the LED. The LED could be powered on the secondary side of the rolled-up
transformer structure being DC insulated from the primary side of the transformer.
|
1709.06942 | 1 | 1709 | 2017-07-26T20:25:03 | A new experimental method to study the influence of welding residual stresses on fatigue crack propagation | [
"physics.app-ph"
] | This paper presents a study on the influence of welding residual stresses (RS) on fatigue crack propagation rate (FCPR) in mode I. The objective of this work is to develop a novel methodology that allows a variation of a RS field in the studied specimen while keeping constant all other variables influencing FCPR. This led to the development of a novel specimen geometry, named CT-RES, in which RS are introduced by weld bead deposition far from the region in which fatigue crack propagation (FCP) occurs. As a consequence, the effect of factors influencing FCPR other than RS such as microstructural changes or plastic deformation, often introduced by welding processes, can be avoided. The welding RS introduced in the CT-RES specimen were determined by the contour method and the weight functions method was used to calculate the stress intensity factor (SIF), Kres, resulting from the RS as the fatigue crack propagates into the specimen. The evolution of cyclic stress ratio at the crack tip, Rlocal, was then computed from Kres to quantify the influence of RS on the cyclic stress ratio. The results show that for a given stress intensity range, dK, the FCPR of the welded geometry with fixed externally low R ratio (R = 0.1), but constantly increasing Rlocal, is the same as for the as-machined geometry without RS, solicited at high cyclic stress ratio (R = 0.7). These observations partially validate the BS7910 standard philosophy in which the remaining life of a flawed structure in presence of tensile RS is calculated from a high cyclic stress ratio (R > 0.5) crack propagation curve to eliminate crack closure effects. | physics.app-ph | physics | A new experimental method to study the influence of
welding residual stresses on fatigue crack propagation
Deschênes, P.-A.a, Lanteigne, J.b, Verreman, Y.a, Paquet, D.b,
Lévesque, J.-B.b, Brochu, M.a
aDepartment of Mechanical Engineering, École Polytechnique de Montréal, Montreal (Québec), H3T 1J4, Canada
bHydro-Québec, Institut de recherche d'Hydro-Québec, Varennes (Québec), J3X 1S1, Canada
Corresponding author : Pierre-Antony Deschênes ([email protected])
© 2017. This manuscript version is made available under the CC-BY-NC-ND 4.0 license
https://doi.org/10.1016/j.ijfatigue.2017.01.031
Abstract
This paper presents a study on the influence of welding residual stresses (RS) on fatigue crack propagation
rate (FCPR). The objective of this work is to develop a novel methodology that allows a variation of a RS
field in the studied specimen while keeping constant all other variables influencing FCPR. This led to the
development of a novel specimen geometry, named CT-RES, in which RS are introduced by weld bead
deposition far from the region in which fatigue crack propagation (FCP) occurs. As a consequence, the
effect of factors influencing FCPR other than RS such as microstructural changes or plastic deformation,
often introduced by welding processes, can be avoided. The welding RS introduced in the CT-RES
specimen were determined by the contour method and the weight functions method was used to calculate
the stress intensity factor (SIF), 𝐾(cid:3045)(cid:3032)(cid:3046) , resulting from the RS as the fatigue crack propagates into the
specimen. The evolution of cyclic stress ratio at the crack tip, Rlocal, was then computed from 𝐾(cid:3045)(cid:3032)(cid:3046) to
quantify the influence of RS on the cyclic stress ratio. The results show that for a given stress intensity
range, K, the FCPR of the welded geometry with fixed externally low R ratio (R = 0.1), but constantly
increasing Rlocal, is the same as for the as-machined geometry without RS, solicited at high cyclic stress
ratio (R = 0.7). These observations partially validate the BS7910 standard philosophy in which the
remaining life of a flawed structure in presence of tensile RS is calculated from a high cyclic stress ratio
(R ≥ 0.5) crack propagation curve to eliminate crack closure effects.
Keywords: Fatigue crack growth, residual stress, welding, weight functions, crack closure
1. Introduction
Francis runners used in the hydropower industry are composed of a crown (1), blades extending from the
crown to the band (2), and a band (3), as shown in Figure 1a. These steel components are assembled by
welding. It was demonstrated that performing a stress relief heat treatment does not completely eliminate
welding RS introduced in turbine runners during their fabrication(Lanteigne, Baillargeon, & Lalonde,
1998). Moreover, during blades repair, E309L austenitic stainless steel is often chosen as the weld material,
since using martensitic stainless steel as filler material requires a post-weld heat treatment. This type of
heat treatment is complex and expensive to do in a turbine environment, as during blade repairs, hence the
preference for E309L. However, the blade is left with a large amount of tensile residual stresses. Thus, like
1
the tensile residual stresses generated during the fabrication, those coming from blade repairs accentuate
the need for this study.
The presence of RS in structures can have either beneficial or detrimental effects on FCPR. A number of
studies have shown that compressive RS increase the fatigue life of cracked components (Beghini &
Bertini, 1990; Fleck, Smith, & Smith, 1983; Ghidini, 2007; Itoh, Suruga, & Kashiwaya, 1989; Jones,
2008)while other studies have shown that tensile RS decrease it (Liljedahl, Zanellato, Fitzpatrick, Lin, &
Edwards, 2010; Ohta, Maeda, Kosuge, Machida, & Yoshinari, 1989; Ohta, McEvily, & Suzuki, 1993;
Trudel, Sabourin, Lévesque, & Brochu, 2014). Thus, there is a direct correlation between the orientation of
the stress component normal to the propagation plane and its influence on FCPR. In those studies, RS were
introduced either by a single overload in a notched geometry(Fleck et al., 1983), by plastic deformation of
the sample(Jones, 2008), by localized heat treatment (Ohta et al., 1993)or by welding (Beghini & Bertini,
1990; Ghidini, 2007; Itoh et al., 1989; Liljedahl et al., 2010; Ohta et al., 1989; Trudel, Sabourin, et al.,
2014). In all these studies, the effect of RS on FCPR was rationalized by its influence on fatigue crack
closure, which is known to significantly affect fatigue crack propagation rates.
However, none of the above mentioned methods adequately represents the RS distribution in Francis turbine
blades or their redistribution during propagation. Indeed, the welded blades are highly constrained by a
clamping effect resulting from the rigidity of the crown and band. Furthermore, all studies investigating
crack propagation in welding RS fields were performed in regions where microstructural modifications
and/or plastic deformations were introduced by the welding process. In order to avoid those microstructural
modifications and reproduce a more representative environment, an alternate method had to be developed.
The purpose of this research work is to design a fatigue crack propagation experiment in which: (i) the
specimen used would replicate the clamping effect observed in actual Francis turbine blades; and (ii) the
FCP would occur in a region of the specimen exempt from any microstructural transformation or plastic
deformation. To the best of the authors' knowledge, such an experiment has not yet been published.
1.1 The CT-RES: A novel specimen geometry
The particular specimen geometry used in this study to characterize the influence of RS on FCPR was
developed following three constraints that were required to reproduce realistic conditions in which cracks
propagate in Francis turbine blades:
1. RS must be introduced in the fatigue test specimen by welding;
2. FCP must not be influenced by any microstructural changes induced by this welding process;
3. Weld beads must be deposited so as to avoid any plastic deformation in the region of propagation;
The second restriction was added to avoid the inevitable microstructural effects (Trudel, Lévesque, &
Brochu, 2014)when a crack propagates from the vicinity of a weld toe towards the weld deposit.
A new specimen geometry was developed in order to design a fatigue crack propagation test methodology
that respects all three listed constraints. A schematic of this sample, named the Compact Tensile Residual
Stress (CT-RES) specimen, is shown in Figure 1b. A Francis turbine is also included in the figure to
emphasize on the similarity between the two geometries: the medallion acts as a turbine blade while the
rigid frame stands for the clamping effect resulting from the rigidity of the crown and the band. This
particular specimen geometry is obtained by welding the medallion into a rigid frame, Figure 1b.
2
Figure 2 presents different views of the CT-RES specimen. The different variables defining its geometry
are also shown in Table 1. The crack length is measured with two quantities, namely a and a* (cf. Fig.2a).
The crack length is quantified using as reference the loading line (a) and the boundary of the medallion (a*)
respectively.
A specimen having the same geometry as the CT-RES was machined from the same heat treated block of
steel, therefore without RS. This sample, named CT-Monoblock, was used as a reference with which the
fatigue crack propagation rate was compared to the welded CT-RES specimen to highlight the influence of
RS on FCPR.
Figure 1 :(a) A typical Francis turbine composed of a crown (1), blades extending from the crown to the band (2), and a band (3)
(Hall, 2005), and (b) Schematic representation of the CT-RES specimen designed to reproduce the fatigue crack behavior of
Francis turbine blades. The assembly is made of a medallion (I) and a rigid frame (II)
Figure 2: a) Front view of the CT-RES specimen; and b) Side view of the CT-RES specimen showing relative thickness
dimensions
3
L[mm]
228.6
H [mm]
228.6
W [mm]
85.16
bf [mm]
25.4
bm[mm]
10.16
Table 1: CT-RES specimen dimensions
2. Materials properties and experimental set-up
2.1 Materials properties
The base metal used for the fabrication of the rigid frame and the medallion of the CT-RES specimen is
UNS S41500 martensitic stainless steel. The samples were cut from Francis turbine blade segments sent by
the manufacturer. Prior to machining, the material was heat treated at a temperature of 1050 C° during one
hour. The material was then cooled to room temperature by slightly opening the oven door. Following this
austenitization and air quenching treatment, Figure 3, the material was further tempered at a temperature of
620 C° for two hours in order to stabilize the reformed austenite and obtain a tougher martensite. Following
the heat treatment, the hardness of the steel is measured at 25 HRC. The average primary austenitic grain
size was measured at 102m by the equivalent spherical diameter method. The average percentage of
reformed austenite measured by x-ray diffraction (XRD) on three different specimens was 14.8% in
volume.
Figure 3 Microstructure following the tempering of the S41500 martensitic stainless steel.
The chemical composition of both base and weld metals is presented in Table 2. Room temperature tensile
properties of S41500 and E309L alloys were characterized according to ASTM E8M standard (ASTM,
2011). They are presented in Table 3. The S41500 tensile tests were performed in L-T direction, that is,
normal to the crack plane using small-size specimen with a 9 mm diameter and a gage length five times the
diameter. The head speed was imposed at 5 mm/min. Fracture toughness of S41500 in L-T direction was
measured by Chen and his coworkers (Chen, Verreman, Foroozmehr, & Lanteigne, 2013). They found
𝐾(cid:3010)(cid:3004) = 316 MPa√m.
4
C
Si
Mn
P
S
Cr
0.026
0.34
0.74
0.021
0.001
13.02
Mo
0.56
Ni
N2
3.91
0.031
<0.05
<0.6
0.5-1
< 0.03
<0.03
11.5-14
0.5-1
3.5-5.5
0.03
0.72
2.12
0.006
0.001
24.5
0.3
13.7
...
...
S41500
ASTM[2]
E309L
Cu
0.15
...
...
...
< 0.03
AWS[4]
Table 2: As-received and nominal composition limits (weight %) of: (i) S41500 alloy used to machine the rigid frame and
medallion of the CT-RES specimen; and (ii) 309L alloy used as the filler metal for welding the medallion to the rigid frame
0.3-0.65
1.0-2.5
< 0.75
12-14
< 0.03
< 0.03
23-25
...
E (GPa)
𝝈𝒚𝒔 (MPa)
𝝈𝒖𝒍𝒕.(MPa) A (%)
S41500
195
660
822
32
63
E309L
Table 3: Room temperature tensile properties of S41500 and 309L alloys
358
644
200
2.2 Fabrication of the CT-RES specimen
The fabrication of the CT-RES specimen used the robotic arm Scompi® developed at Hydro-Québec
Research Institute (Hydro-Québec, 2011). The actual setup designed to weld the CT-RES specimen is
shown in Figure 4. Two torches, one on each side of the specimen, were fixed on a rigid assembly while
the robotic arm moved the specimen during the welding process. It is believed that simultaneous deposition
of beads on each side of the specimen would significantly reduce distortions and lead to a more symmetric
residual stress field in the specimen.
Flux-cored arc welding (FCAW) was used to perform weld deposition between the CT-RES medallion and
its rigid frame. The protection gas was constituted of 75% argon and 25% carbon dioxide. A single bead
on each side of the medallion, top and bottom, was sufficient to assemble the CT-RES sample. That is a
total of four beads. The welding parameters are present in Table 4.
5
Figure 4: Setup used to weld the CT-RES specimen: (A) Scompi® Robotic Arm; (B) CT-RES specimen; (C) welding torches;
and (D) rigid assembly
Voltage
(V)
26
Current
intensity
Wire feed
speed
(mm/s)
Preheat
Interpass
temperature
temperature
(C°)
(A)
100
170
Table 4: Welding parameters used to assemble the CT-RES medallion to the frame
132
804
(C°)
100-115
Travel
speed
(mm/s)
5.5
Deposition
rate
(kg/h)
3.3
Linear
energy
(J/mm)
During the process, particular attention was given so that the temperature reached in the medallion did not
exceed 630 °C at weld interpass. This was done to avoid the austenite-martensite ( 𝛾-𝛼(cid:4593) ) phase
transformation that occurs in this alloy at 300 °C upon cooling. This phase transformation results in volume
increase of approximately 3% that could significantly alter the final residual stress field in the medallion
after the assembly. This expansion is clearly visible on a Thermo-Mechanical Analyzer (TMA) curve
obtained in a previous study (Lanteigne, Sabourin, Bui-Quoc, & Julien, 2008). Maximum temperature in
the medallion never exceeded 350 °C, so that the austenite-martensite transformation was avoided.
2.3 Fatigue crack propagation test
The characterization of crack propagation rate was done according to the ASTM E647 standard (ASTM,
2013). Using a 250 kN servo-hydraulic test machine, tests were performed at a fixed frequency of 6 Hz.
Prior to the FCP test, an initial stress intensity factor range of ∆ K = 11 MPa √m was imposed for
pre-cracking the samples. The pre-cracking stage was maintained until the crack lengths observed with
microscope on each side of the specimen were comparable to within approximately 200 m.
During the FCP test, the crack length was calculated by the compliance method. For this purpose, a crack
mouth opening displacement (CMOD) extensometer was mounted on the medallion. The crack lengths
calculated with this method were frequently compared with optical measurements to ensure their validity.
Crack length data was automatically sampled at each 250 m crack advance and the FCPR was then
calculated using the secant method defined by the following equation:
6
(1.)
𝑑𝑎
𝑑𝑁(cid:3028)
=
(𝑎(cid:3036)(cid:2878)(cid:2869) − 𝑎(cid:3036))
(𝑁(cid:3036)(cid:2878)(cid:2869) − 𝑁(cid:3036))
As for the ∆K, it was calculated using the average crack length of the 𝑑𝑎 interval. Fatigue crack closure
was measured using the P-v curves. To achieve this, a 4% deviation criterion based on a linear regression
fitted to the linear segment (i.e. fully open crack) of the P-v diagram was used to determine the closure
load, Pcl, as show in Figure 12.
To delineate the influence of welding RS on the fatigue crack propagation behavior of the CT-RES
specimen, the experimental plan shown in Table 5 is undertaken. The choice made for the R ratios used to
perform the FCP test in the CT-Monoblock specimen, i.e. without RS, is based on the actual loading
condition present in a turbine runner. In fact, there are two distinct load signatures. As the production
begins, the stress measured in the blades rapidly increases to a steady-state value. This load signature is
characterized by a low cyclic stress ratio (R = 0.1). When the wicket gates are fully open, the hydraulic
dynamic load is combined to the static load from which a high cyclic stress ratio (R = 0.7) is generated
(Sabourin, Bouffard, & Paquet, 2007).
Specimen
CT-RES
CT-
Monoblock
ΔK
[MPa√𝐦]
R
Residual
stresses
Kmax
[MPa√𝐦]
15
15
K
K
min
max
K
K
res
res
)(
a
)(
a
Yes
Kmax=
0.1/0.7
No
Kmax=
K
)(
a
res
)
K
1(
R
K
R
1(
)
Table 5: Experimental plan to highlight the effect of residual stresses on FCP in the CT-RES specimen
3. Computation of stress intensity factor and compliance
The following equation of linear elastic fracture mechanics (LEFM) applicable to standard CT-specimen
geometry is used to correlate the load P to the stress intensity factor :
𝐾 =
𝑃
𝑏(cid:3040)√𝑊
𝑌(𝑎/𝑊)
(2.)
in which the dimensionless solution of K, 𝑌(𝑎/𝑊), is a polynomial expansion.
When this equation is applied to the CT-RES specimen, 𝑃 represents the test force, 𝑏(cid:3040) the thickness of the
medallion, 𝑊 the distance between the loading line and the end of the medallion (cf. Fig. 2a.). The
geometrical factor 𝑌(𝑎/𝑊) had to be calibrated numerically by the finite element method (FEM). To
ensure an adequate calibration of the CT-RES geometry, three methods are used to calculate its SIF: (i) the
J-Integral method (Rice, 1967); (ii) the displacement extrapolation at the crack tip; and (iii) the equation of
an edge crack in a semi-infinite sheet. It can be noted that methods (i) and (ii) are available in the finite
element (FE) software Ansys®. The FE model used represents a quarter of the entire geometry. Symmetry
boundary conditions were applied on the crack propagation plane and on the mid-thickness plane.
The implementation of the first method (J-Integral) requires multiple calculations. In each of these
calculations, a crack of length 𝑎 is introduced by changing the boundary conditions on the crack plane. The
mesh surrounding the crack tip is constituted of 20-nodes quadratic hexahedral elements. The convergence
of the method is verified over 10 paths to insure a good validation.
7
In the second technique, the mesh is modified to respect the square root distribution of displacements near
the crack tip. Thus, the second node of each element defining the crack front is moved at 𝐿/4 from the
crack tip, as proposed by Barsoum (Barsoum, 1977). The parameter 𝐿 is the element length.
The third method uses the analytical equation giving the stress intensity factor of an edge crack in a semi-
infinite sheet to calculate K for mechanically short cracks (a/W 0.5) in the CT-RES specimen. Hence,
𝐾 = 1.1215 𝜎(cid:3041)(cid:3042)(cid:3040)√𝜋𝑎∗
(3.)
where𝜎(cid:3041)(cid:3042)(cid:3040) represents the stress component 𝜎(cid:3052) on the medallion side of the un-cracked CT-RES geometry
subjected to the same load 𝑃 used to calibrate the cracked geometry.
Results obtained with the three methods are presented in Figure 5.
Figure 5: Dimensionless solution of K computed for the CT-RES and CT-Monoblock specimens with: the J-Integral method; the
displacement extrapolation technique; and the analytical equation of a crack edge in a semi-infinite sheet.
Figure 5 show a good agreement between the three different methods. The analytical equation,
corresponding to the dashed line of the figure, is valid for crack lengths up to a/W= 0.52. For convenience,
the function Y (a/W)calculated in this study was obtained from a polynomial fit of the J-Integral results.
The polynomial fit equation is defined as:
(cid:2874)
𝑌(𝑎/𝑊) = (cid:3533) 𝐶(cid:3036) ∙ (𝑎/𝑊)(cid:3036)
(cid:3036)(cid:2880)(cid:2868)
(4.)
The CT-RES specimen compliance has also been calibrated by the FEM. Crack mouth opening
displacements were calculated for a series of crack lengths obtained by incrementing successively a/W by
a value of 0.02. The dimensionless compliance values, 𝑣(cid:3051), were then calculated using the equation proposed
by Saxena (Saxena & Hudak, 1978):
8
𝑣(cid:3051) = (cid:4688)(cid:4684)(cid:3496)
𝐸𝑣𝑏(cid:3040)
𝑃
(cid:2879)(cid:2869)
(cid:4685) + 1(cid:4689)
(5.)
where𝑣corresponds to the crack mouth opening displacement and 𝐸 the elastic modulus. The data obtained
from this method is then smoothed using a polynomial regression, the results of which are used to calculate
the crack length during the FCP test. The polynomial function best representing the compliance behavior
of the CT-RES specimen is shown in Figure 6 and is defined as:
(cid:2874)
(𝑎/𝑊) = (cid:3533) 𝐶(cid:3036) ∙ 𝑣(cid:3051)
(cid:3036)
(cid:3036)(cid:2880)(cid:2868)
(6.)
Figure 6: Dimensionless compliance values computed from FE analysis for the CT-RES and CT-Monoblock specimens
It is noteworthy that the difference between visual measurements and crack lengths calculated with the
compliance method never exceeds 0.5 mm. Since the initial crack length is 37.5 mm, it follows that the
maximum error introduced by the compliance method is approximately 1.3% on the crack length. The
precision of the compliance method is thus sufficient to avoid significant errors on the applied loads.
4. Residual stresses in the CT-RES specimen
4.1 Measurement of the residual stress field
The residual stress field introduced in the CT-RES specimen after welding was measured by the contour
method developed by Prime (Prime, 2001). This technique allows the determination of the distribution of
the normal component of an unknown residual stress field by cutting the specimen along a plane
perpendicular to that stress component (Levesque, 2015).
(cid:3045)(cid:3032)(cid:3046) that is
This method is used herein to calculate the distribution of the residual stress component 𝜎(cid:3052)
perpendicular to the crack propagation plane. To begin, the CT-RES specimen is cut along the crack
9
propagation plane by electrical discharge machining (EDM). Then, an optical profilometer is used to
measure the relaxed normal displacements along the two freshly created surfaces. Finally, the average
displacement map is imposed as boundary conditions on a FE model of the CT-RES specimen to compute
the corresponding residual stress distribution.
However, modification of the original displacement maps obtained with the profilometer is required before
calculating the residual stress field. First, there is an abrupt increase in displacement at the very end of the
cut, seemingly caused by plastic deformation. The displacements in this plastically deformed region were
replaced by a 3D linear extrapolation from the rest of the displacement distribution. Second, the results
obtained near the edges (1 mm) were discarded because it is well known that they are inaccurate
(Hosseinzadeh, Ledgard, & Bouchard, 2012; Sarafan, Lévesque, Wanjara , Gholipour, & Champliaud,
2015). It is also important to note that the contour method is a destructive procedure. As a matter of fact,
two CT-RES specimens were fabricated. The first one was used for the measurement of the residual stress
field in the medallion, while the second was used for the FCP test. Therefore, the analysis relies on the
hypothesis that both samples have identical residual stress fields. Figure 7 illustrates a distribution of the
(cid:3045)(cid:3032)(cid:3046) at mid-thickness of the medallion and calculated by the contour method.
residual stress component 𝜎(cid:3052)
The oscillations are caused by the smoothed displacement map obtained from cubic splines. The variation
of stress across the thickness is not significant.
Figure 7: Mid-thickness distribution of 𝜎(cid:3052)
(cid:3045)(cid:3032)(cid:3046) normal to the crack propagation plane in the medallion of the CT-RES specimen.
It can be seen in Figure 7 that no residual stresses are captured for the first 5 mm due to the pre-cracking of
the specimen. This pre-crack also led to a significant increase of the residual stress ahead of the crack tip
as observed by Liljedahl et al.,(Liljedahl et al., 2010)and Trudel et al. (Trudel, Sabourin, et al., 2014).
The results in Figure 7 also show a continuous increase in the tensile residual stresses along the medallion
width (x'). This is different from the residual stress profiles in self-equilibrated specimens. This particular
feature of the CT-RES makes it a unique specimen for studying the influence of RS on FCPR.
To get an idea of the behavior of the residual stress field redistribution as the crack propagates in the
medallion, a finite element modeling was performed to recreate thermal contraction induced by the welding
of the medallion in the rigid frame. In the simulation, elements within the weld deposit were contracted
artificially. The induced stress in the uncracked planed, 𝜎(cid:3033), was computed as the sum of all reaction forces,
10
𝑓(cid:3036), normal to the crack plane divided by the area of the remaining ligament, 𝐴. This simple equation is
defined as followed:
𝜎(cid:3033) =
Σ𝑓(cid:3036)
𝐴
(7.)
Figure 8 presents the evolution of the normalized mean stress in the remaining ligament as the crack
propagates. We can see that the mean stress constantly increases as the remaining ligament of the medallion
(cid:3045)(cid:3032)(cid:3046) redistribution and thus, the
becomes shorter. That short analysis informs us on the behavior of the 𝜎(cid:3052)
evolution of 𝐾(cid:3045)(cid:3032)(cid:3046) in the CT-RES specimen. It is important to specify that this simulation does not recreate
the real conditions of the welding procedure but therefore give a tendency of the redistribution of the RS in
the CT-RES medallion as the crack propagates.
Figure 8: Evolution of the mean stress 𝜎(cid:3033) in the remaining ligament of the medallion as the crack propagates
Using this novel specimen, our experimental strategy is to perform crack propagation tests at constant ∆K
and R ratio to isolate the effect of residual stresses on crack growth behavior. By maintaining constant ∆K
during a fatigue crack propagation test, 𝜎(cid:3045)(cid:3032)(cid:3046) remains the only varying parameter, hence the influence of RS
on fatigue crack propagation can be studied while avoiding the effect of any other parameter variation.
4.2 Computation of 𝑲𝒓𝒆𝒔with the weight functions method
The variation of 𝐾(cid:3045)(cid:3032)(cid:3046)with the advance of the crack in the CT-RES specimen was analyzed using the weight
functions method developed by Bueckner (Bueckner, 1970). This method allows the calculation of the SIF
resulting from any residual stresses distribution and can be used with a wide variety of specimen geometries.
4.2.1 Weight function theory
Rice (Rice, 1972) proposed a simplified expression to calculate the stress intensity factor, 𝐾(cid:3010),(cid:3010)(cid:3010), in a 2D
cracked structure submitted to surface traction 𝒕 on a boundary Γ and submitted to body forces 𝒇:
𝐾(cid:3010),(cid:3010)(cid:3010) = (cid:3515) 𝒕 ∙ 𝒉 𝑑Γ + (cid:3505) 𝒇 ∙ 𝒉 𝑑𝑉
(8.)
11
where 𝒉, the weight function, is defined as:
𝒉(𝑥, 𝑦, 𝑎) =
𝐸
𝐾
𝜕𝒖(𝑥, 𝑦, 𝑎)
𝜕𝑎
(9.)
The second term of the Eq.8 is discarded in this work because the CT-RES specimen is not submitted to
body forces. Moreover, the two-dimensional function 𝒉 is reduced to a one dimension function, ℎ(𝑥, 𝑎),
since the calibration of the method was done only on the 𝑢(cid:3052) component (𝐾(cid:3010)) of the global displacement
field, 𝒖. The final form of the equation can be defined as:
(cid:3028)
𝐾(cid:3010) = (cid:3505) 𝜎(cid:3052)
(cid:3045)(cid:3032)(cid:3046)(𝑥)
ℎ(𝑥, 𝑎) 𝑑𝑥
(cid:3028)(cid:3116)
(9.)
The work of Glinka & Shen (Glinka & Shen, 1991)was exploited in order to fit the displacement derivative
𝜕𝑢(cid:3052)/𝜕𝑎 with the universal equation developed by the authors, Eq.10, composed of 𝑀(cid:3036) unknown
coefficients. For a deep edge crack, which is assumed in the case of the CT-RES specimen, the work of Fett
et al. (Fett, Mattheck, & Munz, 1987)also contributes to the simplification of the calibration since 𝑀(cid:2870) = 3.
Hence, the universal function is written as:
ℎ(𝑥, 𝑎) =
2
(cid:3493)2𝜋(𝑎 − 𝑥)
(cid:3429)1 + 𝑀(cid:2869) (cid:4672)1 −
(cid:3117)
(cid:3118)
𝑥
𝑎
(cid:4673)
+ 3 (cid:4672)1 −
𝑥
𝑎
(cid:4673) + 𝑀(cid:2871) (cid:4672)1 −
(cid:3119)
(cid:3118)
𝑥
𝑎
(cid:4673)
(cid:3433)
(10.)
4.2.2 Weight function calibration
To perform the calculation of 𝑀(cid:3036), two linearly independent stress distributions must be applied on the
CT-RES specimen boundary as proposed by Shen & Glinka (Shen & Glinka, 1991). They are chosen as:
(i) a constant stress distribution; and (ii) a linearly increasing stress distribution from the load line to the
end of the frame. These two reference stress distributions were applied to the CT-RES rigid frame in the
FE model.
Following the calculation, a validation was executed to ensure the accuracy on the calculation of 𝐾(cid:3045)(cid:3032)(cid:3046). To
do so, a volumetric contraction was imposed on the weld bead elements (cf. Fig.2b) of the FE model. The
J-Integral method was then used to determine K for various crack lengths and the results were compared to
those obtained by the weight functions method. Both methods showed a very good agreement. Indeed, the
largest difference between the SIF calculated by the weight functions method and the one obtained with the
J-Integral method was less than 0.1%.
4.2.2 𝑲𝒓𝒆𝒔 calculation performed by the weight function method
Once validated using known Kres distributions, the weight functions method can be used to evaluate the
stress intensity factor induced by the CT-RES residual stress field for different crack lengths. The results
are shown in Figure 9. The first point in Figure 9 (the black dot) had to be corrected since the value was
too low(𝐾(cid:3045)(cid:3032)(cid:3046) = 8MPa√𝑚)compared to neighboring values. The value was extrapolated from the two
following values. This is probably caused by a lack a precision from the calculation of the stress near the
crack tip in the FE model.
12
Figure 9: Computed 𝐾(cid:3045)(cid:3032)(cid:3046) from the weight functions method as function of crack advance in the medallion
Figure 8 and Figure 9 respectively show an increase of the residual stress and the residual stress intensity
factor 𝐾(cid:3045)(cid:3032)(cid:3046) as the crack progresses into the residual stress field, which is an expected behavior.
5. Experimental results
5.1 Testing the CT-RES: FCP in a residual stress field
Figure 10 shows the evolution of the minimum and maximum stress intensity factors calculated using the
weight functions method plotted against the crack length. The values 𝐾(cid:3040)(cid:3036)(cid:3041)(cid:2879)(cid:3028)(cid:3043)(cid:3043) and 𝐾(cid:3040)(cid:3028)(cid:3051)(cid:2879)(cid:3028)(cid:3043)(cid:3043) represent
the minimum and maximum applied SIF by the servo-hydraulic bench, known as the external load. The
gray band corresponds to the external solicitation i.e ∆K. As mentioned before, the ∆K was kept constant
during the test so as to highlight the potential effect of the RS field on the FCPR in the CT-RES sample.
Consequently, the variation of 𝐾(cid:3045)(cid:3032)(cid:3046) along the medallion modifies the minimum and maximum SIFs, i.e.
𝐾(cid:3040)(cid:3036)(cid:3041) and 𝐾(cid:3040)(cid:3028)(cid:3051), while not affecting the difference between the two. This results in a constant ∆K fatigue
crack propagation test in which the local cyclic stress ratio (𝑅(cid:3039)(cid:3042)(cid:3030)(cid:3028)(cid:3039)) evolves with the crack length from R =
0.55 to R = 0.8 as shown in Figure 10.
13
Figure 10: Plot of the variation of SIF in the CT-RES specimen during an FCP test at constant applied ∆𝐾 = 15 MPa√m
Figure 11 shows the comparison between fatigue crack propagation rates in the CT-RES and
CT-Monoblock specimens at ∆K = 15 MPa√m. The dashed lines correspond to punctual values of FCPR
extracted from da/dN - ∆K curves (R=0.1 and R=0.7) performed at increasing ∆K in the Paris regime.
Figure 11: Crack propagation rates in the CT-RES and CT-Monoblock specimens when solicited at ∆𝐾 = 15 MPa√m
As seen in Figure 11, even with an increase of the cyclic stress ratio at the crack tip, 𝑅(cid:3039)(cid:3042)(cid:3030)(cid:3028)(cid:3039), induced by the
residual stress field, the FCPR remains constant during the test. The average rate of crack propagation in
both the CT-Monoblock and CT-RES specimen are listed in Table 6.
Specimen
CT-Monoblock
CT-Monoblock
CT-RES
Applied R ratio
da/dN[mm/cycle]
0.1
0.7
0.1
2.33E-06
1.56E-05
1.72E-05
Table 6: Fatigue crack propagation rate in CT-RES and CT-Monoblock specimens at ∆𝐾 = 15 MPa√m
14
During the FCP test performed on the welded CT-RES specimen, no crack closure was detected as
presented by the P-v results A and B plotted on Figure 12a. Series A and B were respectively acquired
during the last (32 mm) and the first (6 mm) crack propagation cycles. It is apparent that the tensile RS in
the specimen opened the crack, eliminating crack closure. The closure free behavior was confirmed by the
perfect superposition of a straight line on the two sets of results. On the other hand, the P-v curve C,
corresponding to the data recorded during the test performed at R = 0.1 in the CT-Monoblock specimen,
presents crack closure. The arrow on Figure 12b indicates the crack closure load, Pcl.
a)
b)
Figure 12: (a) The first (A) and last (B) P-v curves monitored in the welded CT-RES specimen, as well as the P-v curve (C)
monitored in the CT-Monoblock specimen at ∆K= 15 MPa√m and R = 0.1, (b) the curves B and C zoomed to highlight the crack
closure load Pcl
In opposition, the P-v results acquired for the CT-Monoblock specimen tested at R = 0.7 showed no
evidence of crack closure. It is therefore expected that the FCPR behavior of the CT-RES specimen will be
closer to the FCPR of the CT-Monoblock tested at R=0.7, as shown by the results in Table 6. Nevertheless,
a small difference persists between these two behaviors. The average fatigue crack growth rate in the CT-
15
RES is approximately 10% higher than that of the CT-Monoblock specimen tested at R = 0.7. Even if this
difference could be attributed to result dispersion, (Virkler, Hillberry, & Goel, 1979), another hypothesis
points towards the effect of RS on 𝐾(cid:3040)(cid:3028)(cid:3051). For a given ∆K, 𝐾(cid:3040)(cid:3028)(cid:3051) is known to influence the fatigue crack
growth rate in certain materials. For example, aluminum alloys are more sensitive to an increase in 𝐾(cid:3040)(cid:3028)(cid:3051),
than mild steels or stainless steels (Koçak, Webster, Janosch, Ainsworth, & Koers, 2006).In fact, the low
values of 𝐾(cid:3010)(cid:3004) in aluminum alloys is penalizing since a low value of 𝐾(cid:3040)(cid:3028)(cid:3051) could triggered static damage
mechanics and increase the FCPR. It is thus proposed to briefly analyze our material sensitivity to 𝐾(cid:3040)(cid:3028)(cid:3051),
using the Nasgro model.
In the NASGRO model(Forman, R. G. & Mettu, 1992), the sensitivity of an alloy to 𝐾(cid:3040)(cid:3028)(cid:3051) is quantified by
an exponent q. The NASGRO equation giving the crack growth rate as a function of the applied SIF range
K, maximum SIF 𝐾(cid:3040)(cid:3028)(cid:3051), cyclic stress ratio R and the effective stress ratio f, function of R, as defined by
Newman 6 is:
𝑑𝑎
𝑑𝑁
= 𝐶 (cid:3428)(cid:3436)
1 − 𝑓
1 − 𝑅
(cid:3440) ∆𝐾(cid:3432)
(cid:4672)1 −
(cid:3012)(cid:3288)(cid:3276)(cid:3299)
(cid:3012)(cid:3258)(cid:3252)
(cid:4673)
(cid:3041) (cid:4672)1 −
∆(cid:3012)(cid:3295)(cid:3283)
∆(cid:3012)
(cid:4673)
(cid:3043)
(cid:3044)
(11.)
In this equation, parameters 𝐶, 𝑛, p, and q are material constants that must be determined experimentally.
For FCP tests performed with a positive R, the parameter q is equal to 1 for the majority of aluminum alloys
and 0.25 for the majority of stainless steels 7(Forman, R.G. , Shivakumar, Cardinal, William, &
McKeighan, 2005; Koçak et al., 2006).
The NASGRO model is used herein to estimate the influence of 𝐾(cid:3040)(cid:3028)(cid:3051) on the FCPR of the CT-RES
specimen. Since the ratio containing both R and f is equal to 1 during the FCP test (crack fully open), the
analysis is limited to the term quantifying the sensitivity of this alloy to 𝐾(cid:3040)(cid:3028)(cid:3051). The coefficients p and q
were assumed to be equal to 0.25 due to the lack of data for SS415 stainless steel in the literature.
To highlight the influence of 𝐾(cid:3040)(cid:3028)(cid:3051) on the FCPR, the starting and ending values of 𝐾(cid:3040)(cid:3028)(cid:3051) 𝐾(cid:3010)(cid:3004)⁄
during FCP in the CT-RES specimen are presented in Table 7.
ratio obtained
𝐾(cid:3040)(cid:3028)(cid:3051) 𝐾(cid:3010)(cid:3004)⁄
(cid:3436)1 −
(cid:2868).(cid:2870)(cid:2873)
(cid:3440)
𝐾(cid:3040)(cid:3028)(cid:3051)
𝐾(cid:3010)(cid:3004)
Starting
0.101
0.973
Ending
.190
0.948
Difference
+ 46 %
- 2.57 %
Table 7: Evaluation of the effect of 𝐾(cid:3040)(cid:3028)(cid:3051) 𝐾(cid:3010)(cid:3004)⁄
ratio on the predicted FCPR in CT-RES specimen by the NASGRO model
As we can see from Table 7, a significant increase of the ratio 𝐾(cid:3040)(cid:3028)(cid:3051) 𝐾(cid:3010)(cid:3004)⁄
generates a slight decrease of the
denominator in the NASGRO model, Eq.11. Such a small difference is negligible relative to the results
dispersion.
This entails that stainless steels suffer no particular effect from residual stresses, other than a R effect, with
respect to FCPR for the tested loading conditions. Consequently, a tensile residual stress field mainly
influences the FCPR by preventing crack closure as would high stress ratio do. These observations partially
16
validate the philosophy of the BS7910 standard 8 (BSI, 2013)recommending that the remaining life of a
cracked structured subjected to tensile RS should be calculated from propagation curves characterizing fully
opened cracks (R ≥ 0.5). However, this philosophy could lead to significant underestimation of the
remaining life in the case of partial or complete closure of the crack. Thus, the design optimization and
improvement in fatigue life prediction of structures facing dynamic load will inevitably require broad
knowledge of both metallurgical and mechanical notions. To name a few, a thorough comprehension of the
effect of phase transformation in materials, the calculation of the residual stress field in assembled structures
by multiphysics FE simulation and the effect of multiaxial, proportional or non-proportional, loading on
the damage nucleation and crack propagation.
6. Conclusion
In this study, a novel specimen geometry was developed to study the influence of welding residual stresses
RS on fatigue crack propagation in SS415, while avoiding the influence of other factors such as
microstructural changes or plastic deformations. The CT-RES specimen was manufactured by welding a
medallion in a rigid frame. Both the medallion and rigid frame were composed of SS415 stainless steel.
The resulting residual stress field is believed to be very similar to the RS field observed in turbine runners
after fabrication. The crack propagation rate measured in the welded specimen is nearly identical to the
crack propagation rate observed in the CT-Monoblock specimen, i.e. without RS, at an applied R ratio
of 0.7. From the findings presented in this study, the following conclusions may be drawn:
The CT-RES specimen can be used to study the influence of welding residual stress on FCPR;
The CT-RES can be used for testing different welding procedures as well as the effect of the
following post-weld treatment on the RS level;
The fabrication of the CT-RES specimen does not alter the material microstructure in the FCP
region of the specimen, or its consolidation state. In other words, this sample can be used to isolate
the effect of welding RS on fatigue crack propagation behavior;
The absence of crack closure accounts for most of the crack propagation rate increase observed in
the welded CT-RES specimen. This phenomenon is directly related to the presence of tensile
residual stresses resulting from the welding process;
The redistribution of the residual stress field in the constrained medallion has no effect on
propagation as long as the tensile residual stress magnitude is high enough to fully open the crack
lips. As a consequence, crack propagation rates measured in the CT-RES and CT-Monoblock
specimens solicited at R = 0.7 are very similar.
Acknowledgments
The authors gratefully acknowledge CReFaRRE, General Electric, Hydro-Québec, CRSNG, MITACS for
their financial support. P.-A. Deschênes would also like to acknowledge Carlo Baillargeon,
Manon Provencher, René Dubois, Yvan Laroche and Jean-Benoît Hudon for their technical help during the
project.
References
17
y
6Newman, J.C.84A crack opening stress equation for fatigue crack growthInternational Journal of
Fatigue, 24, 131-135.78
ASTM. (2011). Standard Test Method for Tension Testing of Metallic Materials. E8/E8M-11. West
Conshohocken, PA: ASTM International. doi:10.1520/E0008_E0008M-11
ASTM. (2012). Standard specification for Chromium and Chromium-Nickel Stainless Steel Plate, Sheet,
and Strip for Pressure Vessels and for General Applications A240/A240M-16. West
Conshohocken, PA: ASTM International. doi:10.1520/A0240_A0240M-16
ASTM. (2013). Standard Test Method for Measurement of Fatigue Crack Growth Rates. E647-13a. West
Conshohocken: ASTM International. doi:10.1520/E0647-13A
AWS. (2006). Specifications for Bare Stainless Steel Welding Electrodes and Rods (A5.9/A5.9M).
Barsoum, R. S. (1977). Triangular quarter-point elements as elastic and perfectly-plastic crack tip
elements. International Journal for Numerical Methods in Engineering, 11(1), 85-98.
doi:10.1002/nme.1620110109
Beghini, M., & Bertini, L. (1990). Fatigue crack propagation through residual stress fields with closure
phenomena. Engineering Fracture Mechanics, 36(3), 379-387.
BSI. (2013). Guide to methods for assessing the acceptability of flaws in metallic structures. BS 7910.
Bueckner, H. F. (1970). A novel principle for the compution on stress intensity factors. Journal of
Applied Mathematics and Mechanics, 50(9), 529-546.
Chen, J., Verreman, Y., Foroozmehr, F., & Lanteigne, J. (2013). Fracture toughness of 13Cr4NiMo high-
strengh steels used in hydraulic turbine runners. Communication présentée à Materials Science
and Technology, Montréal, Québec, Canada.
Fett, T., Mattheck, C., & Munz, D. (1987). On the calculation of crack opening displacement from the
stress intensity factor. Engineering Fracture Mechanics, 27(6), 697-715. doi:10.1016/0013-
7944(87)90159-7
Fleck, N. A., Smith, I. F. C., & Smith, R. A. (1983). Closure Behaviour of Surface Cracks. Fatigue &
Fracture of Engineering Materials and Structures, 6(3), 225-239. doi:10.1111/j.1460-
2695.1983.tb00339.x
Forman, R. G., & Mettu, S. R. (1992). Behavior of surface and corner cracks subjected to tensile and
bending loads in Ti–6Al–4V allo. Fracture mechanics: Twenty-Second Symposium, I(1131), 519-
546.
Forman, R. G., Shivakumar, V., Cardinal, J. W., William, L. C., & McKeighan, P. C. (2005). Fatigue
crack growth database for damage tolerance analysis.
Ghidini, T., Donne, C. D. (2007). Fatigue crack propagation assessment based on residual stresses
obtained through cut-compliance technique. Fatigue & Fracture of Engineering Materials and
Structures, 30(3), 214-222. doi:10.1111/j.1460-2695.2006.01059.x
Glinka, G., & Shen, G. (1991). Universal features of weight functions for cracks in mode I. Engineering
Fracture Mechanics, 40(6), 1135-1146. doi:10.1016/0013-7944(91)90177-3
Hall, R. J. (2005). Turbine runner glen canyon. Tiré de
https://commons.wikimedia.org/wiki/File:Turbine_runner_glen_canyon.jpg
Hosseinzadeh, F., Ledgard, P., & Bouchard, P. J. (2012). Controlling the Cut in Contour Residual Stress
Measurements of Electron Beam Welded Ti-6Al-4V Alloy Plates. Experimental Mechanics,
53(5), 829-839. doi:10.1007/s11340-012-9686-1
Hydro-Québec. (2011). Scompi Robot. Tiré de
http://www.hydroquebec.com/innovation/en/pdf/2010G080-18A-SCOMPI.pdf
Itoh, Y. Z., Suruga, S., & Kashiwaya, H. (1989). Prediction of fatigue crack growth rate in welding
residual stress field. Engineering Fracture Mechanics, 33(3), 397-407.
Jones, K. W., Dunn, M. L. (2008). Fatigue crack growth through a residual stress field introduced by
plastic beam bending. Fatigue & Fracture of Engineering Materials & Structures, 31(10), 863-
875. doi:10.1111/j.1460-2695.2008.01274.x
18
Koçak, M., Webster, S., Janosch, J. J., Ainsworth, R. A., & Koers, R. (2006). FITNET Fitness-for-Service
(FSS) procedure.
Lanteigne, J., Baillargeon, C., & Lalonde, S. L., R. (1998). Analyse d'un prélèvement de la roue #10 de la
centrale de Beauharnois : Évaluation sommaire de la cause de la fissuration. Varennes, Qc:
Institut de recherche d'Hydro-Québec.
Lanteigne, J., Sabourin, M., Bui-Quoc, T., & Julien, D. (2008). The characteristics of the steels used in
hydraulic turbine runers. Communication présentée à 24th Symposium on Hyraulic Machinery
and Systems, Foz Do Iguassu, Brazil.
Levesque, J.-B. (2015). Programme de calcul des contraintes résiduelles par la méthode du contour.
Varennes, Canada: Institut de Recherche d'Hydro-Québec (IREQ).
Liljedahl, C. D. M., Zanellato, O., Fitzpatrick, M. E., Lin, J., & Edwards, L. (2010). The effect of weld
residual stresses and their re-distribution with crack growth during fatigue under constant
amplitude loading. International Journal of Fatigue, 32(4), 735-743.
doi:10.1016/j.ijfatigue.2009.10.012
Ohta, A., Maeda, Y., Kosuge, M., Machida, S., & Yoshinari, H. (1989). Fatigue crack propagation curve
for design of welded structures. Transaction of the Japan Welding Society, 20(1), 17-23.
Ohta, A., McEvily, A. J., & Suzuki, N. (1993). Fatigue crack propagation in a tensile residual stress field
under a two-step programmed test. International Journal of Fatigue, 15(1), 9-12.
Prime, M. B. (2001). Cross-Sectional Mapping of Residual Stresses by Measuring the Surface Contour
After a Cut. Journal of Engineering Materials and Technology, 123(2), 162.
doi:10.1115/1.1345526
Rice, J. R. (1967). A Path Independent Integral and the Approximate Analysis of Strain Concentration by
Notches and Cracks. Journal of Applied Mechanics, 35, 379-386.
Rice, J. R. (1972). Some remarks on elastic crack-tip elastic stres fields International Journal of Solids
and Structures, 8, 751-758.
Sabourin, M., Bouffard, D. A., & Paquet, F. (2007). Life prediction of hydraulic runners using fracture
mechanics. Communication présentée à Waterpower XV, Chattanooga, Tennessee.
Sarafan, S., Lévesque, J.-B., Wanjara , P., Gholipour, J., & Champliaud, H. (2015). Distortion and
Residual Stresses in Electron Beam Welded Hydroelectric Turbine Materials. Science and
Technology of Welding and Joining.
Saxena, A., & Hudak, S. J. (1978). Review and extension of compliance information for common crack
growth specimens. International Journal of Fracture, 14(5), 453-468. doi:10.1007/bf01390468
Shen, G., & Glinka, G. (1991). Determination of weight functions from reference stress intensity factors.
Theoretical and Applied Fracture Mechanics, 15, 237-245.
Trudel, A., Lévesque, M., & Brochu, M. (2014). Microstructural effects on the fatigue crack growth
resistance of a stainless steel CA6NM weld. Engineering Fracture Mechanics, 115, 60-72.
doi:10.1016/j.engfracmech.2013.11.013
Trudel, A., Sabourin, M., Lévesque, M., & Brochu, M. (2014). Fatigue crack growth in the heat affected
zone of a hydraulic turbine runner weld. International Journal of Fatigue, 66, 39-46.
doi:10.1016/j.ijfatigue.2014.03.006
Virkler, D. A., Hillberry, B. M., & Goel, P. K. (1979). The statistical nature of fatigue crack propagation.
Transaction of the ASME, 101, 148-153.
19
|
1903.12109 | 1 | 1903 | 2019-03-28T16:47:35 | Absolute Seebeck coefficient of thin platinum films | [
"physics.app-ph",
"cond-mat.mes-hall"
] | The influence of size effects on the thermoelectric properties of thin platinum films is investigated and compared to the bulk. Structural properties, like the film thickness and the grain size, are varied. We correlate the electron mean free path with the temperature dependence of the electrical conductivity and the absolute Seebeck coefficient $S_{\text{Pt}}$ of platinum. We use a measurement platform as a standardized method to determine $S_{\text{Pt}}$ and show that $S_{\text{Pt,film}}$ is reduced compared to $S_{\text{Pt,bulk}}$. Boundary and surface scattering reduce the thermodiffusion and the phonon drag contribution to $S_{\text{Pt,film}}$ by nearly the same factor. A detailed discussion and a model to describe the temperature dependence of the absolute Seebeck coefficient and the influence of size effects of electron-phonon and phonon-phonon interaction on $S_{\text{Pt}}$ is given. | physics.app-ph | physics | Absolute Seebeck coefficient of thin platinum films
Absolute Seebeck coefficient of thin platinum films
M. Kockert,1, a) R. Mitdank,1 A. Zykov,2 S. Kowarik,2, 3 and S. F. Fischer1, b)
1)Novel Materials Group, Humboldt-Universität zu Berlin, 10099 Berlin, Germany
2)Nanoscale processes and growth, Humboldt-Universität zu Berlin, 10099 Berlin,
Germany
3)Bundesamt für Materialforschung und -prüfung (BAM), 12203 Berlin, Germany
(Dated: 29 March 2019)
The influence of size effects on the thermoelectric properties of thin platinum films is investigated and com-
pared to the bulk. Structural properties, like the film thickness and the grain size, are varied. We correlate
the electron mean free path with the temperature dependence of the electrical conductivity and the absolute
Seebeck coefficient SPt of platinum. We use a measurement platform as a standardized method to determine
SPt and show that SPt,film is reduced compared to SPt,bulk. Boundary and surface scattering reduce the
thermodiffusion and the phonon drag contribution to SPt,film by nearly the same factor. A detailed discussion
and a model to describe the temperature dependence of the absolute Seebeck coefficient and the influence of
size effects of electron-phonon and phonon-phonon interaction on SPt is given.
I.
INTRODUCTION
II. EXPERIMENTAL DETAILS
Platinum is the most commonly used thermoelectric
reference material and is used with other materials, e.g.
as commercially available bulk thermocouples1. How-
ever,
in recent years micro- and nanopatterning have
become more interesting2,3. Popular examples are thin
films4,5 and nanowires6,7. New challenges for metrology
and its interpretation are coming with this trend.
In order to determine the thermoelectric transport
properties of nanowires, measurements are usually per-
formed relative to thin films6 -- 9. For this purpose, mi-
croelectromechanical systems (MEMS) with thin plat-
inum conduction lines of a few hundred nanometer thick-
ness have been developed as measurement platforms10 -- 12.
However, thin metal films have a reduced absolute See-
13 -- 18.
beck coefficient Sfilm compared to the bulk Sbulk
Especially for metal-metal junctions, it is important to
know the absolute Seebeck coefficient of the reference ma-
terial. Deviations in the single-digit microvolt per Kelvin
range can easily lead to misinterpretations of the mea-
surement results.
For this reason, we present a measurement platform
to investigate the temperature-dependent thermoelectric
transport properties of thin metal films. We demonstrate
the platforms usability by investigating platinum films
with a thickness of 134 nm and 197 nm. To understand
the influence of the microstructure on the absolute See-
beck coefficient S, we adapted and improved a model19,20
that allows the decomposition of S into a thermodiffusion
and a phonon drag contribution. We discuss the influence
of electron-phonon interaction compared to the phonon-
phonon interaction on the phonon drag part of SPt,film
and SPt,bulk. This work shows that thin platinum films
differ significantly from the bulk in terms of S by nearly
400 % at T = 290 K.
a)Electronic mail: [email protected]
b)Electronic mail: [email protected]
A thermoelectric micro lab (TML) was designed with
microlithography on a 5 mm x 9 mm glass substrate and
it is shown in figure 1 a. The TML involves a thermo-
couple, which consists of a thin sputtered platinum film
(sputter target: 99, 99 % platinum) and an attached bulk
gold wire (diameter: 25 µm, purity: 99.99 %). This gold
wire creates a thermoelectric connection between the up-
per and lower part of the platinum film. A platinum
line heater is used to generate a temperature difference
δT along the sample, which can be determined by four-
terminal resistance thermometers. The resulting temper-
ature difference along both materials due to the different
junction temperatures (T1 > T2) produces a thermovolt-
age US. The relative Seebeck coefficient between the gold
wire and the thin platinum film with respect to the cold
side is given by
SAu,Pt = − USAu,Pt
δT
.
(1)
Here, we prepared and investigated platinum films with
a thickness of 134 nm and 197 nm. After the thermoelec-
tric characterization of the samples in a flow cryostat in
helium atmosphere at ambient pressure, a heat treatment
was performed on the same samples. The heat treat-
ment was carried out in a rapid thermal annealer in vac-
uum. The temperature of the annealer was gradually in-
creased from 115 ◦C to 250 ◦C and finally to a maximum
of 400 ◦C. The temperature plateaus were held for two
minutes each. A thermoelectric characterization of the
same samples was performed after the heat treatment.
For the X-ray investigations, 10 mm x 10 mm large sam-
ples were prepared according to the same procedure with
a thickness of 139 nm and 203 nm. X-ray experiments
have been conducted with a lab-based diffractometer,
with a Cu-Kα rotating anode source with a wavelength
of λ = 0.154 nm. The thickness of all samples was deter-
mined by atomic force microscopy.
9
1
0
2
r
a
M
8
2
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
1
v
9
0
1
2
1
.
3
0
9
1
:
v
i
X
r
a
Absolute Seebeck coefficient of thin platinum films
2
Sample t (nm) Tmax (◦C)
Pt 1
Pt 2
Pt 3
Pt 4
134 ± 5
134 ± 5
197 ± 5
197 ± 5
−
400
−
400
RRR
−
8.45 ± 0.01
3.00 ± 0.01
8.26 ± 0.01
TABLE I. Thickness, heat treatment parameters and
residual resistance ratio. Overview of thickness t, max-
imum heat treatment temperature Tmax and residual resis-
tance ratio RRR of four thin platinum films. The thickness
was determined by atomic force microscopy. The residual
resistance ratio RRR was determined as the ratio of the re-
sistance at 290 K divided by the resistance of 20 K.
III. RESULTS
A. X-ray
X-ray measurements of polycrystalline platinum films
with thicknesses of 139 nm and 203 nm exhibit four Bragg
peaks corresponding to the platinum (111), (200), (220)
and (311) reflections. The most intense Bragg peak of
the 203 nm thin film is the (111) reflection (see figure
2 a, indicating a preferential orientation of the crystal-
lites with a surface parallel (111) plane. Additional heat
treatment and increased temperature of the heat treat-
ment lead to an increase in the peak intensity and a
shift of the peak position to larger detector angles, corre-
sponding to smaller lattice constants. The position of the
Bragg reflection at (111) of the annealed platinum films
is in agreement with literature21. The average crystal-
lographic grain size DS of the crystallites with a (111)
orientation was estimated by the Scherrer equation22,23
DS =
Kλ
∆(2θ) cos(θ)
.
(2)
K is a dimensionless shape factor with a value of 0.9,
λ = 0.154 nm is the X-ray wavelength. The broadening
∆(2θ) is given by the full-width at half-maximum FWHM
of the X-ray diffraction peak shown in figure 2 a and θ is
the Bragg angle. DS of 139 nm and 203 nm from as sput-
tered thin platinum films are DS,139nm = (33 ± 2) nm
and DS,203nm = (35 ± 2) nm, respectively. This size in-
creases with heat treatment at 400 ◦C to DS,139nm,400C =
(41±1) nm and DS,203nm,400C = (41±1) nm, respectively.
Furthermore, the mosaicity Γ, which is a measure of
the spread of crystal plane orientations, was determined
for the (111) Bragg reflection. Figure 2 b shows a rock-
ing scan where the sample angle ω is varied for a fixed
detector angle 2θ. As the (111) lattice planes are not
all perfectly parallel to the substrate surface, intensity
is found within an angular distribution of FWHM. Γ of
139 nm and 203 nm as sputtered thin platinum films are
Γ139nm = (11.0 ± 0.4) ◦ and Γ203nm = (8.9 ± 0.2) ◦, re-
spectively. Thicker platinum films therefore have a more
perfect texture corresponding to a narrower distribution
of crystallite tilt angles. The tilt decreases and so the
FIG. 1. Sketches of the measurement setup. a Sketch of
the thermoelectric micro lab. Platinum (Pt) was sputtered on
a (5 x 9) mm2 glass substrate. A line heater creates a temper-
ature difference δT = T1 − T2 along the sample by applying
a current at the contacts IH+ and IH-. A gold (Au) wire
(diameter d = 25 µm) was bonded between the hot (red tem-
perature T1) and cold (blue temperature T2) four-terminal
resistance thermometers indicated by the corresponding cur-
rent contacts (IT1 ,T2) and voltage contacts (UT1 ,T2). The
thermovoltage US was measured at the Pt pads with respect
to the cold side of the sample. The Pt pads were kept at the
same temperature T0 in order to minimize parasitic thermo-
voltages. b Sketch of the measurement setup as a thermocou-
ple consisting of gold (Au) and platinum (Pt). A temperature
difference along both materials resulting from different junc-
tion temperatures (T1 > T2) between Au and Pt produces a
thermovoltage US.
Absolute Seebeck coefficient of thin platinum films
crystal quality increases with heat treatment at 400 ◦C to
Γ139nm,400C = (9.5±0.2) ◦ and Γ203nm,400C = (7.7±0.2) ◦,
respectively.
B. Electrical measurements
3
The Bloch-Grüneisen equation was used to fit the
temperature-dependent resistance of the platinum films,
which show the expected metallic behavior given in fig-
ure 3 a, in order to determine the Debye temperature
ΘD of the material. All platinum films with heat treat-
ment and the bulk material (wire diameter d = 25 µm)
agree with the literature value of ΘD = 240 K 25. Pt 3
(197 nm, without heat treatment) has a reduced Debye
temperature ΘD,Pt 3 = 191 K compared to the literature.
This can be attributed to the microstructure.
Furthermore, the residual resistance ratio RRR was
determined as the ratio of the resistance at 290 K divided
by the resistance of 20 K in order to compare the quality
of the sputtered platinum films and the influence of the
heat treatment. Bulk platinum has the highest residual
resistance ratio RRRbulk = 50.3 ± 0.1. Thin films have
a reduced residual resistance ratio compared to the bulk
given in table I.
Figure 3 b shows the temperature coefficient αPt of the
resistance, which depends on the thickness of the plat-
inum films. Larger film thickness and additional heat
treatment lead to an increase of the temperature coeffi-
cient. Like the residual resistance ratios, the temperature
coefficients of Pt 2 (134 nm, with heat treatment) and Pt
4 (197 nm, with heat treatment) are similar. The four-
terminal resistance of the cold thermometer was used to
determine the electrical conductivity σ of the thin plat-
inum films shown in figure 3 c. The electrical conduc-
tivity of the platinum bulk wire σPt,bulk is larger than
σPt,film. σPt,film depends on the film thickness, can be
increased by heat treatment and reaches a maximum at
low bath temperatures.
Figure 3 d features the electron mean free path Λel
of the of the thin films and of the bulk. The electron
mean free path of the thin films is reduced compared to
the bulk and increases with increasing film thickness and
can be further increased by heat treatment. At low bath
temperatures, the mean free path reaches a maximum
limited by the film thickness and the structural proper-
ties.
FIG. 2. X-ray diffractometry of a thin platinum film
with a thickness of 203 nm as sputtered and with ad-
ditional heat treatment. a Intensity of the (111) Bragg
reflection as a function of detector angle 2Θ as sputtered,
with heat treatment at a maximum temperature of 400 ◦C
and 600 ◦C. The average crystallographic grain size DS was
determined by the Scherrer equation. b Intensity of the (111)
Bragg peak as a function of the rocking angle variation ω as
sputtered, with heat treatment at a maximum temperature
of 400 ◦C and 600 ◦C. The mosaicity Γ of the platinum crys-
tallites with a (111) texture is given by the full width at half
maximum of the peak.
C. Seebeck measurements
In addition to the electrical characterization,
the
temperature-dependent Seebeck coefficient of thin plat-
inum films and bulk platinum relative to a bulk gold wire
(wire diameter d = 25 µm) was determined by measuring
the thermovoltage US as a function of the heater current
of the line heater. Figure 4 shows the parabolic behavior
of the thermovoltage indicating the thermoelectric effect.
The temperature difference δT was determined by four-
terminal resistance measurements of the thermometers at
the hot and cold side of the measurement platform. The
slope of the function US(δT ) gives the relative Seebeck
coefficient
3839404142 as sputtered 400 °C 600 °CIntensity (a.u.) (°)DS 35 nm DS 41 nm DS 44 nm-10-505100.00.20.40.60.81.0 as sputtered 400 °C 600 °C = 8.9° Intensity (a.u.) (°) = 7.7° = 7.0°(111) Mosaicity(111) Peak203 nmab Absolute Seebeck coefficient of thin platinum films
4
FIG. 3. Electrical measurements of thin platinum films as-sputtered and with heat treatment and of the bulk.
a Four-terminal resistance R as a function of the bath temperature T . The Bloch-Grüneisen equation (BGE) was fitted to
the data in order to determine the temperature dependence of the resistance and to calculate the Debye temperature. b
Temperature coefficient αPt of the resistance as a function of the bath temperature T . The inset shows temperature coefficient
of the bulk material. c Electrical conductivity σ as a function of the bath temperature T . d Mean free path of the electrons
Λel as a function of the bath temperature T . The mean free path of as sputtered thin platinum films is mainly limited by grain
boundaries at low temperatures. The mean free path of thin platinum films with heat treatment is mainly limited by the film
thickness.
IV. DISCUSSION
SAu,Pt = SAu − SPt = − USAu,Pt
δT
.
(3)
A. Structural and electrical characterization
The relative Seebeck coefficient of the thin films and
the bulk material decreases with decreasing bath temper-
ature as depicted in figure 5 a. In order to determine the
absolute Seebeck coefficient of platinum SPt, the absolute
Seebeck coefficient of bulk gold SAu was taken from24.
SPt is given in figure 5 b.
X-ray diffraction analysis gives the average crystal-
lographic grain size DS, which was determined by the
Scherrer equation22,23 from the (111) Bragg reflections
of figure 2 a. The determined values indicate only a
lower limit for the crystallographic grain size. The ac-
tual grain size can be larger. Furthermore, the X-ray
measurements only provide information about the grain
size in the growth direction. The in-plane grain size can
10100101001000film thickness Pt bulk Pt 4, 197 nm, 400 °C Pt 3, 197 nm Pt 2, 134 nm, 400 °C Pt 1, 134 nmMean free path el (nm)Temperature T (K)40200grain boundaries050100150200250300012345678 Pt 1, 134 nm, BGE fit Pt 2, 134 nm, 400 °C, BGE fit Pt 3, 197 nm, BGE fit Pt 4, 197 nm, 400 °C, BGE fit Pt bulk, BGE fitResistance R ()Temperature T (K)101000.1110100 Pt 1, 134 nm Pt 2, 134 nm, 400 °C Pt 3, 197 nm Pt 4, 197 nm, 400 °C Pt bulkElectrical conductivity (107 m-1)Temperature T (K)050100150200250300024681012141618 Pt 1, 134 nm Pt 2, 134 nm, 400 °C Pt 3, 197 nm Pt 4, 197 nm, 400 °CTemperature coefficient Pt (10-3 K-1)Temperature T (K)0501001502002503000102030405060 Pt bulkTemperature coefficient Pt (10-3 K-1)Temperature T (K)abcd Absolute Seebeck coefficient of thin platinum films
5
FIG. 4. Thermovoltage as a function of the heater
current and temperature difference of Pt 4. a Ther-
movoltage US as a function of heater current IH at different
bath temperatures T . The sign of the thermovoltage changes
below T ≈ 150 K. b Thermovoltage US as a function of tem-
perature difference δT at different bath temperatures T . The
slope of the fitted solid lines gives the relative Seebeck co-
efficient between the thin platinum films and the bulk gold
wire.
be significantly larger than the grain size in the growth
direction.
In addition to the average crystallographic
grain size, the morphological grain size is discussed in
literature26,27. It was proposed that the morphological
grains, which were determined by scanning tunneling mi-
croscopy, are agglomerates of crystallographic grains and
that the morphological grain size is increasing with in-
creasing film thickness much further than the crystallo-
graphic grain size26,27.
The structural properties like film thickness and grain
size can be linked with the mean free path of the electrons
in order to investigate the transport properties of the
thin platinum films. The electron mean free path Λel
was determined from the electrical conductivity of the
FIG. 5. Relative and absolute Seebeck coefficient of
thin platinum films as sputtered and with heat treat-
ment and bulk material. a Relative Seebeck coefficient
SAu,Pt of platinum relative to a gold wire with a diameter of
25 µm as a function of the bath temperature T . b Absolute
Seebeck coefficient SPt of platinum as a function of the bath
temperature T . The corresponding solid lines represent the
Seebeck fit of the data. Absolute Seebeck coefficient of bulk
gold SAu was taken from24.
thin films with Matthiessen rule7.
Λel,f(T )−1 = Λel,b(T )−1 + Λ−1
el,s,gb
(4)
is the inverse thin film electron mean free path,
Λel,b(T )−1 is the inverse mean free path of the bulk and
Λ−1
el,s,gb is the inverse temperature-independent scattering
length of the electrons due to surface and grain boundary
scattering.
and
Λel,f(T ) =
σf(T )
σf(RT)
Λel,f(RT)
Λel,b(T ) =
σb(T )
σb(RT)
Λel,b(RT)
(5)
(6)
0246810-60-50-40-30-20-10010Thermovoltage US (VK-1)Temperature difference T (K)84 K30 K177 K212 K268 K299 K-100-50050100-60-45-30-1501530 K84 K177 K212 K268 K299 KThermovoltage US (VK-1)Heater current IH (mA)abPt 4, 197 nm, 400 °C050100150200250300-6-4-20246 Pt 1, 134 nm Pt 2, 134 nm, 400 °C, Seebeck fit Pt 3, 197 nm, Seebeck fit Pt 4, 197 nm, 400 °C, Seebeck fit Pt bulk, Seebeck fitAbsolute Seebeck coefficient SPt (VK-1)Temperature T (K)-6-4-202468 Pt 1, 134 nm Pt 2, 134 nm, 400 °C Pt 3, 197 nm Pt 4, 197 nm, 400 °C Pt bulkRelative Seebeckcoefficient SAu,Pt (VK-1)ab Absolute Seebeck coefficient of thin platinum films
6
are the electron mean free paths of the thin films and
of the bulk material, respectively. Λel,b(RT) = 23 nm
is given in literature at room temperature25. Figure 3
d shows the electron mean free path as a function of
the bath temperature. Table II shows the electron mean
free path at room temperature and the temperature-
independent scattering length of the electrons. Pt 3
(197 nm, without heat treatment) has a mean free path
which saturates at low bath temperatures at Λel,s,gb =
(48 ± 1) nm. Pt 2 (134 nm, with heat treatment) and Pt
4 (197 nm, with heat treatment) have nearly the same
temperature dependence of the electron mean free path,
which is in the order of magnitude of the film thickness.
Heat treatment increases the mean free path. From the
structural properties, we can conclude that the electron
mean free path at low bath temperatures is mainly lim-
ited by grain boundaries for thin films without heat treat-
ment. For thin films with heat treatment, the main lim-
itation is set by the film thickness.
Sample t (nm) Tmax (◦C) Λel,f(RT) (nm) Λel,s,gb (nm)
Pt 1
Pt 2
Pt 3
Pt 4
36 ± 1
212 ± 2
48 ± 1
219 ± 2
14 ± 1
21 ± 2
16 ± 1
22 ± 2
-
400
-
400
134
134
197
197
TABLE II. Electron mean free path. Mean free path
of the samples with thickness t and maximum heat treat-
ment temperature Tmax. Λel,f(RT) is the electron mean free
path of the thin films at room temperature and Λel,s,gb is the
temperature-independent mean free path of the electrons due
to surface and grain boundary scattering. Λel,b(RT) = 23 nm
of bulk is given in literature at room temperature25.
B. Seebeck coefficient
In general, the Seebeck coefficient is the sum of two
parts. The thermodiffusion part and the phonon drag
part14,19,20:
S = Sdiff + Sph.
(7)
Sdiff is the contribution due to the thermodiffusion of the
charge carriers as described by Mott's formula14,20,28:
(cid:18) ∂ ln σ()
∂
(cid:19)(cid:12)(cid:12)(cid:12)(cid:12)F
Sdiff =
π2k2
BT
3e
with kB is the Boltzmann constant, T is the bath tem-
perature, e is the elementary charge and ∂ ln σ()
is the
derivative of the electrical conductivity according to the
energy at the Fermi energy.
∂
Sph is the contribution due to the phonon drag effect.
The phonon drag effect is based on the electron-phonon
interaction. A momentum transfer from phonons to elec-
trons leads to an increase of the Seebeck coefficient19.
(cid:12)(cid:12)F
(8)
Finally, we combine the approach of the phonon drag
part with the specific heat of the phonons and the ther-
modiffusion part Sdiff = Fdiff
to the formula, which
describes the Seebeck coefficient over a wide temperature
range
T
ΘD
The Seebeck fit and measurement results are given in
figure 5 b. The phonon drag peak appears between bath
(cid:18) T
(cid:19)3(cid:90) θD
T
ΘD
0
The phonon drag part of the Seebeck coefficient is con-
nected with the specific heat of the phonons19,20
Cph(T ) = 9nkB
x4 exp(x)
(exp (x) − 1)2 dx
(9)
3ne
ωD
kBT = ΘD
with the number of charge carriers per volume n and
T which is given by the reduced Planck
x =
constant and by the Debye frequency ωD.
There are low and high temperature approaches to de-
scribe the phonon drag part of the Seebeck coefficient19.
The low temperature approach can be described by Sph =
Cph,low
. According to the Debye model the specific heat
Cph,low goes with T → 0 to Cph,low ∝ T 3 and for this rea-
son Sph,low ∝ T 3 19. The resulting Seebeck coefficient
can be written as Slow = Sdiff + Sph,low = aT + bT 3. a
and b are variables.
The high temperature approach has to be modified
compared to the low temperature approach because the
phonon-phonon scattering time τpp and the scattering
time of the electron-phonon interaction τep has to be
taken into account19,20. The Debye model predicts a
constant specific heat (Cph,high ≈ 3nkB) for temperatures
above the Debye temperature ΘD and due to phonon-
phonon Umklapp scattering Sph,high ∝ 1
19. This re-
T
sults in Shigh = Sdiff + Sph,high = cT + d 1
T . c and d are
variables.
The combination of the low temperature approach, the
high temperature approach and additionally a character-
ization of the intermediate regime gives
Sph =
Cph(T )
3ne
γ =
Cph(T )
τpp
3ne
τpp + τep
19,20,
(10)
hence
Sph =
Cph(T )
3ne
1
Cph(T )
=
1 + τep
τpp
3ne
= Fτ T exp(cid:0)− ΘD
1
1 + Fτ T exp(cid:0)− ΘD
(cid:1) .
(cid:1) was suggested by
(11)
T
The relation τep
τpp
and adapted from20,29. This equation assumes that the
phonon-phonon interaction becomes dominant at high
temperatures20,29. This results in the γ-factor
T
γ =
1
1 + τep
τpp
=
1 + Fτ T exp(cid:0)− ΘD
1
(cid:1) .
T
(12)
(cid:16) T
(cid:17)3(cid:82) θD
1 + Fτ T exp(cid:0)− ΘD
ΘD
0
T
T
(cid:1)
x4 exp(x)
(exp(x)−1)2 dx
.
(13)
S = Fdiff
Fph
T
ΘD
+
Absolute Seebeck coefficient of thin platinum films
7
temperatures of 65 K and 75 K. A possible reason for the
deviation of the Seebeck fit from the measurement results
below 50 K is that the phonon drag part is very sensitive
to impurities14,24,30. The Seebeck fit provides different
parameters, which are given in table III.
The thermodiffusion part is described by the fit pa-
rameter Fdiff. The platinum films without heat treat-
ment have the smallest thermodiffusion contribution to
the absolute Seebeck coefficient, which can be attributed
to an increased scattering rate at grain boundaries. Pt
2 (134 nm, with heat treatment) and Pt 4 (197 nm,
with heat treatment) exhibit a film thickness depen-
dence of the thermodiffusion part on the bath temper-
ature. The fit parameter of the thermodiffusion part of
Pt 2 is Fdiff,Pt 2 = (−2.6 ± 0.1) µVK−1 and of Pt 4 is
Fdiff,Pt 4 = (−4.7 ± 0.1) µVK−1. This difference can be
explained by size effects like surface scattering, which is
more likely to happen in Pt 2 than in Pt 4. The thermod-
iffusion part of Pt 4 and bulk have nearly the same tem-
perature dependence and magnitude and it seems that
Pt 4 reaches the upper limit of the thermodiffusion part,
which is provided by the bulk material.
Compared to the thermodiffusion part, the phonon
drag part vanishes at high bath temperatures. Fph es-
timates the strength of the phonon drag on the abso-
lute Seebeck coefficient. Pt 3 (197 nm, without heat
treatment) has the smallest value of the fit parameter
Fph,Pt 3 = (24 ± 2) µVK−1. This value increases with
increasing film thickness and with additional heat treat-
ment and reaches its maximum for the bulk material with
Fph,bulk = (47 ± 2) µVK−1.
The ratio Fph/Fdiff of bulk and of the thin films is
approximately 10. Except for Pt 4 (197 nm, with heat
treatment), which is slightly lower. This ratio indicates
that the thermodiffusion and phonon drag part are re-
duced by nearly the same factor, when the film thickness
is decreasing. To further illustrate that the thermodiffu-
sion and the phonon drag part are related to each other,
we introduce Fτ , which gives the ratio of the scattering
time of the electron-phonon and phonon-phonon inter-
action and determines the γ-factor. The γ-factor, see
equation 12, is a number between 0 and 1, which de-
pends on the interaction between phonons and electrons.
For T (cid:28) ΘD, γ ≈ 1, means electron-phonon interac-
tion is dominant compared to phonon-phonon interac-
tion. Phonon-phonon interaction is dominant compared
to electron-phonon for γ ≈ 0.
The γ-factor as a function of temperature for thin films
and bulk is given in figure 6. For all temperatures ap-
plies: γfilm > γbulk. This means that there is an in-
creased amount of electron-phonon interaction compared
to phonon-phonon interaction in the thin films than in
the bulk. The influence of the phonon drag part on the
absolute Seebeck coefficient dominates in thin films com-
pared to the bulk. For example, the thermodiffusion part
of Pt 3 is reduced by 77 % towards bulk, resulting in a
significant effect of the phonon drag part even at room
temperature. This difference can be explained by the
inner and outer interfaces of the the thin films and the
resulting grain boundary scattering.
Decreasing temperatures lead to an increase of γfilm
and γbulk. This indicates that the electron-phonon inter-
action is becoming more dominant compared to phonon-
phonon interaction. At temperatures below 50 K, the
thermodiffusion part tends to 0 and the γ-factor tends to
1. The reason for this behavior can be attributed to the
phonon-phonon interaction, which is negligible compared
to the electron-phonon interaction.
Pt 2 (134 nm, with heat treatment) and Pt 4 (197 nm,
with heat treatment) have the same γ-factor but dif-
ferent absolute Seebeck coefficients, indicating that the
ratio of electron-phonon interaction compared to the
phonon-phonon interaction is the same but the absolute
amount of electron-phonon and phonon-phonon interac-
tion is larger in thicker platinum films with heat treat-
ment, because the essential limitation is no longer caused
by grain boundaries, but by the film thickness.
Sample t (nm) Tmax (◦C) Fph (µVK−1)
Pt 1
Pt 2
Pt 3
Pt 4
Bulk
24 ± 2
12 ± 1
35 ± 2
47 ± 2
134
134
197
197
-
-
400
-
400
-
-
-
Fdiff (µVK−1) Fph/Fdiff SPt (µVK−1) at 280 K
Fτ (K−1)
−2.6 ± 0.1
0.05 ± 0.01
0.020 ± 0.005 −1.1 ± 0.1
0.05 ± 0.01
−4.7 ± 0.1
−4.8 ± 0.1
0.09 ± 0.01
−0.6 ± 0.4
−2.3 ± 0.3
−0.7 ± 0.1
−4.7 ± 0.2
−4.8 ± 0.1
9.2 ± 0.8
10.9 ± 1.3
7.4 ± 0.5
9.8 ± 0.5
-
-
TABLE III. Seebeck fit parameters. Parameters of Seebeck fit for each sample with thickness t and maximum heat
treatment temperature Tmax. Fph is the parameter which describes the intensity of the phonon drag part. Fτ gives the ratio
of the scattering time of the electron-phonon interaction and of the phonon-phonon interaction. Fdiff describes the intensity of
the thermodiffusion part. Fph/Fdiff gives the modulus ratio between the fit parameter of the thermodiffusion part and phonon
drag part. SPt (µVK−1) is the absolute Seebeck coefficient of platinum at 280 K. Due to the lack of low temperature data, the
fit parameters of Pt 1 (197 nm, without heat treatment) are not given.
V. CONCLUSION
The influence of heat treatment and film thickness on the
In this work, we performed thermoelectric and struc-
tural characterizations of thin sputtered platinum films.
Absolute Seebeck coefficient of thin platinum films
8
VII. REFERENCES
1J. Machin, D. Tucker and J. V. Pearce. A comprehensive survey
of thermoelectric homogeneity of commonly used thermocouples
types. Measurement Science and Technology 29, 075418 (2018).
2A. M. Burke, D. J. Carrad, J. G. Gluschke, K. Storm, S. Fahlvik
Svensson, H. Linke, L. Samuelson and A. P. Micolich. InAs
Nanowire Transistors with Multiple, Independent Wrap-Gate
Segments. Nano Letters 15, 2836 (2015).
3M. S. Dresselhaus, G. Chen, M. Y. Tang, R. Yang, H. Lee, D.
Wang, Z. Ren, J.-P. Fleurial and P. Gogna. New Directions for
Low-Dimensional Thermoelectric Materials. Advanced Materials
19, 1043 (2007).
4R. Venkatasubramanian, E. Siivola, T. Colpitts and B. O'Quinn.
Thin-film thermoelectric devices with high room-temperature fig-
ures of merit. Nature 413, 597 (2001).
5M. V. Daniel, M. Lindorf and M. Albrecht. Thermoelectric
properties of skutterudite CoSb3 thin films. Journal of Applied
Physics 120, 125306 (2016).
6A. I. Hochbaum, R. Chen, R. D. Delgado, W. Liang, E. C. Gar-
naett, M. Najarian, A. Majumdar and P. Yang. Enhanced ther-
moelectric performance of rough silicon nanowires. Nature 451,
163 (2008).
7D. Kojda, R. Mitdank, M. Handwerg, A. Mogilatenko, M. Al-
brecht, Z. Wang, J. Ruhhammer, M. Kroener, P. Woias and S. F.
Fischer. Temperature-dependent thermoelectric properties of in-
dividual silver nanowires. Physical Review B 91, 024302 (2015).
8J. Kim, G. Kim, J.-H. Bahk, J.-S. Noh and W. Lee. Enhanced
thermoelectric properties in Bi/Te core/shell heterostructure
nanowires through strain and interface engineering. Nano En-
ergy 32, 520 (2017).
9D. Kojda, R. Mitdank, A. Mogilatenko, W. Töllner, Z. Wang,
M. Kröner, P. Woias, K. Nielsch and S. F. Fischer. The effect
of a distinct diameter variation on the thermoelectric properties
of individual Bi0.39Te0.61 nanowires. Semiconductor Science and
Technology 29, 124006 (2014).
10Z. Wang, S. S. Adhikari, M. Kroener, D. Kojda, R. Mitdank, S. F.
Fischer, W. Toellner, K. Nielsch and P. Woias. Electrical conduc-
tivity and Seebeck coefficient measurements of single nanowires
by utilizing a microfabricated thermoelectric nanowire character-
ization platform. IEEE 26th International Conference on Micro
Electro Mechanical Systems (MEMS) (IEEE, Pscataway, NJ, 508
(2013).
11S. H. Moosavi, D. Kojda, M. Kockert, S. F. Fischer, M. Kroener
and P. Woias. The Effect of the MEMS Measurement Plat-
form Design on the Seebeck Coefficient Measurement of a Single
Nanowire. Nanomaterials 8, 219 (2018).
12V. Linseis, F. Völklein, H. Reith, P. Woias and K. Nielsch. Plat-
form for in-plane ZT measurement and Hall coefficient determi-
nation of thin films in a temperature range from 120 K up to 450
K. Journal of Materials Research 31, 3196 (2016).
13G. Brändli and J. L. Olsen. Size Effects in Electron Transport in
Metals. Materials Science and Engineering 4, 61 (1969).
14R. P. Huebener. Size Effect on Phonon Drag in Platinum. Phys-
16W. F. Leonard and H.-Y. Yu. Thermoelectric power of thin cop-
per films. Journal of Applied Physics 44, 5320 (1973).
17H.-Y. Yu and W. F. Leonard. Thermoelectric power of thin silver
films. Journal of Applied Physics 44, 5324 (1973).
18V. D. Das and N. Soundararajan. Size and temperature effects
on the Seebeck coefficient of thin bismuth films. Physical Review
B 35, 5990 (1987).
19D. K. C. MacDonald. Thermoelectricity − An introduction to
the principles. Dover Publications, Mineola, New York (2006).
20R. P. Huebener. Phonon scattering by lattice vacancies in plat-
inum. Physical Review 146, 490 (1966).
15R. P. Huebener. Thermoelectric Size Effect in Pure Gold. Phys-
ical Review 140, 490 (1965).
ical Review 136, 1740 (1964).
FIG. 6. The panel shows the γ-factor of thin platinum films
as sputtered and with heat treatment and bulk material as a
function of bath temperature. The γ-factor is determined by
the electron-phonon (e-p) and phonon-phonon (p-p) interac-
tion. The gray shaded area around the solid lines marks the
uncertainty. γ = 1 means that the electron-phonon interac-
tion compared to the phonon-phonon interaction is dominant.
γ = 0 means that the phonon-phonon interaction compared
to the electron-phonon interaction is dominant. The curves
of Pt 2 and Pt 4 have a similar temperature dependence and
are lying on top of each other.
electrical conductivity and the absolute Seebeck coeffi-
cient were investigated. Additional heat treatment and a
larger film thickness increase the crystal quality of sput-
tered platinum films. The electrical conductivity and
the absolute Seebeck coefficient are reduced compared
to the bulk due to size effects like surface and boundary
scattering. We find that structural properties like grain
size and film thickness, which limit the electron mean
free path, influence the absolute Seebeck coefficient. For
the phonon drag part of the absolute Seebeck coefficient,
the electron-phonon interaction compared to the phonon-
phonon interaction plays a more dominant role in thin
films than in bulk. If the mean free path of thin metallic
films is in the order of the film thickness, the absolute
Seebeck coefficient of bulk is no more valid. This has to
be taken into account, when using thin platinum films
as a reference material for the determination of the ab-
solute Seebeck coefficient. Due to the influence of the
microstructure, metallic interconnects can be tailored in
a way, that the relative Seebeck coefficient can be reduced
to zero, which is interesting for low-noise applications.
VI. ACKNOWLEDGEMENTS
We acknowledge partial financial support by the Ger-
man Science Foundation (DFG-SPP1386).
with heat treatmentwithout heat treatmentp-p interaction Pt 2, 134 nm, 400 °C Pt 3, 197 nm Pt 4, 197 nm, 400 °C Pt bulk-factorTemperature T (K)e-p interaction Absolute Seebeck coefficient of thin platinum films
9
21R. W. G. Wyckoff. Crystal Structures. Interscience Publisher,
Physical Review B 82, 075418 (2010).
New York (1963).
22P. Scherrer. Bestimmung der Groesse und der inneren Struktur
von Kolloidteilchen mittels Röntgenstrahlen. Nachrichten von
der Königlichen Gesellschaft der Wissenschaften zu Göttingen
2, 98 (1918).
23C. Weber, L. Pithan, A. Zykov, S. Bommel, F. Carla, R. Felici, C.
Knie, D. Bleger and S. Kowarik. Multiple timescales in the photo-
switching kinetics of crystalline thin films of azobenzene-trimers.
Journal of Physics: Condensed Matter 29, 434001 (2017).
24R. P. Huebener. Thermoelectric Power of Lattice Vacancies in
Gold. Physical Review 135, A 1281 (1964).
25N. Stojanovic, D. H. S. Maithripala, J. M. Berg and M. Holtz.
Thermal conductivity in metallic nanostructures at high tem-
perature: Electrons, phonons, and the Wiedemann-Franz law.
26M. C. Salvadori, L. L. Melo, A. R. Vaz, R. S. Wiederkehr, F. S.
Teixeira and M. Cattani. Platinum and gold thin films deposited
by filtered vacuum arc: morphological and crystallographic grain
sizes. Surface & Coatings Technology 200, 2965 (2006).
27L. L. Melo, A. R. Vaz, M. C. Salvadori and M. Cattani. Grain
Sizes and Surface Rougness in Platinum and Gold Thin Films.
Journal of Metastable and Nanocrystalline Materials 20-21, 623
(2004).
28M. Cutler and N. F. Mott. Observation of Anderson Localization
in an Electron Gas. Physical Review 181, 1336 (1969).
29C. T. Walker and R. O. Pohl. Phonon scattering by point defects.
Physical Review 131, 1432 (1963).
30M. Bailyn. Phonon-Drag Part of the Thermoelectric Power in
Metals. Physical Review 157, 480 (1967).
|
1801.02557 | 1 | 1801 | 2018-01-08T17:03:12 | The effect of heavy tars (toluene and naphthalene) on the electrochemical performance of an anode-supported SOFC running on bio-syngas | [
"physics.app-ph",
"cond-mat.mtrl-sci"
] | The effect of heavy tar compounds on the performance of a Ni-YSZ anode supported solid oxide fuel cell was investigated. Both toluene and naphthalene were chosen as model compounds and tested separately with a simulated bio-syngas. Notably, the effect of naphthalene is almost negligible with pure H2 feed to the SOFC, whereas a severe degradation is observed when using a bio-syngas with an H2:CO = 1. The tar compound showed to have a remarkable effect on the inhibition of the WGS shift-reaction, possibly also on the CO direct electro-oxidation at the three-phase-boundary. An interaction through adsorption of naphthalene on nickel catalytic and electrocatalytic active sites is a plausible explanation for observed degradation and strong performance loss. Different sites seem to be involved for H2 and CO electro-oxidation and also with regard to catalytic water gas shift reaction. Finally, heavy tars (C>=10) must be regarded as a poison more than a fuel for SOFC applications, contrarily to lighter compounds such benzene or toluene that can directly reformed within the anode electrode. The presence of naphthalene strongly increases the risk of anode re-oxidation in a syngas stream as CO conversion to H2 is inhibited and also CH4 conversion is blocked. | physics.app-ph | physics | The effect of heavy tars (toluene and naphthalene) on the electrochemical
performance of an anode-supported SOFC running on bio-syngas
Davide Papurello*, Andrea Lanzini, Pierluigi Leone, Massimo Santarelli
a Department of Energy (DENERG), Politecnico di Torino, Corso Duca degli Abruzzi, 24, 10129,
Turin, Italy.
1
Abstract
The effect of heavy tar compounds on the performance of a Ni-YSZ anode supported solid oxide
fuel cell was investigated. Both toluene and naphthalene were chosen as model compounds and
tested separately with a simulated bio-syngas. Notably, the effect of naphthalene is almost
negligible with pure H2 feed to the SOFC, whereas a severe degradation is observed when using a
bio-syngas with an H2:CO = 1. The tar compound showed to have a remarkable effect on the
inhibition of the WGS shift-reaction, possibly also on the CO direct electro-oxidation at the three-
phase-boundary. An interaction through adsorption of naphthalene on nickel catalytic and
electrocatalytic active sites is a plausible explanation for observed degradation and strong
performance loss. Different sites seem to be involved for H2 and CO electro-oxidation and also with
regard to catalytic water gas shift reaction. Finally, heavy tars (C≥10) must be regarded as a poison
more than a fuel for SOFC applications, contrarily to lighter compounds such benzene or toluene
that can directly reformed within the anode electrode. The presence of naphthalene strongly
increases the risk of anode re-oxidation in a syngas stream as CO conversion to H2 is inhibited and
also CH4 conversion is blocked.
Keywords: Bio-syngas, SOFC, Tar compounds, Electro-chemical performance, Naphthalene,
Toluene.
*Corresponding author. Tel.:+393402351692. Email address: [email protected]
2
Nomenclature:
ASC, anode supported cell,
ASR, area specific resistance,
C10H8, naphthalene,
C2H2, acetylene,
C2H4, ethylene,
C7H8, toluene,
ECN, energy research centre of the netherlands,
EIS, electrochemical impedance spectroscopy,
ESC, electrolyte supported cell,
FU, fuel utilization,
LHV, low heating value,
Ni-GDC, nickel mixed with the ceramic material GDC,
Ni-YSZ, nickel mixed with the ceramic material YSZ,
ppm(v), parts per million by volume,
RWGS, reverse water gas shift reaction,
SEM/EDS, scanning electron microscope/energy dispersive x-ray spectrometry,
SOFC, solid oxide fuel cell,
TPB, three phase boundary,
WGS, water gas shift reaction.
3
Introduction
Solid oxide fuel cells (SOFCs) coupled with biomass source offer the potential of highly efficient
combined heat and power generation [1]. The theoretical energy performance expected from
integrated biomass gasifier and SOFC systems has been investigated by several authors [2–6] and
efficiency as high as 50% (biomass-to-electricity, LHV basis) can be reached. However, the direct
use of biosyngas may degrade the performance of SOFCs as it contains minor species, including
particulates, hydrogen sulfide, chlorides, alkali compounds, and tars. These species can deactivate
or damage SOFC anodes [4,7–10] when contained in the feeding stream. Among these trace
species, tars have been identified as a major concern in developing gasifier–SOFC power systems
as they can potentially deactivate the anode catalysts and degrade fuel cells performance mainly
through carbon deposition [11]. Tar removal is still a critical issue in the effort toward integrated
biomass-fed gasifier fuel cell systems [12]. The amount of tars in the syngas depends much on the
gasifier configuration (up-draft, down-draft, fluidized circulating bed) as well as operating
conditions (gasifier agent, temperature and pressure). A study of Phuphuakrat et al., (2010) showed
for a downdraft dry sewage sludge gasifier, a syngas with the following tar content: essentially
benzene, toluene, xylene, styrene, phenol and naphthalene [13]. Furthermore, when seeking for a
hot integration of the gasifier with the SOFC generator, tar removal at high temperature becomes
generally costly as a dedicated catalytic tar cracker reformer is mandatory. There is still the risk of
trace amount of heavy hydrocarbons that passes through the catalytic bed to achieve the SOFC
anode. In Aravind et al., (2008) was studied the impact of naphthalene (110 ppm(v)) on the
performance of electrolyte supported cells with a graded Ni-GDC anode [7]. The cell showed
feasible performance with that concentration. However, only a H2-H2O anode feeding mixture was
tested and no results were provided for CO-, CH4- containing fuel. Recently, Liu et al., (2013)
studied the effect of toluene on the performance of Ni-GDC anodes in electrolyte-supported type
cells [14]. Toluene was well tolerated during the experiments up to 20 g/Nm3 (corresponding to
4
around 5,000 ppm(v) of contaminant in the anode syngas feed). The performance of Ni-GDC
composite anodes is generally recognized to be more resistant to carbon deposition than Ni-YSZ, as
also shown in the study of Lorente et al., (2013), where the two materials were tested with real tar
fractions in a micro-reactor configuration [15]. The short-term operation (~7 hrs) on a high tar load
(>10 g/Nm3) has proven that tars did not cause immediate problems to Ni-GDC anode operation
[7]. Indeed, no performance losses were observed and no carbon or other product gas trace
constituents contamination of the anodes was found when the SOFC membranes were examined
with SEM/EDS after the tests. To avoid carbon deposition it seems to be a synergic effect of having
a thin Ni-GDC anode support on an electrolyte supports during the electrochemical operation. The
Ni-phase is mostly involved (active) toward electrochemical reactions and it is not deactivated by
carbon forming reactions or other surface adsorption phenomena. However, limited studies were
conducted so far on Ni-YSZ anode-supported cells that still represent the dominant design for
commercial systems, especially because lower temperature operation is feasible (650-750 °C)
compared to ESC. The presented experimental results of stable SOFC operation on real heavily tar-
laden biomass derived product gas are promising when compared with the findings reported from
tests with synthetically introduced tar at comparable loads on similar anode material, where the
introduction of different tar levels, i.e. naphthalene (0.29 – 6 g Nm3), phenanthrene (1 g Nm3),
pyrene (0.2 g Nm3) caused a rapid voltage drop within a few hours before leveling out at a lower
voltage. Usually, tar model compounds chosen for SOFC performance testing are benzene or
toluene [13–15]. Only a very limited number of studies, according to our knowledge, are devoted to
heavier tars such phenols and naphthalene [2,16,17]. At ECN was tested ASC and ESC cells with
various tars, including toluene (0.4 %vol.), naphthalene (525 ppm(v)), phenanthrene (126 ppm(v))
and pyrene (22 ppm(v)) at 750 °C, 0.3 A/cm², FU = 60% in a reformate gas mixture. Cell
degradation was clearly observed during toluene feeding, and much steeper with naphthalene.
However, a very high-tar containing anode feed was used, and safe limits of operation with such
contaminants were not identified. In the same report, ECN also evaluated the performance of ESC
5
cell with a Ni-GDC graded anode at 850 °C, 26%vol. H2O, 0.16 A/cm², FU = 60%. In this case a
higher tolerance to toluene was observed, still with a degradation trend. Naphthalene was inserted
gradually this time, with concentrations of 50, 100, 250 and 525 ppm(v), respectively. Even with
the lowest amount (50 ppm(v)), degradation was observed. In the same experiment, anode feeds
containing phenanthrene (126 ppm(v)) and pyrene (22 ppm(v)) were also tested showing strong
degradation. From such experiments, more refined with the ESC-type cell, the clearly emerging
trend is that the higher the C amount (i.e., the heavier the tar), the higher is also the degradation.
Carbon deposition was the recognized source of degradation, even though very fast deactivation
occurs which may be due to also to other phenomena that we have tried to better understand with
the present work. The effect of acetylene (C2H2) and C2H4 (ethylene) was also studied, resulting in
no degradation when feeding up to 0.1 %vol. of C2H2 and 1.4 %vol. of C2H4. Again, lower C-
compounds are easily converted as fuel in the SOFC. This is true also for toluene, however de-
activation and related cell degradation start becoming evident. For heavier compounds (i.e., C ≥
10), degradation becomes more marked and fuel conversion is strongly inhibited (also for lighter
hydrocarbons present in the anode feed).
In this study, we have chosen toluene and naphthalene as tar model compounds to study the
electrochemical performance of a commercial Ni-YSZ anode-supported SOFC running on
simulated bio-syngas containing up to few ppm(v) of such hydrocarbons. Only few studies report
the effect of such contaminants on the SOFC performance when running with such anode mixture.
A real syngas fuel was reproduced and a full understanding of the interaction of naphthalene with a
Ni-YSZ under relevant electrochemical operating conditions is still lacking. Our focus is limited to
the impact of few ppm(v) (up to 50 ppm(v)) of naphthalene in the anode feed. We focus on
commercial-type anode-supported Ni-YSZ cells, for which almost no literature studies are available
for SOFCs fed by heavy tars in a syngas mixture during an electrochemical experiment.
6
Material and methods
The experiments are intended to simulate the behavior of wood gas from biomass gasification on a
SOFC anode. In the test bench, synthetic wood gas consisting of H2, CO, CH4, CO2 and N2 that are
mixed using five mass flow controllers (Bronkhorst, The Netherlands). A liquid mass-flow
controller in connection with a controller mixer evaporator was used to feed and vaporize water, to
be mixed with the other dry gaseous components. Naphthalene was co-fed with H2 using a certified
gas cylinder (Siad Spa, Italy) containing 50 ppm(v) of C10H8 in H2. The cylinder pressure was
limited to 28 bar(g) in order to avoid tar condensation and thus segregation from the gas matrix.
Toluene was co-fed with CH4 using a certified gas cylinder (Siad Spa, Italy) containing 200 ppm(v)
of C7H8 in CH4. The cylinder pressure was limited to 46 bar(g) in order to avoid tar condensation
and thus segregation from the gas matrix. Such approach is useful when co-feeding, with anode
mixture, low or ultra-low concentration of contaminants in an extremely accurate way. Nonetheless,
conventional systems based on evaporator and carrier gas, where the contaminant is available as
pure compound and a carrier gas is used to extract the vapor phase formed above the liquid (or
solid) volume fraction, are the only option for higher concentrations. In this study, the interest was
to find the threshold limit of heavy tar compounds affecting the SOFC performance. Moreover, in
the real-configuration system, testing low concentrations could be also relevant as they could
represent the case of a tar-cracker able to convert most of the tar in the producer gas, leaving
however some hydrocarbons unconverted. Nonetheless, the main motivation of using low
concentrations to investigate the underlying degradation phenomena occurring in the presence of
heavy-tars in the anode feed stream. A commercial anode-supported cell (SOLIDpower Spa, Italy)
was used for all experiments. The cells consist of a Ni-YSZ anode support of ~240 µm, an anode
active layer, a YSZ dense electrolyte and a LSCF cathode. The cells consisted of a disk of 50 mm of
anode diameter which were tested in oven configuration with seal-less ceramic housings. A detailed
description of the experimental apparatus can be found elsewhere Errore. L'origine riferimento
7
non è stata trovata.. The operating temperature was 740 – 750 °C: the cell temperature was
measured through a thermocouple (TC) placed at the center of the cell on the anode side. In this
way, the monitored TC signal is quite representative of variation occurring on the anode electrode
due to chemical reactions (e.g., endothermic steam-reforming, slightly exothermic water-gas shift,
etc.). The volumetric dry syngas composition used for all experiments is H2/CO/CH4/CO2/H2O/N2
19/17/2/8/7/47, yielding an H2/CO ratio quite close to 1. A (molar) steam-to-carbon of 0.4 (related
to amount of carbon contained in C-fuels, i.e., CO and CH4) was set to prevent carbon deposition
while avoiding an excessive amount of steam that could lead to enhanced Ni degradation. On
another test bench, the catalytic conversion study of a gas mixture with tars was carried out with a
concentration of C10H8 on nickel anode half-cell. The reactor was placed into a furnace whose
operative temperature was leaded by a controller driven (Horst Gmbh, Germany) by a K–type
thermocouple placed in a thermo well cantered in the catalyst bed. Two additional thermocouples
(K–type) were located at the inlet and outlet of the catalytic bed to measure temperature of inlet and
outlet gas flows (Tersid, Milano Italy). The heating apparatus was calibrated in such a way that all
temperature measuring points had the same constant temperature. An alumina fixed-bed reactor
(int.=14 mm; hbed=180 mm) with a previously reduced anode + electrolyte cell was heated up to
700 °C in a ceramic electric furnace. The reactor was fed by two different inlet streams:
H2/CO2 (50/50 %vol.) mixture with 25 ppm(v) of C10H8,
The syngas mixture with a steam to carbon ratio fixed at 0.4 with 10 ppm(v) of pollutant
(concentration selected as the maximum value possible depending on the H2 molar
concentration adopted).
According to Guerra et al., (2014), we selected the fuel flow and the nickel anode amount: the
nickel anode half-cell reduced was about 0.55 g [18]. An HPR-20 mass spectrometer (Hiden ltd,
UK) has been adopted to monitor the furnace outlet gas concentration.
8
Results
Naphthalene effect on SOFC performance
The effect of Naphthalene was first check against an anode feed containing H2-fuel only,
humidified at 6.1%. The fuel cell performance was not affected by the presence of a varying amount
of C10H8 up to 50 ppm(v) under an operating temperature of 750 °C and a current density of 0.33
A/cm2 (fig. 1). In fig.2 the EIS were measured during the galvanostatic experiment shown in fig. 1
(portion of the graph on the left, where H2 fuel is used). Essentially the impedance response was
observed with/without the tar. Actually, a slightly shift on the left (meaning lower overall
polarization) was observed when feeding naphthalene as a result of an activation process still
occurring. Adding at the hydrogen stream, naphthalene the test improves due to the more fuel
adopted that is reformed on the cell. Considering the intercept with the real axis a slight
improvement is recorded for the H2 + C10H8 case. However, the shape of circles observed is
essentially the same, underlying that no interaction of tar with electrochemical performance were
observed with H2 fuel only at such low tar concentrations in the feeding fuel. In fact, the ASR value
is 0.52 cm2 with pure H2 and 0.49 cm2 adding 25 ppm(v) of C10H8. This is also in agreement to
what found in Errore. L'origine riferimento non è stata trovata.. The response was completely
different when feeding the cell with the syngas mixture, humidified at 14%, with a varying amount
of naphthalene up to 10 ppm(v). As the naphthalene concentration gets higher, a sudden voltage
drop is observed followed by stable voltage operation up to ~5 ppm(v). For concentrations above 5
ppm(v), both initial voltage drop and subsequent performance degradation becomes steep. At 10
ppm(v), the voltage drop is such to reverse the potential of the fuel cell indicating a strong re-
oxidation occurring on the Ni-anode. In fig. 3 the portion of experiment when naphthalene is added
to the synthetic syngas stream is enlarged, showing well how the cell temperature results reduced by
almost 20 °C while switching from tar-free syngas to a maximum concentration of C10H8 of about
10 ppm(v). Naphthalene is recognized to slow/inhibit steam-reforming kinetics of methane fuel.
9
However it is not clear its effect on water-gas shift reactions considering that a potential is also
applied to the electrode. Considering that the syngas contains both CH4 (2%vol.) and CO
(15%vol.), the involved chemical reactions within the Ni-YSZ anode electrode are:
𝐶𝐻4 + 𝐻2𝑂 ⇋ 3𝐻2 + 𝐶𝑂, Δ𝐻0 = 206 𝑘𝐽/𝑚𝑜𝑙
steam-reforming (SR) (1)
𝐶𝑂 + 𝐻2𝑂 ⇋ 𝐻2 + 𝐶𝑂2, Δ𝐻0 = −41 𝑘𝐽/𝑚𝑜𝑙 water-gas shift (WGS)
(2)
The temperature drop in fig. 3 when inserting C10H8 in the anode syngas stream is consistent with a
strong inhibition of the WGS reactions (Eq. 2) rather SR suppression, that would have caused an
opposite temperature behavior. The regular oscillation observed in the temperature profile is noise
from the electronic instrumentation used for data acquisition.
Figure 1 – The effect of naphthalene (C10H8) on the ASC performance when feeding an H2-
H2O only fuel and bio-syngas.
10
0510152025303501002003004005006007008009001.000182184186188190192194196198200202204206208210C10H8feed (ppmv)Cell potential (mV) Temperature ( C)Test time (h)H2feed + C10H8Syngasfeed + C10H8Clean syngasfeed ASC cellFU =15% @ 0.33 A/cm2 Figure 2 – EIS with clean H2 fuel and H2 + 25 ppm(v) of naphthalene (current density 0.33
A/cm2, temperature 750 °C).
11
0,000,050,100,150,200,250,250,300,350,400,450,500,550,60-Z Imag (ohm*cm2)Z Real (ohm*cm2)H2 - FU 21%H2 - FU 21% - C10H8 - 25ppmv
Figure 3 – The effect of naphthalene (C10H8) on ASC performance when feeding syngas. The
temperature behavior is also plotted.
In order to further understand the impact of naphthalene on the WGS reaction, and thus to analyse
what stated before, polarization experiments were run using anode fuel compositions without
methane (fig. 4). In this way, the only relevant chemical reaction occurring on the Ni-anode is the
WGS reaction (or its reverse). Being the naphthalene concentration always around few ppm(v)s, the
tar in the anode feed is always considered as a fuel impurity rather than a fuel able to provide a
meaningful amount of electricity. However, the fate of naphthalene remains unclear and should be
further investigated with gas analysis of the anode outlet. Polarization results show and confirm that
the impact of heavy tar (naphthalene in our study) does not affect significantly the fuel cell
performance under H2 fuel only. Performance are only slightly reduced with naphthalene at higher
current density, not necessarily meaning an effect of the tar. The performance is within the
variability observed among repetition of the i-V trace under same operating and fuel conditions.
12
720725730735740745750-300-100100300500700900190192194196198200202204Temperature ( C)Cell potential (mV) Test time (hr)Syngas feed + C10H8(up 10 ppm)Clean syngasfeed FU =15% @ 0.33 A/cm2Anode feed (% vol.): H2/CO/CH4/CO2/H2O/N219/17/2/8/7/479 ppmC10H810 ppmC10H8Temperature dropcasued byWGS reaction (exothermic) inhibition(a thermocouple is placed at thecell center on the anode electrode side) The performance with H2/CO 70/30 %vol. in absence of any steam was instead producing quite
different results depending on the presence of tar. In fact, naphthalene strongly reduced the cell
performance. Most notably, already the OCV is reduced, when switching from a clean H2/CO feed
to one with tar showing that, even at open circuit WGS is inhibited thus reducing the H2 partial
pressure at anode electrode. The performance with CH4 and H2 plus steam with and without the
pollutant impact of naphthalene are reported in figure 4. A strong degradation is reported especially
increasing the current density required, the reforming of methane is supposed to be limited with a
decreasing on the performance around 32%.
Figure 4 – Polarization curves with/without the presence of naphthalene under different
anode fuel compositions (pure H2, H2/CO 70/30, H2/CO 50/50 and H2/CH4).
13
02004006008001000120000,10,20,30,40,50,6Cell potential (mV)Current density (A/cm2)H2H2 + C10H8 (25 ppm)H2/CO 70/30H2/CO 70/30 + C10H8 (18 ppm)H2/CO 50/50 + steamH2/CO/ 50/50 + steam + C10H8 (10 ppm)H2/CH4 + steamH2/CH4 + steam + C10H8 (13 ppm)SyngasSyngas + C10H8 (9 ppm)FU =21% @ 0.2 A/cm2for all i-V traces.Temperature (center of the cell): 740-750 C.When present, the amount (Nmol/s or g/s) of naphthalene (C10H8) reachingthe Ni-anode is always fixed. Different concentrations only result because ofvaried anode gas compositions, and thus contaminant dilution, established in each experiment. Toluene effect on SOFC performance
Due to results achieved in the previous section toluene was tested only with syngas mixture in order
to investigate the tar effect on SOFC performance. As pointed out in figures 5 and 6, when feeding
the syngas mixture with two different amount of toluene, 3.8 and 24.2 ppm(v) the response was
completely different. When 3.8 ppm(v) of toluene was added to the syngas, the EIS behavior and
cell voltage value were slightly higher than the values recorded in clean syngas. In figure 5
(galvanostatic mode) it is shown that for almost 100 hrs the cell performance has been improved, as
the toluene acts as a fuel decreasing the FU. The cell voltage passes from almost 780 mV with clean
syngas up to 830 mV for 60 h and 820 mV stable, for 80 h with syngas adding 3.8 ppm(v) of
toluene. This voltage decreasing is related to an unstable behavior of the cell with a low toluene
concentration. Adding 24.2 ppm(v) of toluene in the syngas mixture, the cell performance shows a
strong deterioration. The cell voltage remains quite stable to 800 mV for 60 h, after that a strong
decrease is reported. A decrease of almost 0.57 mV/h is depicted. This behavior is in accordance
with the behavior recorded in case of naphthalene, demonstrating how syngas + tar (relative low
concentration) can be detrimental for SOFC applications.
14
Figure 5 – Cell voltage curves (galvanostatic mode) with/without the presence of toluene
under syngas fuel composition.
A reversible behavior for toluene was reported for concentrations below 10 ppm(v). In fact EIS
curves for clean syngas and syngas + 4 ppm(v) of C7H8 show similar behavior (figure 6) while with
the concentration around 24 ppm(v), a strong performance decrease is reported. Figure 6 describes
the EIS values for pure H2, syngas at steam to carbon of 0.4 and 0.35. Moreover it describes the EIS
behavior of syngas plus 3.8 ppm(v) and 24.2 ppm(v) of toluene. Comparing the behavior of the cell
fed by H2with clean syngas at S/C 0.35 and 0.4 a strong influence, especially on the polarization
resistance are reported. The ohmic voltage for H2 is 0.18 cm2 comparable with 0.22 cm2 for the
syngas mixture. All the syngas mixtures with and without toluene show the same ohmic value. The
ASR discrepancy for syngas (S/C 0.35) and pure hydrogen is around 38%. Increasing the steam to
carbon value the ASR increases of almost 27%. This is mainly due to the more dilution effect of
15
6006507007508008509009501.000100200300400500600Voltage (mV)Elapsed time (hr)syngas + 3,8 ppmv C7H8FU 30%syngas cleanFU 30%syngas clean FU 14%syngas + 24,2 ppmv C7H8FU 30% water decreasing the electrochemical fuel available. Adding to the syngas mixture 3.8 ppm(v) of
toluene, the AR value remains almost unchanged with the clean fuel. This demonstrates how the
cell behavior is not affected by toluene at such concentrations. Increasing the toluene concentration
the ASR increases due to problems related to the syngas direct internal reforming on the cell,
hampered by the difficulty of reforming such tar compound with a syngas mixture. A similar
behavior was recorded when it was tested naphthalene (10 ppm(v)) with a syngas mixture.
Figure 6 – EIS with clean syngas and syngas + 3.8 and + 24 ppm(v) of toluene (current density
0.33 A/cm2, temperature 750 °C).
Considering the method investigated by Papurello et al., (2016), figure 7 shows the different
contribution on the ASR value for the syngas with and without the toluene concentration [19].
16
0,000,050,100,150,200,250,300,000,200,400,600,801,001,201,40-Z Imag (ohm*cm2)Z Real (ohm*cm2)S/C 0.4S/C 0.4 + 3.8 ppmv C7H8H2 FU 21% humidifiedS/C 0.35S/C 0.35 + 24.2 ppmv C7H8 Figure 7 – EIS contribution for clean syngas and syngas + 3.8 and + 24 ppm(v) of toluene
(current density 0.33 A/cm2, temperature 750 °C).
The ohmic contribution, Re, remains stable for all the case studied. The most affected term is Rp2,
increasing the toluene from 3.8 ppm(v) to 24.2 ppm(v). This value is related to the low frequency
term that considers mass transport phenomena. Removing from the syngas mixture the ohmic
contribution remains stable, while Rp2 change.
Naphthalene catalytic study on Nickel anode compartment
The anode nickel based fuel cell have been fed with H2/CO2 and syngas mixtures at 700 °C in a
catalytic reactor (without electrochemical reactions). The goal of this section is to verify the
difficulty to complete the reverse water gas shift reaction (RWGS) (eq. 3) using naphthalene as tar
model in the gas mixture.
17
02468101203.824.20Resistance (m)C7H8concentration (ppm(v))Rp1Rp2Re 𝐶𝑂2 + 𝐻2 ⇋ 𝐻2𝑂 + 𝐶𝑂, Δ𝐻0 = 41 𝑘𝐽/𝑚𝑜𝑙 reverse water-gas shift (RWGS)
(3)
Figure 7 shows the gas concentration at the reactor outlet in case of H2/CO2 50/50 %vol. inlet
mixture at 700 °C with and without naphthalene.
In accordance with results achieved on electrochemical tests, the catalytic study showed how 25
ppm(v) of C10H8 decrease the CO and water concentration content in the reactor outlet. This
demonstrate how naphthalene strongly reduces the cell performance via the RWGS inhibition:
reaction (3) consumes H2 and CO2 to produce additional CO at clean conditions, while considering
the case with naphthalene (10 ppm(v)) the CO and H2O concentration decreases whereas increase
the concentration of hydrogen.
Figure 8 – Gas outlet concentration for H2/CO2 mixture at 700 °C.
The decreasing rate of CO and H2O is more pronounced in the syngas case where a direct internal
reforming has to be performed on nickel active sites, see figure 8. These results are in accordance
18
0102030405060COCO2CH4H2H2OGas concentration (% vol.)Inlet reactorOutlet reactor - C10H8 effectOutlet reactor - clean with those achieved during the electrochemical tests. The RWGS reaction is thus inhibited
increasing the H2 partial pressure at the anode electrode.
Figure 9 – Gas outlet concentration for syngas mixture at 700 °C.
Conclusions
Assuming the SOFC operating at a constant current density the fuel utilization depends very much
on the reforming rate. Since the steam reforming and WGS reactions occurs on the Ni catalyst
surface according, it is possible that adsorbed hydrocarbons in case of slow conversion reaction
kinetics – as it is the case of Naphthalene – they also slow electrochemical reactions by occupying
active sites of the TPB. The steps of the reforming reaction mechanism are: adsorption of methane
on the Ni surface, decomposition, surface reaction and desorption of the products from the Ni
surface. The kinetic inhibition is due to the many steps of the reforming reaction that can lead to a
decreasing reactive surface caused by the adsorption on the Ni particle. The adsorption and
desorption of naphthalene is obviously interfering with the reforming of methane and WGS (i.e.,
19
051015202530354045COCO2CH4H2H2OGas concentration (% vol.)Inlet reactorOutlet reactor - cleanOutlet reactor - C10H8 effect CO conversion to H2) by decreasing the reactive surface, and increasing the actual FU experienced
by the fuel cell. The latter fact could lead to anode re-oxidation that would be very critical as
yielding a reduced overall SOFC lifetime.
Acknowledgements
The research leading to these results has received funding from the Italian Minister for Education
(MIUR) under the national project PRIN-2009 "Analisi sperimentale ed energetico/strategica
dell'utilizzo di syngas da carbone e biomassa per l'alimentazione di celle SOFC integrate con
processo di separazione della CO2" and part of the SOFCOM project which is carried out by
Politecnico di Torino and other European partners (FCH-JU) (www.sofcom.eu).
20
References
[1]
Papurello D, Lanzini A, Tognana L, Silvestri S, Santarelli M. Waste to energy: Exploitation
of biogas from organic waste in a 500 Wel solid oxide fuel cell (SOFC) stack. Energy
2015;85:145–58. doi:10.1016/j.energy.2015.03.093.
[2] Hauth M, Lerch W, K??nig K, Karl J. Impact of naphthalene on the performance of SOFCs
during operation with synthetic wood gas. J Power Sources 2011;196:7144–51.
doi:10.1016/j.jpowsour.2010.09.007.
[3] Namioka T, Naruse T, Yamane R. Behavior and mechanisms of Ni/ScSZ cermet anode
deterioration by trace tar in wood gas in a solid oxide fuel cell. Int J Hydrogen Energy
2011;36:5581–8. doi:10.1016/j.ijhydene.2011.01.165.
[4] Hofmann P, Panopoulos KD, Fryda LE, Schweiger A, Ouweltjes JP, Karl J. Integrating
biomass gasification with solid oxide fuel cells: Effect of real product gas tars, fluctuations
and particulates on Ni-GDC anode. Int J Hydrogen Energy 2008;33:2834–44.
doi:10.1016/j.ijhydene.2008.03.020.
[5] Hofmann P, Panopoulos KD, Aravind P V., Siedlecki M, Schweiger A, Karl J, et al.
Operation of solid oxide fuel cell on biomass product gas with tar levels >10 g Nm-3. Int J
Hydrogen Energy 2009;34:9203–12. doi:10.1016/j.ijhydene.2009.07.040.
[6]
Lorente E, Millan M, Brandon NP. Use of gasification syngas in SOFC: Impact of real tar on
anode materials. Int J Hydrogen Energy 2012;37:7271–8.
doi:10.1016/j.ijhydene.2011.11.047.
[7] Aravind P V., Ouweltjes JP, Woudstra N, Rietveld G. Impact of Biomass-Derived
Contaminants on SOFCs with Ni/Gadolinia-Doped Ceria Anodes. Electrochem Solid-State
Lett 2008;11:B24. doi:10.1149/1.2820452.
[8]
Papurello D, Lanzini A, Fiorilli S, Smeacetto F, Singh R, Santarelli M. Sulfur poisoning in
Ni-anode solid oxide fuel cells (SOFCs): Deactivation in single cells and a stack. Chem Eng
J 2016;283:1224–33. doi:10.1016/j.cej.2015.08.091.
[9] Madi H, Lanzini A, Diethelm S, Papurello D, Van herle J, Lualdi M, et al. Solid oxide fuel
cell anode degradation by the effect of siloxanes. J Power Sources 2015;279:460–71.
doi:10.1016/j.jpowsour.2015.01.053.
[10] Papurello D, Borchiellini R, Bareschino P, Chiodo V, Freni S, Lanzini A, et al. Performance
of a Solid Oxide Fuel Cell short-stack with biogas feeding. Appl Energy 2014;125:254–63.
21
[11] Mermelstein J, Millan M, Brandon N. The impact of steam and current density on carbon
formation from biomass gasification tar on Ni/YSZ, and Ni/CGO solid oxide fuel cell
anodes. J Power Sources 2010;195:1657–66. doi:10.1016/j.jpowsour.2009.09.046.
[12] Aravind P V., De Jong W. Evaluation of high temperature gas cleaning options for biomass
gasification product gas for Solid Oxide Fuel Cells. Prog Energy Combust Sci 2012;38:737–
64. doi:10.1016/j.pecs.2012.03.006.
[13] Phuphuakrat T, Nipattummakul N, Namioka T, Kerdsuwan S, Yoshikawa K.
Characterization of tar content in the syngas produced in a downdraft type fixed bed
gasification system from dried sewage sludge. Fuel 2010;89:2278–84.
doi:10.1016/j.fuel.2010.01.015.
[14] Liu M, van der Kleij A, Verkooijen AHM, Aravind P V. An experimental study of the
interaction between tar and SOFCs with Ni/GDC anodes. Appl Energy 2013;108:149–57.
doi:10.1016/j.apenergy.2013.03.020.
[15] Lorente E, Berrueco C, Millan M, Brandon NP. Effect of tar fractions from coal gasification
on nickel-yttria stabilized zirconia and nickel-gadolinium doped ceria solid oxide fuel cell
anode materials. J Power Sources 2013;242:824–31. doi:10.1016/j.jpowsour.2013.05.158.
[16] Liu M, Millan-Agorio MG, Aravind P V, Brandon NP. Influence of Operation Conditions on
Carbon Deposition in SOFCs Fuelled by Tar-Containing Biosyngas. ECS Trans 2011;35
:2701–12. doi:10.1149/1.3570269.
[17] Gerdes K. Exposure Limits of Higher Hydrocarbons for SOFC. Fuel Cells 2010.
[18] Guerra C, Lanzini A, Leone P, Santarelli M, Brandon NP. Optimization of dry reforming of
methane over Ni/YSZ anodes for solid oxide fuel cells. J Power Sources 2014;245:154–63.
doi:10.1016/j.jpowsour.2013.06.088.
[19] Papurello D, Lanzini A, Drago D, Leone P, Santarelli M. Limiting factors for planar solid
oxide fuel cells under different trace compound concentrations. Energy 2016;95:67–78.
doi:10.1016/j.energy.2015.11.070.
22
|
1704.07918 | 1 | 1704 | 2017-04-25T22:05:55 | Rectangular Photonic Crystal Nanobeam Cavities in Bulk Diamond | [
"physics.app-ph",
"cond-mat.mtrl-sci",
"physics.atom-ph",
"physics.optics",
"quant-ph"
] | We demonstrate the fabrication of photonic crystal nanobeam cavities with rectangular cross section into bulk diamond. In simulation, these cavities have an unloaded quality factor (Q) of over 1 million. Measured cavity resonances show fundamental modes with spectrometer-limited quality factors larger than 14,000 within 1nm of the NV center's zero phonon line at 637nm. We find high cavity yield across the full diamond chip with deterministic resonance trends across the fabricated parameter sweeps. | physics.app-ph | physics | Rectangular Photonic Crystal Nanobeam Cavities in Bulk Diamond
Sara Mouradian,1, ∗ Noel H. Wan,1, ∗ Tim Schroder,2 and Dirk Englund1, †
1Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge MA 02139
2Niels Bohr Institute, University of Copenhagen, 2100 Copenhagen, Denmark
We demonstrate the fabrication of photonic crystal nanobeam cavities with rectangular cross
section into bulk diamond. In simulation, these cavities have an unloaded quality factor (Q) of over
1 × 106. Measured cavity resonances show fundamental modes with spectrometer-limited quality
factors ≥ 14 × 103 within 1 nm of the NV center's zero phonon line at 637 nm. We find high
cavity yield across the full diamond chip with deterministic resonance trends across the fabricated
parameter sweeps.
7
1
0
2
r
p
A
5
2
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
1
v
8
1
9
7
0
.
4
0
7
1
:
v
i
X
r
a
A central aim of quantum information science is the
efficient generation of large entangled states of station-
ary quantum memories with high-fidelity single and two-
qubit gates. Among solid-state qubits, a leading system
consists of the nitrogen vacancy (NV) center in diamond.
Recently, entanglement between distant NV nodes has
been demonstrated via single-photon measurements of
the zero phonon line (ZPL) emission [1, 2]. Such her-
alded entanglement can be used to build large cluster
states which are a resource for universal quantum compu-
tation [3] or quantum repeater networks [4, 5]. However,
the coherent ZPL optical transition of the NV accounts
for only 3% of the emission due to phonon interactions
even at cryogenic temperatures. This severely limits the
entanglement rate, even with broadband collection en-
hancement structures such as a solid immersion lens [2].
To improve the entanglement rate, the collection rate
into the desired frequency and spatial mode must be in-
creased. Much recent work has focused on maximizing
the coherent ZPL collection efficiency via the Purcell ef-
fect in a photonic crystal nanocavity.
One major challenge in coupling the NV to a nanocav-
ity is that high-quality diamond cannot currently be
grown in thin (wavelength-scale) waveguiding mem-
branes. Hybrid structures have been used to enhance the
ZPL emission [6–8], though the maximum enhancement
rate is limited as the NV must be placed out of the cav-
ity's mode maximum. Advances in diamond patterning
have enabled the fabrication of photonic crystal cavities
in diamond [9] with two main methods: fabrication into
thinned diamond membranes [10–17] and angled etching
of bulk diamond [18–21]. However, the measured quality
(Q) factors have been limited to a few thousand near the
NV ZPL. Low cavity yield or inconsistency across the di-
amond substrate also presents a major scaling challenges,
especially for membrane-based approaches.
Recently, Khanaliloo et al. [22, 23] introduced a tech-
nique to undercut lithographically defined structures in
bulk diamond, allowing for the fabrication of high qual-
ity free-standing diamond optomechanical devices.
In
this Letter, we apply a similar isotropic undercut pro-
∗ S.M. and N.H.W. contributed equally to this work
† [email protected]
cess to the fabrication of photonic crystal nanobeam cav-
ities. Figure 1(a) shows a schematic of the rectangu-
lar nanobeam photonic crystal fabricated from bulk dia-
mond. We find that the quasi-isotropic etching procedure
is highly repeatable and consistent across the full chip. In
addition, the process requires minimal fabrication opti-
mization as electron-beam lithography determines all pa-
rameters except for the nanobeam height, which can be
precisely tuned during the relatively slow, isotropic etch-
ing. This fabrication process enables instrument-limited
optical quality (Q) factors exceeding 14,000 within 1 nm
of the NV− center ZPL wavelength of 637 nm, as well as
uniform nanocavity fabrication across a full chip.
FIG. 1. (a) Schematic of an array of rectangular nanobeam
cavities fabricated from bulk diamond. (b) Re(E) for the first
order TE mode. (c) Re(E) for the first order TM mode. Scale
bar for (b) and (c): 1 µm
We design photonic crystal nanobeam cavities to sup-
port a low-mode-volume (V ) mode with high quality fac-
tor Q at the NV ZPL (637 nm), and with the electric-
field maximum concentrated in the diamond. The design
process begins with approximate cavity parameters de-
rived from band structure simulations and optimizes the
cavity Q of the lowest order TE mode (intensity pro-
file shown in Figure 1(b)) at λN V = 637 nm by Finite
Difference Time Domain (FDTD) simulations. The final
design consists of a diamond waveguide (W = 250 nm
(a)(c)(b)HWa2(cid:83) 2
FIG. 2.
(a-f) Fabrication steps for rectangular nanobeam photonic crystal cavities, where white, blue, and gray indicate
bulk diamond, SiN, and Al2O3, respectively. (a) Anisotropic etch into diamond with SiN hard mask following electron-beam
lithography. (b) Atomic layer deposition of Al2O3.
(d) Anisotropic etch
into diamond (e) Quasi-isotropic etch of diamond nanobeam cavity (f) Suspended nanobeam cavity following mask removal.
Scanning electron micrographs (g),(h),(i) correspond to processes (a),(e) and (f), respectively. Scale bar: 1 µm (j) Overview of
an array of photonic crystal nanobeam cavities. Scale bar: 10 µm
(c) Selective removal of the top layer of Al2O3.
and H = 230 nm) periodically patterned with holes with
radius r = 58 nm and spacing a = 192 nm. The defect
supporting the cavity mode is introduced by linearly de-
creasing the hole spacing to a = 171 nm over 5 periods.
With 25 holes on either side of the cavity region, this cav-
ity has a radiation-limited Q factor of > 1 million. The
cavity also supports a TM mode (Figure 1(c) shows the
intensity profile) at 615 nm with a Q of 13,000 in simu-
lation, as well as other TE modes with mode maxima in
the outer hole regions with lower Q values – 5,500 and
2,500 in simulation at 649 nm and 653 nm respectively.
These higher order modes are denoted as M3 and M4 in
the measured spectra in Figures 3 and 4.
We fabricated these nanobeam cavity designs in
a 3 mm×3 mm×0.3 mm single-crystal diamond with a
{100} top face grown by chemical vapor deposition
(CVD, Element6) with a nitrogen defect density of less
than 1 ppm, and thus a native NV density of approx-
imately 1ppb. High-resolution X-ray diffraction con-
firmed the crystal orientation to align the nanobeams
along the facet with the fastest etch rate ({110}), which
is 45◦ from the diamond chip facets on all measured dia-
mond, as illustrated in Fig 2(j). The quasi-isotropic etch
rate is facet-dependent [23], and ensures that the beam
has a rectangular cross-section. Unlike nanobeam cavi-
ties with triangular cross sections which mix the TE and
TM modes, these cavities have low out-of-plane scatter-
ing loss [18, 20].
Fig. 2(a-f) outlines the essential steps of the fabrication
process. A 180 nm-thick low-stress SiN layer, deposited
with plasma-enhanced chemical vapor deposition, func-
tions as a hard mask to pattern the diamond. We ob-
served a selectivity of approximately 30:1 for the oxygen
anisotropic etch parameters described below. Electron-
beam lithography defines the nanobeam cavities (ZEP
520A exposed at 500 µC/cm2 and developed at 0◦C in
ortho-xylene for 90 s). Following resist development, a
tetrafluoromethane (CF4) plasma reactive-ion etch step
transfers the pattern into the the SiN hard mask. This
pattern is transferred into the diamond (Figure 2(a) us-
ing an inductively coupled oxygen plasma (ICP) with a
working pressure of 0.15 Pa with ICP and RF powers of
(a)(b)(c)(d)(e)(f)chipaxes[110]plane(j)(g)(h)(i) 3
FIG. 3. (a) PL spectrum of Cavity A with lowest order TE and TM cavity peaks fit with Lorentzian functions. (b) Zoom in of
the fundamental TE mode of Cavity A. The red line shows a fit to the spectrometer-limited cavity peak, fit with a pseudo-Voigt
function. (c) PL spectrum of Cavity B showing the inhomogeneously broadened ZPL emission from the ensemble of excited
NVs at 637 nm and a cavity peak at 637.1 nm. (d) Photoluminescence excitation (PLE) spectrum of Cavity B showing a PLE
peak at NV ZPL position and a cavity-enhanced PLE peak 100 GHz detuned center from the inhomogeneous ZPL distribution.
The inset shows a PLE measurement of the same nanobeam 7 µm from the cavity center showing only the inhomogeneous
distribution of NV centers in the sample.
500 W and 240 W, respectively. This anistropic etch step
is 2.5 times as deep as the final desired cavity height for
precise tuning of the nanobeam height in the subsequent
isotropic undercut etch step. This step produces smooth
and straight sidewalls as seen in the scanning electron
micrograph (SEM) in Fig. 2(g).
A conformal
layer of 20 nm of aluminum oxide
(Al2O3) produced by atomic layer deposition (ALD), see
Fig. 2(b), protects all sides of the nanobeam cavity for
the subsequent etch steps. A CF4 reactive ion etch re-
moves the top Al2O3 layer, leaving only the sides covered,
as shown in Fig. 2(c). A second anisotropic oxygen etch,
using the same parameters as above, removes an addi-
tional 1 µm of diamond, as shown in Fig. 2(d). A quasi-
isotropic etch then undercuts the nanobeam structure at
200◦C and 3 Pa with 900 W ICP and no forward bias
(Fig. 2(e)). The elevated temperature and pressure in-
crease the chemical interaction rate with the diamond to
increase the etch rate. As seen in the SEM in Fig. 2(h),
the Al2O3 is thin enough to allow periodic SEM mea-
surements of the nanobeam height. Once the desired
nanobeam height of 230 nm (∼85 minutes with an intial
beam height of 575 nm) and undercut are achieved, the
residual SiN and Al2O3 are removed using 49% hydroflu-
oric acid (Fig. 2(f)). An SEM of the final structure after
mask removal is shown in Fig. 2(i). The measured chip
contains 125 cavities, with 5 copies at each of 25 param-
eters. The parameter sweep modified the beam widths
and hole radii by ± 4 % and ± 16 % respectively, to cover
a large wavelength range.
We measured the fabricated cavities at 4 K by pho-
toluminescence (PL) and photoluminescence excitation
(PLE) spectroscopy. The native population of NV cen-
ters, excited using a 532 nm laser, provides an internal
light source for the cavity. The cavity PL is collected
with a 0.9 NA objective and resolved on a spectrometer
with a resolution of 0.06 nm (Q ∼14,000). Fig. 3 shows
two cavity spectra: Cavity A (W = 260 nm, r = 55 nm)
625630635640645Intensity (arb)0.51.52650NV ZPL(a)116366372Wavelength (nm)(b)Frequency (THz)-2000Intensity (cps)345610020034567x10-4Intensity (cps)050150Frequency (GHz)(c)7100200-50-100-150-200-100Q > 14,700Q > 14,000Wavelength (nm)TETMM3M412108642Intensity (arb)Wavelength (nm)637637.2636.8(d) and Cavity B (W = 245 nm, r = 58 nm). Fig. 3(a) shows
the cavity-modified PL spectrum of the NV ensemble
at Cavity A under 532 nm excitation. Four modes ap-
pear in the spectrum, corresponding to the first-order
TE and TM modes (profiles shown in Figure 1(b,c)), as
well as the higher-order modes, M3 and M4. Fig. 3(b)
shows the high-Q first-order TE mode at 636.1 nm, as
well as the inhomogeneously-broadened ZPL of the ex-
cited ensemble of NV centers at 637 nm. The linewidth
of the cavity mode is limited by the spectrometer's reso-
lution, as confirmed by comparing to the spectrum of a
sub-0.5 MHz laser at the same wavelength. Fitting the
spectrum with a pseudo-Voigt function [24] takes into
account the effect of the Gaussian spectrometer result
on the Lorentzian cavity spectrum. This fit indicates a
lower bound of Q ≥ 16700, above the resolution of the
spectrometer (Q ≥ 14, 000).
The spectrum of Cavity B (Fig. 3(c)) reveals the first-
order TE mode overlapping with the inhomogenous dis-
tribution of ZPLs in the NV ensemble. We performed
PLE spectroscopy to better measure the Q of this cavity
and study the cavity-enhanced excitation of the NV en-
semble. The PLE consisted of NV phonon side band
(PSB) detection (filtered > 650 nm) while scanning a
< 0.5 MHz-linewidth laser over the NV ZPL and cavity
frequencies. The inset of Fig. 3(c) shows the PLE spec-
trum of the unpatterned nanobeam 7 µm from the cavity
center. This reveals an inhomogeneously broadened ab-
sorption spectrum of the excited population of NV cen-
ters centered at 470.48 THZ (637.2 nm) with a full width
half maximum of 42.7 GHz. A PLE scan at the center of
the cavity shows a second peak that is absent on the rest
of the sample; we attribute this to the cavity-enhanced
absorption of NV centers coupled to the cavity mode. A
Lorentzian fit of the peak reveals a measured cavity Q of
14,700. This provides a lower bound on the Q of the bare
cavity mode, as the PLE spectrum may be broadened by
the interaction with the inhomogeneous distribution of
emitters coupled to this mode of the cavity [25].
A survey of all of the fabricated devices demonstrates
the consistency and high yield of our fabrication tech-
nique. Fig. 4(a) shows the measured cavity resonances
of mode M4 across the full range of parameters. The
error bars show the standard deviation of the wavelength
position for the 5 cavities fabricated with each parameter
set. We used here mode M4 as it has the highest vertical
loss of the four modes considered here, and thus the
highest SNR in spectrometer measurements. This survey
reveals the expected trends: the resonance wavelengths
increase with larger beam widths and decrease with
larger holes radii. The standard deviation within each
parameter is low with an average of ± 2.2 nm deviation
from the mean in each parameter set, showing the
consistency of the fabrication process. The spectra of
M3 and M4 of the 5 cavities with W = 260 nm and
r = 58 nm (circled in Fig. 4(a) are shown in Fig. 4(b).
4
FIG. 4. (a) Frequency distribution of resonant frequencies of
Mode 4 for all 25 sets of parameters. The error bars show the
standard deviation ±σ of the resonant frequencies for the 5
cavities fabricated for each set of parameters. (b) The spectra
of the 5 cavities fabricated with W = 260 nm and r = 58 nm
(circled in (a)) showing the distribution of M3 and M4.
In summary, we have demonstrated a method to fab-
ricate high-Q photonic crystal nanobeam cavities from
bulk diamond at the NV ZPL wavelength. We measured
cavity resonances within 1 nm of the NV ZPL wavelength,
with instrument-limited Q factors larger than 14,000.
The process showed consistent cavity properties across
all fabricated devices. Future work will apply this fab-
rication process to high-purity diamond with a nitrogen
concentration below 100 ppb, which should enable the
coupling of individual NV ZPLs with high Purcell en-
hancement. The rectangular cross section of these cavi-
ties should enable efficient mode conversion between di-
amond waveguides and on-chip ridge or channel waveg-
uides [26]. This fabrication technique applies to other
diamond color centers, such as the germanium vacancy
and silicon vacancy centers. These emitters have natu-
rally narrow emission linewidth [27] so that the nanocav-
ity parameters achieved here should allow for the strong
coupling regime of cavity quantum electrodynamics.
ACKNOWLEDGMENTS
This research was supported in part by the Army
Research Laboratory Center
for Distributed Quan-
tum Information (CDQI). S.M. was supported in part
by the NSF IQuISE program and the NSF program
ACQUIRE:"Scalable Quantum Communications with
Error-Corrected Semiconductor Qubits." N.W was sup-
ported by CDQI.
Intensity (arb)12108642Wavelength (nm)635645240245250255260Width (nm)Wavelength (nm)610620630640650660(a)(b)NV ZPL 5
[1] Emre Togan, Yiwen Chu, AS Trifonov, Liang Jiang,
Jeronimo Maze, Lilian Childress, MV Gurudev Dutt, An-
ders Søndberg Sørensen, PR Hemmer, AS Zibrov, et al.,
"Quantum entanglement between an optical photon and
a solid-state spin qubit," Nature 466, 730–734 (2010).
[2] H Bernien, B Hensen, W Pfaff, G Koolstra, M S Blok,
L Robledo, T H Taminiau, M Markham, D J Twitchen,
L Childress, and R Hanson, "Heralded entanglement be-
tween solid-state qubits separated by three metres," Na-
ture 497, 86–90 (2013).
[3] Sean D Barrett and Pieter Kok, "Efficient high-fidelity
quantum computation using matter qubits and linear op-
tics," Physical Review A 71, 060310 (2005).
[4] L Childress, JM Taylor, Anders Søndberg Sørensen,
and Mikhail D Lukin, "Fault-tolerant quantum repeaters
with minimal physical resources and implementations
based on single-photon emitters," Physical Review A 72,
052330 (2005).
[5] Scott E Vinay and Pieter Kok, "Practical repeaters
for ultra-long distance quantum communication," arXiv
preprint arXiv:1607.08140 (2016).
[6] Michael Gould, Emma R Schmidgall, Shabnam Dad-
gostar, Fariba Hatami,
and Kai-Mei C Fu, "Effi-
cient extraction of zero-phonon-line photons from sin-
gle nitrogen-vacancy centers in an integrated gap-on-
diamond platform," Physical Review Applied 6, 011001
(2016).
[7] Dirk Englund, Brendan Shields, Kelley Rivoire, Fariba
Hatami, Jelena Vuckovic, Hongkun Park, and Mikhail D
Lukin, "Deterministic coupling of a single nitrogen va-
cancy center to a photonic crystal cavity," Nano letters
10, 3922–3926 (2010).
[8] Janik Wolters, Andreas W Schell, Gunter Kewes, Nils
Nusse, Max Schoengen, Henning Doscher, Thomas Han-
nappel, Bernd Lochel, Michael Barth, and Oliver Ben-
son, "Enhancement of the zero phonon line emission from
a single nitrogen vacancy center in a nanodiamond via
coupling to a photonic crystal cavity," Applied Physics
Letters 97, 141108 (2010).
[9] Tim Schroder, Sara L Mouradian, Jiabao Zheng,
Matthew E Trusheim, Michael Walsh, Edward H Chen,
Luozhou Li, Igal Bayn, and Dirk Englund, "Quantum
nanophotonics in diamond [invited]," JOSA B 33, B65–
B83 (2016).
[10] Andrei Faraon, Paul E Barclay, Charles Santori, Kai-
Mei C Fu, and Raymond G Beausoleil, "Resonant en-
hancement of the zero-phonon emission from a colour
centre in a diamond cavity," Nature Photonics 5, 301–
305 (2011).
[11] Andrei Faraon, Charles Santori, Zhihong Huang, Vic-
tor M Acosta, and Raymond G Beausoleil, "Coupling of
nitrogen-vacancy centers to photonic crystal cavities in
monocrystalline diamond," Physical review letters 109,
033604 (2012).
[12] BJM Hausmann, BJ Shields, Q Quan, Y Chu,
NP De Leon, R Evans, MJ Burek, AS Zibrov,
M Markham, DJ Twitchen, et al., "Coupling of nv cen-
ters to photonic crystal nanobeams in diamond," Nano
letters 13, 5791–5796 (2013).
Andrich, Benjamin Alem´an, Kasey J Russell, Andrew P
Magyar, et al., "Deterministic coupling of delta-doped
nitrogen vacancy centers to a nanobeam photonic crystal
cavity," Applied Physics Letters 105, 261101 (2014).
[14] Luozhou Li, Tim Schroder, Edward H Chen, Michael
Walsh, Igal Bayn, Jordan Goldstein, Ophir Gaathon,
Matthew E Trusheim, Ming Lu, Jacob Mower, et al.,
"Coherent spin control of a nanocavity-enhanced qubit
in diamond," Nature communications 6 (2015).
[15] Janine Riedrich-Moller, Carsten Arend, Christoph Pauly,
Frank Mucklich, Martin Fischer, Stefan Gsell, Matthias
Schreck, and Christoph Becher, "Deterministic coupling
of a single silicon-vacancy color center to a photonic
crystal cavity in diamond," Nano letters 14, 5281–5287
(2014).
[16] Janine Riedrich-Moller, Laura Kipfstuhl, Christian
Hepp, Elke Neu, Christoph Pauly, Frank Mucklich,
Armin Baur, Michael Wandt, Sandra Wolff, Martin Fis-
cher, Stefan Gsell, Matthias Schreck,
and Christoph
Becher, "One- and two-dimensional photonic crystal mi-
crocavities in single crystal diamond," Nature Nanotech-
nology 7, 69–74 (2012).
[17] Luozhou Li, Igal Bayn, Ming Lu, Chang-Yong Nam,
and
Tim Schroder, Aaron Stein, Nicholas C Harris,
Dirk Englund, "Nanofabrication on unconventional sub-
strates using transferred hard masks," Scientific reports
5 (2015).
[18] Igal Bayn, Boris Meyler, Joseph Salzman,
and Rafi
Kalish, "Triangular nanobeam photonic cavities in single-
crystal diamond," New Journal of Physics 13, 025018
(2011).
[19] Igal Bayn, Sara Mouradian, Luozhou Li, Jordan A.
Goldstein, Tim Schroder, Jiabao Zheng, Ed H. Chen,
Ophir Gaathon, Ming Lu, Aaron Stein, Carl A. Ruggiero,
J Salzman, R Kalish, and Dirk Englund, "Fabrication
of triangular nanobeam waveguide networks in bulk di-
amond using single-crystal silicon hard masks," Applied
Physics Letters 105, 211101 (2014).
[20] Michael J Burek, Yiwen Chu, Madelaine SZ Liddy, Parth
Patel, Jake Rochman, Srujan Meesala, Wooyoung Hong,
Qimin Quan, Mikhail D Lukin,
and Marko Loncar,
"High quality-factor optical nanocavities in bulk single-
crystal diamond," Nature communications 5 (2014).
[21] M Schukraft, J Zheng, T Schroder, SL Mouradian,
M Walsh, ME Trusheim, H Bakhru, and DR Englund,
"Invited article: Precision nanoimplantation of nitrogen
vacancy centers into diamond photonic crystal cavities
and waveguides," APL Photonics 1, 020801 (2016).
[22] Behzad Khanaliloo et al., "High-q/v monolithic diamond
microdisks fabricated with quasi-isotropic etching," Nano
letters 15, 5131–5136 (2015).
[23] Behzad Khanaliloo et al., "Single-crystal diamond
nanobeam waveguide optomechanics," Physical Review
X 5, 041051 (2015).
[24] Roland Albrecht, Alexander Bommer, Christian
Deutsch, Jakob Reichel, and Christoph Becher, "Cou-
pling of a single nitrogen-vacancy center in diamond to
a fiber-based microcavity," Physical review letters 110,
243602 (2013).
[13] Jonathan C Lee, David O Bracher, Shanying Cui,
Kenichi Ohno, Claire A McLellan, Xingyu Zhang, Paolo
[25] Daniel Valente, Jan Suffczy´nski, Tomasz Jakubczyk,
Adrien Dousse, Aristide Lemaıtre, Isabelle Sagnes, Loıc
Lanco, Paul Voisin, Alexia Auff`eves, and Pascale Senel-
lart, "Frequency cavity pulling induced by a single semi-
conductor quantum dot," Physical Review B 89, 041302
(2014).
[26] Sara L Mouradian, Tim Schroder, Carl B Poitras, Lu-
ozhou Li, Jordan Goldstein, Edward H Chen, Michael
Walsh, Jaime Cardenas, Matthew L Markham, Daniel J
Twitchen, et al., "Scalable integration of long-lived quan-
tum memories into a photonic circuit," Physical Review
X 5, 031009 (2015).
[27] Tim Schroder, Matthew E Trusheim, Michael Walsh,
Luozhou Li, Jiabao Zheng, Marco Schukraft, Jose L
Pacheco, Ryan M Camacho, Edward S Bielejec, Alp
Sipahigil, et al., "Scalable focused ion beam creation of
nearly lifetime-limited single quantum emitters in dia-
mond nanostructures," arXiv preprint arXiv:1610.09492
(2016).
6
|
1903.07818 | 3 | 1903 | 2019-12-18T02:02:52 | Orbital Angular Momentum Waves: Generation, Detection and Emerging Applications | [
"physics.app-ph"
] | Orbital angular momentum (OAM) has aroused a widespread interest in many fields, especially in telecommunications due to its potential for unleashing new capacity in the severely congested spectrum of commercial communication systems. Beams carrying OAM have a helical phase front and a field strength with a singularity along the axial center, which can be used for information transmission, imaging and particle manipulation. The number of orthogonal OAM modes in a single beam is theoretically infinite and each mode is an element of a complete orthogonal basis that can be employed for multiplexing different signals, thus greatly improving the spectrum efficiency. In this paper, we comprehensively summarize and compare the methods for generation and detection of optical OAM, radio OAM and acoustic OAM. Then, we represent the applications and technical challenges of OAM in communications, including free-space optical communications, optical fiber communications, radio communications and acoustic communications. To complete our survey, we also discuss the state of art of particle manipulation and target imaging with OAM beams. | physics.app-ph | physics | Orbital Angular Momentum Waves: Generation,
Detection and Emerging Applications
Rui Chen, Member, IEEE, Hong Zhou, Marco Moretti, Member, IEEE, Xiaodong Wang, Fellow, IEEE, and
Jiandong Li, Senior Member, IEEE
9
1
0
2
c
e
D
8
1
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
3
v
8
1
8
7
0
.
3
0
9
1
:
v
i
X
r
a
Abstract -- Orbital angular momentum (OAM) has aroused a
widespread interest in many fields, especially in telecommu-
nications due to its potential for unleashing new capacity in
the severely congested spectrum of commercial communication
systems. Beams carrying OAM have a helical phase front and
a field strength with a singularity along the axial center, which
can be used for information transmission, imaging and particle
manipulation. The number of orthogonal OAM modes in a single
beam is theoretically infinite and each mode is an element of a
complete orthogonal basis that can be employed for multiplexing
different signals, thus greatly improving the spectrum efficiency.
In this paper, we comprehensively summarize and compare the
methods for generation and detection of optical OAM, radio
OAM and acoustic OAM. Then, we represent the applications
and technical challenges of OAM in communications, including
free-space optical communications, optical fiber communications,
radio communications and acoustic communications. To complete
our survey, we also discuss the state of art of particle manipu-
lation and target imaging with OAM beams.
Index Terms -- Orbital angular momentum (OAM), optical,
radio, acoustic, vortex, generation, detection, communications,
particle manipulation, imaging.
I. INTRODUCTION
E Lectromagnetic waves carry both energy and momentum,
where momentum comprises linear momentum P and
angular momentum L. In particular, angular momentum has
an additive component linked to polarization, spin angular
momentum (SAM), and another one associated with spatial dis-
tribution, which is called orbital angular momentum (OAM).
The relationship between SAM and OAM can be explained by
referring to the model of electron rotation around the nucleus:
the momentum generated by the circular motion of electrons
around the nucleus is equivalent to OAM, and the momentum
generated by the spin of electrons is equivalent to SAM.
In 1992, Allen et al. first combined the concept of OAM
with the idea of optical vortex [1]:
in an optical vortex
the planes of constant phase of the electric and magnetic
vector fields form a corkscrew or helicoid running in the
direction of propagation. The vortex is characterized by a
This work was supported in part by the Fundamental Research Funds for the
Central Universities (JB180109), the 111 Project (B08038) and the University
of Pisa under the PRA 2018-2019 Research Project CONCEPT.
Rui Chen, Hong Zhou and Jiandong Li are with the State Key Laboratory of
ISN, Xidian University, Shaanxi 710071, China (e-mail: [email protected],
hzhou [email protected], [email protected]). Rui Chen is also with
the Science and Technology on Communication Networks Laboratory.
Marco Moretti is with the University of Pisa, Dipartimento di Ingegneria
dell'Informazione, Italy (e-mail: [email protected]).
Xiaodong Wang is with the Electrical Engineering Department, Columbia
University, New York, NY 10027 USA (e-mail: [email protected]).
number, called the topological charge, which indicates the
number of twists the light does in one wavelength. The
larger the number of twists, the faster the light is rotating
around the axis. Accordingly, the OAM carried by the optical
vortex theoretically has an infinite number of eigenstates and
is defined in an infinite-dimensional Hilbert vector space.
Because of this, the application potential of OAM in the field
of communications are enormous, even if there are still some
problems to be solved before a full deployment. If the OAM
dimension of photons can be fully utilized for information
modulation or multiplexing,
the information capacity of a
single photon can be significantly improved, thereby leading to
an increase of the transmission capacity of single-wavelength
and single-mode fibers. In addition, since the vortex beam has
a helical wavefront, its axial center field in the direction of
propagation is null, creating the potential for applications in
particle manipulation and imaging.
The potential of employing OAM for communications are
not restricted to electromagnetic waves at light frequencies. In
2007, Thid´e et al. [2] proposed to apply the concept of optical
vortex to the field of wireless communications, i.e., in a range
of lower radio frequencies than light. Moreover, a different line
of research has showed that, unlike electromagnetic waves,
sound waves do not have polarization or spin effects and
cannot carry SAM but only carry OAM. Although the concept
of acoustic vortex was first proposed in 1979 [3], twenty
years later, Hefner and Marston [4] derived the relationship
between sound pressure and angular momentum, perfecting
the theory of the acoustic vortex. The application of OAM
to wireless communications and acoustic communications,
especially underwater communications, is expected to open
new fields of research and possibly break the limits of existing
communication systems.
A. Contribution with respect to existing literature
This paper is devoted to reviewing the research progress on
the use of OAM waves for communications in the last three
decades. Our survey includes optical, radio and acoustic OAM
generation, detection and applications in communications, as
well as the applications of OAM waves in particle manipu-
lation and imaging. Main technical challenges are addressed
and potential solutions are analyzed. The key contributions of
this survey can be summarized as
• The most common methods for generating and detecting
optical, radio and acoustic OAM waves are introduced
and compared from multiple perspectives.
2
• Recent advances in OAM applications in free-space
optical communications, optical fiber communications,
radio communications and acoustic communications are
carefully reviewed, so that the main technical problems
and their potential solutions are discussed in details.
Indeed, integrating OAM into existing communication
systems presents some very serious challenges such as
OAM beam divergence, misalignment between transmit-
ter and receiver, atmospheric turbulence effects in free-
space optical links, mode coupling in fiber links, and
multipath effects in radio communication links. These
effects need to be carefully considered in link design.
• Since OAM has been considered as a potential solution
for precise particle manipulation and target imaging, the
application prospects of OAM in these fields are also
briefly discussed.
This is not the first survey on OAM waves, since there are
other works on the subject, such as [5], [6], and [7]. However,
[5] is more inclined to discuss the physical properties of OAM
waves, and [6] addresses only the application of OAM multi-
plexing in free-space optical and radio communications. The
work in [7] surveys both generation/detection and application
of OAM beams but mainly for optical communications and
there is no mention of the potential of OAM in acoustics.
Compared with these surveys, our work has the merit of
comprehensively and consistently discussing the application of
OAM waves for optical, radio and acoustic communications.
For each system we present a complete and accurate review
of the generation/detection methods, their application and the
most pressing technical challenges.
B. Paper Organization
The rest of the paper is organized as follows. Section
II describes the main principles of OAM, introducing the
basic characteristics of OAM waves and their application
potential. Subsequently, the generation and detection methods
for optical, radio and acoustic OAM communication systems
are summarized and compared in Section III and Section IV,
respectively. Section V focuses on the most recent advances in
OAM applications to free-space optical, optical fiber, radio and
acoustic communications including the corresponding techni-
cal challenges and potential solutions. Atmospheric turbulence
effects in free-space optical links, mode coupling in fiber
links, and multipath effects in radio communication links are
all significant practical obstacles to the implementation of
OAM-based communication systems. In addition, OAM beam
divergence and misalignment also need to be considered in
link design. Section VI discusses the application of OAM
in other areas, including particle manipulation and imaging.
Conclusions and perspectives then follow in Section VII.
II. OAM WAVES AND THEIR POTENTIAL
The angular momentum of the electromagnetic field asso-
ciated with a volume V can be expressed as [8]
(cid:90)
J =
ε0r × Re{E × B∗}dV,
(1)
(a)
(b)
(c)
Fig. 1: The helical phase fronts of OAM waves: (a) (cid:96) = 0, (b)
(cid:96) = 1, and (c) (cid:96) = 3.
where ε0 is the dielectric constant
in vacuum, r is the
radius vector of a point in the electric field, E and B are
the electric field intensity and magnetic flux density vector,
respectively, Re{·} represents the real part operation, and (·)∗
represents the conjugate operator. The angular momentum J
can be decomposed into a component S (SAM) associated
with polarization and a component L (OAM) related to the
spatial distribution of electromagnetic waves, i.e.
where
(cid:90)
(cid:90)
S = ε0
L = ε0
J = S + L,
Re{E∗ × A} dV,
(cid:110)
iE∗(cid:16)L · A
(cid:17)(cid:111)
Re
(2)
(3)
dV.
In (3), L = −i (r × (cid:79)) is the OAM operator, and A is the
vector potential.
It has been demonstrated that vortex waves having heli-
cal phase structure can carry OAM [1]. Due to the phase
singularity, the wavefront of electromagnetic waves will be
twisted during the propagation, forming vortex waves. The
vortex waves are characterized by helical phase fronts, as
shown in Fig. 1. The amplitude of the vortex wave field is
null along the axis center so that there is a "dark" core at the
location of the phase singularity. The helical phase structure is
described by the phase term exp(i(cid:96)θ), where θ is the transverse
azimuthal angle and (cid:96) is known as the topological charge or
OAM mode and can be any real number. The magnitude and
sign of topological charge (cid:96) determine the number and chirality
of the wave torsions in one wavelength, respectively. When an
OAM beam has integer desired topological charges the OAM
modes are called pure. In practice, it might happen that some
of the energy of one pure mode leaks into other modes. Mode
purity is measured as the ratio of the power of the desired
mode and the overall OAM power.
There are several special OAM waves that can be well
described by their radial intensity distributions, among which
Laguerre-Gaussian (LG) beams are probably the most widely
known. LG beams are paraxial solutions of the wave equation
in cylindrical coordinates and homogeneous media (e.g., free
space). In cylindrical coordinates,
the complex amplitude
distribution of a LG beam propagating along the z-axis is given
(cid:28595)(cid:28595)(cid:28595) (cid:19)
3
by [1], [5]
(cid:115)
LGp,(cid:96)(r, θ, z) =
(cid:35)(cid:96)
(cid:34)
(cid:20)
√
r
2
w(z)
ikr2z
πw2(z)(p + (cid:96))!
(cid:21)
2p!
(cid:20) −r2
(cid:20)
exp
exp
exp
2(z2 + z2
w2(z)
R)
−i(2p + (cid:96) + 1) tan−1
(cid:18) 2r2
(cid:19)(cid:21)
exp(i(cid:96)θ)
w2(z)
L(cid:96)
p
(cid:21)
(cid:18) z
zR
,
(4)
R)/z2
where w(z) = w0[(z2 + z2
R]1/2 is the radius of the beam,
w0 is the waist radius, zR is the Rayleigh range, k is the
(cid:96)
wave number, L
p (x) is the associated Laguerre polynomial
and (2p + (cid:96) + 1) tan−1 (z/zR) is the Gouy phase. It can be
seen from (4) that LG beams are characterized by two indexes:
the azimuthal index (cid:96) and the radial index p. The former is
the topological charge, which characterizes the beam OAM,
the latter is related to the number of radial nodes on the cross
section of the beam intensity. Therefore, the azimuth and radial
wavefronts of LG beams can be described by the (cid:96) and p
indexes.
The complex amplitude distribution of Bessel beams, an-
other special OAM wave described by their radial intensity
distributions, is
B(r, θ, z) = J(cid:96)(krr) exp(ikzz) exp(i(cid:96)θ),
(5)
where J(cid:96)(krr) is the (cid:96)-order Bessel function of the first kind,
kr and kz are radial and axial wave numbers, respectively.
OAM beams have many attractive properties but probably
the most important one is the orthogonality between modes.
Considering two OAM waves with topological charges (cid:96)1 and
(cid:96)2, it can be shown that the two OAM modes are orthogonal
through the inner product
ei(cid:96)1θ(cid:0)ei(cid:96)2θ(cid:1)∗
dθ =
(cid:40)
(cid:90) 2π
0
(cid:96)1 (cid:54)= (cid:96)2
0
2π (cid:96)1 = (cid:96)2
.
(6)
Thanks to the inherent orthogonality between OAM modes,
OAM waves with different (cid:96) can be used as a separate
set of data transmission channels adding a new dimension
independent of time, frequency, and polarization. Accordingly,
this new multiplexing dimension can be combined with other
existing multiplexing strategies to improve system capacity.
Moreover, OAM has certain application potential in particle
manipulation and imaging. Due to the presence of the "dark"
core, the pattern of OAM beams has a specific doughnut shape.
This particular shape allows OAM beams to capture particles
or drive the particles to rotate around the beam axis, with great
potential in biological and medical applications. Besides, this
ring beam with helical phase structure is expected to obtain
resolutions beyond the Rayleigh limit for imaging systems, and
provide new instruments for accurate radar targets imaging.
III. GENERATION OF OAM WAVES
Since Allen et al. in 1992 proved that the optical vortex
with a spiral wavefront carries OAM, the research on OAM
has gradually deepened and broadened. In this section, we
discuss various state-of-the-art methods for generating OAM
in the fields of light, radio, and sound waves. Because of the
Fig. 2: Taxonomy diagram of generation of OAM waves.
large number of methods discussed, Fig. 2 shows a diagram
that summarizes a taxonomy of the techniques presented in
this section.
A. Optical OAM
In optics, a vortex beam can either be obtained as the direct
output of a laser cavity, or can be generated by feeding a
Gaussian beam into a converter, such as a cylindrical lens,
a spiral phase plate, a phase hologram, a metamaterial or a
q-plate.
In [1], [11], cylindrical
Even before Allen showed that the vortex beam carries
OAM, research has been focused on vortex beams and on
how to generate them. In [9], the resonance phenomenon is
exploited to generate a vortex from an optical cavity. The
laser cavity normally produces a mixture of multiple modes
including the basic mode with (cid:96) = 0. By placing a component
inside the laser cavity, such as a spot-defect mirror [10], the
laser cavity can be forced to resonate on a specific OAM mode.
lenses are used to transform a
Hermite-Gaussian (HG) laser beam into a helically-phased LG
beam. An example of a 2nd order HG mode decomposition and
LG mode synthesis is shown in Fig. 3 (a). The 45◦ angle HG
mode can be decomposed into a series of HG modes, and this
series of HG modes can be rephased to obtain the LG mode.
This rephasing can be achieved by changing the Gouy phase
shift in the HG mode. Two cylindrical lens mode converters,
the π/2-converter and the π-converter, are proposed in [11],
as shown in Fig. 3 (b). The functions of the mode converters
are similar to the polarization of birefringence λ/4 and λ/2
plates. The π/2-converter can convert a HG mode with indices
m, n, oriented at 45◦ angle to the lens axis, into a LG mode
with topological charge (cid:96) = m − n and number of radial
nodes in the intensity distribution p = min(m, n). The π-
converter changes the index of the input mode, that is, HGm,n
turns into HGn,m or LGm,n turns into LGn,m, which has an
azimuthal dependence of the opposite sign. Cylindrical lens
Cylindrical lenses [1], [11]Spiral phase plates (SPPs) [12]Holographic gratings [13]-[15]Spatial light modulators (SLMs) [16], [17]Metamaterials [20]-[24]Liquid crystal q-plates [25], [26]Spiral phase plates (SPPs) [29]-[31], [35], [36]Spiral reflectors [33], [34]Holographic gratings [32]Uniform circular antenna arrays (UCAAs) [2], [8], [37]-[41]Metamaterials [44]-[49]Dielectric q-plates [50]Spiral phase plates (SPPs) [51], [52]Uniform circular transducer arrays (UCTAs) [4], [55]-[58]Spiral diffraction gratings [59]-[62]Metamaterials [64]-[66]Optical OAMRadio OAMAcoustic OAMGeneration of OAM waves(cid:28595) 4
(a)
(b)
Fig. 3: (a) An example of 2nd order HG mode decomposition
and LG mode synthesis: HG02 at 45◦ = 1
2HG02 + 1√
HG11 +
2HG20 and − 1
HG11 + 1
2HG20 = LG02. (b)
1
Schematic diagram of π/2-converter and π-converter. Both
converters consist of two identical cylindrical lenses of focal
length f [5].
2HG02 + i√
2
2
Fig. 4: Schematic diagram of SPP with step height s [5].
mode converters have high conversion efficiency and generate
OAM with high purity, but they require high construction
precision, and, at the same time, have poor flexibility because
they require a very precise incident field angle.
The spiral phase plate (SPP) is another way to implement
the vortex beam [12]. As shown in Fig. 4, the SPP is designed
like a rotating step so that its thickness increases as the azimuth
angle increases but is uniform radially. The step height is
expressed as s = (cid:96)λθ/[2π(n − n0)], where n is the refractive
index of the transparent dielectric material of the plate, n0 is
the refractive index of the external medium, (cid:96) is the topological
charge, λ is the wavelength of the incident light, and θ is the
spatial azimuth. When a plane wave of Gaussian light passes
through the phase plate, the light beam experiences a different
phase in the azimuth direction due to the spiral thickness of the
phase plate and is converted into a helically-phased beam with
topological charge (cid:96). The SPP has high conversion efficiency
and can be used for high power laser beams but has also the
limit that it can generate a single mode only and it has very
tight precision requirements.
Instead of producing a complex refractive optical element
Fig. 5: Phase holograms with (cid:96) = 3 in the form of (a) a
spiral phase hologram, (b) an (cid:96)-fold forked hologram, and (c)
a binarized (cid:96)-fold forked grating.
to generate a vortex wave, one can employ a computer-
generated hologram to design a diffractive optical element
whose transmittance function is related to the helical phase
exp(i(cid:96)θ). Holographic gratings, such as Fresnel spiral diffrac-
tion gratings and (cid:96)-fold forked diffraction gratings [13] -- [15],
are generated by using a photolithographic process to record an
interference pattern on a photorestist-coated substrate. When
Gaussian light is incident on the (cid:96)-fold forked diffraction
grating, a vortex beam with the topological charge of n(cid:96) can be
obtained at the nth diffraction order. Holographic diffraction
gratings are relatively simple and fast to produce and provide
good wavefront flatness and high efficiency for a single po-
larization plane. However, as the number of diffraction orders
increases, the quality of the vortex beam is seriously degraded
and this method is generally only used to generate low-order
vortex beams.
Unlike the SPP and the diffraction grating described above,
the spatial light modulator (SLM) can generate vortex beams
with different topological charges [16], [17]. A SLM is a
pixelated liquid crystal device whose liquid crystal molecules
can be programmed to dynamically change the incident beam
parameters, including the beam phase in the transverse plane,
to create a vortex beam. In detail, a phase hologram with a
transmittance function of exp(i(cid:96)θ) is digitally generated and
loaded on the SLM so that the SLM will determine the phase
of each point in the two-dimensional space according to the
value of each pixel of the input phase hologram. In general,
to generate a vortex beam the phase hologram loaded onto the
SLM is in the form of: a) a spiral phase hologram, b) a (cid:96)-
fold forked hologram, and c) a binarized (cid:96)-fold forked grating
(shown in Figs. 5(a), 5(b), and 5(c), respectively). The spiral
phase hologram produces a vortex beam, which exits in the
direction perpendicular to the hologram plane, with only a
single topological charge. The forked hologram can control
the exit direction of the vortex beam, while the binarized
forked grating has multiple diffraction orders, which produce
different topological charges at different diffraction orders.
Liquid crystal SLMs are very flexible since they can control
various parameters of the vortex beam but they are relatively
expensive and have an energy threshold that makes their use
impossible with high power laser beams. Recently, in addition
(cid:28595)(cid:28595)(cid:28595)(cid:894)(cid:258)(cid:895)(cid:3)(cid:894)(cid:271)(cid:895)(cid:3)(cid:894)(cid:272)(cid:895)(cid:3)(cid:3) 5
Fig. 6: (a) A metamaterial composed by a sub-wavelength V-
shaped antenna array. The entire structure consists of eight
regions and the phase response difference between adjacent
regions is π/4. (b) Partially enlarged view of the center
part of (a). (c) Measured pattern with a spiral fringe created
by the interference of the vortex beam with (cid:96) = 1 and a
co-propagating Gaussian beam. (d) Measured pattern with a
dislocated fringe created by the interference of the vortex beam
with (cid:96) = 1 and a Gaussian beam when the two are tilted with
respect to each other [20].
to the liquid crystal SLM, a diffractive optical element using a
digital micro-mirror device (DMD) has been introduced [18],
[19]. In contrast to liquid crystal devices, DMDs cost less and
are faster but their diffraction efficiency is lower.
Transformation optics is a recently developed discipline
that employs complex artificial materials, called metamate-
rials, to make transformations in optical space. Generation
of optical OAM is one of the fields where the findings of
transformation optics can be applied: the metamaterials are
planar ultra-thin optical components, usually composed of sub-
wavelength constitutional units, such as V-shaped antennas
[20], [21], L-shaped antennas [22], rectangular apertures [23],
and rectangular split-ring resonators [24]. Rather than relying
on the continuous transitions of conventional optics, these
planar ultra-thin optical components operate by forcing a
sudden change of beam phase, amplitude or polarization at
the interface. Thus, optical OAM is obtained by controlling
the geometrical parameters (shape, size, direction, etc.) of the
metamaterial to manipulate the phases of different azimuths
and change the spatial phase of the incident light. As shown
in Fig. 6, the metamaterial structure is composed by a sub-
wavelength V-shaped antenna array, and the beam phase can
be controlled by adjusting the angle between the two arms
of the V-shaped antenna. The entire surface is divided into
eight regions, and the phase response difference between the
adjacent regions is π/4. This object generates a helical phase
shift relative to the front of the incident beam producing a
vortex with the topological charge of (cid:96) = 1. The biggest
advantage of the use of metamaterials is their small size that
allows for easy integration. Unfortunately, only a vortex beam
with a particular topological charge can be generated by a
Fig. 7: Four examples of q-plate patterns with: (a) (q, α0) =
(1/2, 0), (b) (q, α0) = (1, 0), (c) (q, α0) = (1, π/2) and
(d) (q, α0) = (2, 0). The segments indicate the optical axis
orientation in the transverse plane. (e) Pictorial illustration
of the optical action of a tuned q-plate with q = 1 on an
input circularly polarized plane beam. A left-circular (or right-
circular) polarized Gaussian beam passing through a tuned
q-plate with q = 1 turns into a helically phased beam with
(cid:96) = +2 (or (cid:96) = −2) and right-circular (or left-circular)
polarization [25].
single device.
The q-plate [25], [26] is a liquid crystal panel that has a
uniform birefringence phase retardation δ and a transverse
optical axis pattern with a non-zero topological charge. The
angle between the optical axis orientation and the x-axis in
xy plane can be expressed as α(r, ϕ) = qϕ + α0, where
α0 is an initial optical axis orientation. Q-plate patterns with
different values of q and α0 are shown in Fig. 7. When q-plate
is optimally tuned, i.e. δ = π, a light beam incident on it is
modified to have a topological charge variation ∆(cid:96) = ±2q.
For example a circularly polarized Gaussian passing through
a tuned q-plate with q = 1 has a helical phase front with
topological charges (cid:96) = ±2, where the sign depends on the
chirality of the input polarized light, as shown in Fig. 7(e).
The characteristic of q-plates is that they generate OAM as a
result of optical spin-to-orbital angular momentum conversion,
exploiting the law of conservation of angular momentum, and
for q = 1,
the maximum topological charge that can be
generated is (cid:96) = 2. The q-plate is essentially a Pancharatnam-
Berry phase optical component that controls the wavefront
shape by transitioning the polarization state. In addition to
the q-plate, the computer-generated sub-wavelength dielectric
grating [27], [28] and the L-shaped antenna array metamaterial
proposed in [22] are also Pancharatnam-Berry optical devices.
The q-plate is capable of producing a pure OAM mode but,
being made of liquid crystal, has also an energy threshold
and thus cannot be used in conjunction with high power laser
beams.
Table I compares the different methods of generating op-
tical OAM with respect to several parameters: cost, speed,
(cid:3)(cid:894)(cid:258)(cid:895)(cid:3)(cid:894)(cid:272)(cid:895)(cid:3)(cid:894)(cid:271)(cid:895)(cid:894)(cid:282)(cid:895)(cid:3)(cid:3)(cid:894)(cid:286)(cid:895)(cid:3)(cid:3)(cid:3) 6
TABLE I: Comparison of Optical OAM Generation Methods.
Features
Cylindrical lenses
SPPs
Metamaterials Liquid crystal
Holographic
gratings
Low
Fast
Low
SLMs
High
Normal
Normal
Cost
Speed
Conversion efficiency High
Normal
Normal
OAM mode
Flexibility
Working frequency
Processing difficulty
System complexity
Market readyness
Multiple modes;
Pure mode
Low
All
High
High
Yes
Low
Normal
High
Single mode;
Non-pure mode Multiple mode Multiple modes;
Composite mode
High
Low
One point
All
Low
High
Low
Low
Yes
Yes
Low
All
High
Low
Yes
Low
Normal
Relatively high
Single mode
Low
A range
High
Low
No
q-plates
High
Fast
Relatively high
Single mode;
Pure mode
High
All
Low
Low
Yes
conversion efficiency, the ratio of outgoing power to incident
power, OAM mode, the capacity of generating pure integer
modes, multiple modes, i.e. several different pure modes, or
composite modes OAM, flexibility, the capacity of controlling
the OAM wavefront parameters such as the beam direction or
the size of OAM beam ring, working frequency, processing
difficulty, the difficulty in manufacturing the generator, system
complexity, the level of difficulty of integrating the generator
in a transmission system, market readyness, the availabilty on
the market of the technology as a product. As can be seen,
the lesson learned is that each method has advantages and
disadvantages. Taking the cylindrical lens as an example, it has
high mode conversion efficiency and generates OAM modes
with high purity, but processing difficulty and system complex-
ity are high. The OAM generation method should be chosen
according to practical considerations. Flexibility is the biggest
advantage of SLMs. If different OAM modes or superimposed
modes are required, SLMs should be the first choice. Besides,
SPPs perform well in cost, conversion efficiency and structural
complexity. If mode purity and processing difficulty are not
considered, SPPs are more recommended in a single OAM
mode application due to the low cost and the simple structure.
In optical OAM communication systems, SLMs loaded with
phase holograms are widely used to generate OAM, and SPPs
are second only to SLMs. In general, cylindrical lenses require
a special incident field that limits their application. At the
same time, low conversion efficiency makes the application of
holographic gratings much less practical than SLMs and SPPs.
Metamaterials are low cost, small size and high efficiency,
so they are more suitable for small integrated vortex beam
generators. The q-plate is a spin-orbital angular momentum
conversion device and as such is not recommended yet for
OAM communications because of the low value of the max-
imum topological charge and is more suitable for particle
manipulation.
B. Radio OAM
OAM is a fundamental characteristic of electromagnetic
waves at all frequencies. It is not limited to the optical band
and can be generated also in the radio band [2].
Fig. 8: (a) Single-stage and multi-stage SPPs [29], planar SPPs
with varying borehole densities (b) [30] and radius (c) [31],
(d) forked gratings [32], (e) stepped spiral reflecting surface
[33] and (f) spiral parabolic antenna [34].
In analogy with the generation of optical OAM, SPPs,
whose thickness increases with the increase of the azimuth
angle, have been the first method used to generate OAM [35],
[36] at high radio frequencies. To generate a high-order vortex
beam,
the spiral surface is usually replaced by a stepped
surface to form a stepped SPP [29] as shown in Fig. 8 (a).
Recently, also planar SPPs [30], [31] have been proposed for
generating radio OAM. Unlike conventional SPPs, which use
thickness to change the beam phase, planar phase plates adjust
the wavefront of the electromagnetic wave by controlling the
change in dielectric constant using different borehole densities
[30] or radius [31]. Although these devices have a planar
structure, like ordinary phase plates, the requirements in terms
of structural precision is very high and the plate is difficult to
manufacture.
To create OAM, the stepped spiral surface can be also
built as a reflector illuminated by a conventional antenna [33].
The spiral reflector causes the reflected electromagnetic wave
to have a wave path difference at different positions of the
cross section, so that the desired electromagnetic vortex is
created. As in Fig. 8 (e), the reflecting stepped spiral surface
is divided into N discrete regions, each of which introduces a
(cid:3)(cid:894)(cid:258)(cid:895)(cid:894)(cid:282)(cid:895)(cid:895)(cid:3)(cid:895)(cid:3)(cid:894)(cid:271)(cid:895)(cid:3)(cid:894)(cid:286)(cid:895)(cid:3)(cid:894)(cid:272)(cid:895)(cid:3)(cid:894)(cid:296)(cid:895)(cid:3)(cid:3) 7
Single mode
Low
Fast
Relatively high
Metamaterials Dielectric
q-plates
Low
Normal
/ (Not discussed)
Single mode;
Pure mode
Low
One point
High
Low
No
Low
One point
High
Low
No
TABLE II: Comparison of Radio OAM Generation Methods.
Features
SPPs
Cost
Speed
Conversion efficiency High
Low
Normal
Holographic
gratings
Low
Fast
Low
OAM mode
Flexibility
Working frequency
Processing difficulty
System complexity
Market readyness
Single mode;
Non-pure mode Multiple mode
Low
One point
High
Low
No
Low
All
High
Low
No
Spiral reflectors UCAAs
Low
Normal
Normal
Single mode;
Non-pure mode
Low
One point
High
Low
No
High
Normal
Normal
Multiple modes;
Composite mode
High
All
Low
High
Yes
Fig. 9: Schematic diagram of a UCAA with N elements
uniformly distributed on the circumference, where the azimuth
difference between adjacent elements is δφ = 2π/N [8].
discontinuous phase difference 2π/N. The topological charge
of radio OAM of stepped surfaces is (cid:96) = 2µ0(N + 1)/(λN ),
where N is the number of regions into which the reflection
surface is partitioned, µ0 is the total surface spacing, and λ is
the wavelength of the incident electromagnetic wave. When
N tends to infinity, the stepped spiral reflector becomes a
continuous form, such as the spiral parabolic antenna used in
the 2012 OAM wireless communication experiment [34]. The
spiral reflector is simple and easy to implement. However, it
can only respond to specific frequencies and the OAM mode
produced by this method is difficult to determine.
Holographic gratings designed by means of computer-
generated holograms are also a method of generating radio
OAM in analogy with the generation method at optical fre-
quencies [32].
Nevertheless, a uniform circular antenna array (UCAA),
as shown in Fig. 9,
is the most commonly used antenna
for generating electromagnetic vortex [2], [8]. The UCAA
is composed by N elements, uniformly distributed on the
circumference. Each array element is controlled by an input
signal of the same amplitude but different phase and the phase
difference between adjacent array elements is ∆ϕ = 2π(cid:96)/N,
where (cid:96) is the OAM mode, so that
the phase difference
over the whole array is 2π(cid:96). Theoretically, a UCAA with N
array elements can produce a distortion-free vortex wave with
(cid:96) < N/2. By precisely controlling the amplitude and the
Fig. 10: Pictorial
reflective metamaterial [44].
illustration of generating OAM with a
phase of the excitation signals, UCAAs have the potential
to generate multiplexed radio OAM. In practice, UCAAs
can be used in different forms of antennas, such as dipole
antennas [2], [8], Vivaldi antennas [37], horn antennas [38],
Yagi antennas [39], and microstrip patch antennas [40], [41].
In addition to UCAAs, radio OAM generation with imperfect
uniform circular arrays (IUCAs) has also be discussed in [42].
In [43] a circular time switched array (TSA) is employed
to generate vortex electromagnetic waves. This TSA method
uses high-speed RF switches to activate the array elements
sequentially, enabling the simultaneous generation of all OAM
modes at the harmonic frequencies of the TSA switching cycle.
The mode of the electromagnetic vortex generated by the
UCAA can be controlled flexibly, but this usually requires
a complicated feed network.
Similar to the generation method at optical frequencies,
metamaterials, which can be classified as reflective metama-
terials [44] -- [48] and transmissive metamaterials [49], are also
used to generate vortex electromagnetic waves. The schematic
diagram of generating OAM using a reflective metamaterial is
shown in Fig. 10, where the incident wave at radio frequencies
is reflected and converted into a vortex wave.
In addition to metamaterials, also q-plates, made by carving
(cid:28595) 8
a series of concentric rings on a dielectric plate, can be used
to generate a vortex wave at radio frequencies [50].
Table II compares the most common methods of generating
vortex waves at radio frequencies. Normally, because their
size depends on the signal wavelength, SPPs and holographic
gratings are suitable for the generations of vortex waves
at high frequencies such as millimeter-waves, while spiral
reflectors and UCAAs are more suitable for OAM gener-
ation at lower frequencies. It is well known that massive
multiple-input multiple-output (MIMO) technology is going
to be one of the key technologies for 5G mobile networks.
Therefore, UCAAs have attracted a wide attention and it is
currently under study the introduction of OAM into 5G or
beyond 5G networks. In addition, thanks to their capability of
controlling the wavefront, radar imaging systems frequently
uses UCAAs to generate OAM waves. The low cost, high
conversion efficiency and simple structure of SPPs make them
recommended for radio OAM communication. Since the spiral
parabolic antenna can focus the OAM wave while it is being
generated, it is more suitable for long distance transmissions.
Moreover, SPPs, spiral parabolic antennas and other single
transmit antennas are expected to be combined with MIMO
technology to further increase capacity. Consistently with what
happens with optical OAM, low conversion efficiency makes
holographic gratings less useful, and metamaterials are more
suitable for miniaturized integrated circuits due to the small
size and low cost. Also in this scenario, q-plates are not
particularly recommended for radio OAM communications.
C. Acoustic OAM
OAM has found applications in audio bands as well [4]. Be-
cause of their simple structure and high conversion efficiency,
SPPs are widely used to generate optical vortices and radio
vortices. Similarly, SPPs have also been employed to generate
acoustic vortices. The acoustic SPP in [51] has thickness h that
0 −c−1)], where c0 is the
varies with azimuth, h = (cid:96)θ/[2πf (c−1
sound speed in the surrounding medium, c is the sound speed
in the phase plate, and f is the frequency of the incident wave.
Another absorbing SPP using optoacoustic technology was
proposed in 2004 to generate OAM in the ultrasonic band [52].
The difference is that the optoacoustic conversion efficiency
of this SPP is very low, and this device requires high-energy
short pulse excitation. In addition to passive SPPs, active sound
sources with spiral thickness [53], [54] are used to generate
acoustic vortices. The dimensions of those spiral structures are
generally dozens of wavelengths and their large volume limits
the application in low frequency sound waves.
In analogy with antenna arrays in radio OAM, transducer
arrays especially uniform circular transducer arrays (UCTAs)
are also widely used to generate acoustic vortices carrying
OAM. A simple four-panel transducer has first been proposed
to produce acoustic vortices [4]. Subsequently, hexagonal
arrays [55] and UCTAs [56] -- [58] have also been studied:
UCTAs use fewer transducers than hexagonal arrays, and, by
adjusting the phase difference between adjacent transducers,
are capable to flexibly generate acoustic vortices with different
OAM modes. Fig. 11 shows a UCTA composed by N uni-
formly spaced elements: by driving each transducer with a sine
Fig. 11: Schematic of an acoustic vortex generation system
using a UCTA composed by N uniformly distributed elements.
The positions of the n-th sound source Sn and the observation
point Q are (R, ϕ0n, 0) and (r, ϕ, z), respectively [58].
Fig. 12: (a) The multi-arm coiling slits using logarithmic
spiral gratings for generating the stable acoustic vortices. (b)
Schematic diagram of the one-armed, two-armed, three-armed,
and four-armed coiling slits [59].
wave with frequency f and phase difference ∆ϕ = 2π(cid:96)/N, the
array can generate an acoustic vortex with (cid:96) ≤ (cid:98)(N − 1)/2(cid:99),
where (cid:98)x(cid:99) indicates the greatest integer less than or equal
to x. By precisely controlling the amplitude and the phase
of the excitation signals, UCTAs can generate multiplexed
acoustic vortices. It should be noted that each transducer can
be regarded as a point source in low frequency acoustic fields,
but not at high frequencies. In high frequency acoustic fields,
the directivity of the sound source needs to be considered
because the size of the transducer is larger than the wavelength
of the sound wave. Like UCAAs, UCTAs also require complex
control circuits, and the complexity grows with the number of
transducers.
Another method for generating acoustic vortices is to
employ sub-wavelength spiral diffraction gratings depicted
on rigid materials, which, due to the diffraction effect, can
generate acoustic vortices in the paraxial region. Higher mode
acoustic OAM can be produced by increasing the number
of arms of spiral gratings. Using a grating with m arms,
an acoustic vortex with the topological charge of (cid:96) = mn
is obtained at the nth diffraction order. Spiral gratings can
be designed as Archimedes spiral type [60], [61], Fresnel
(cid:3)(cid:894)(cid:258)(cid:258)(cid:895)(cid:3)(cid:894)(cid:271)(cid:895)(cid:3)(cid:3) TABLE III: Comparison of Acoustic OAM Generation Methods.
9
Features
SPPs
UCTAs
Cost
Speed
Conversion efficiency High
Low
Normal
OAM mode
Flexibility
Working frequency
Processing difficulty
System complexity
Market readyness
Single mode;
Non-pure mode
Low
One point
High
Low
No
Spiral diffraction
gratings
Low
Normal
Low
High
Normal
Normal
Multiple modes;
Composite mode Multiple modes
High
All
Low
High
Yes
Low
A range
High
Low
No
Metamaterials
Low
Normal
Relatively high
Single mode
Low
One point
High
Low
No
acoustic wave can be adjusted to be spirally distributed along
the azimuthal direction. A planar metamaterial structure for
converting planar acoustic waves into vortex acoustic waves
using acoustic resonance is shown in Fig. 13. The metamaterial
is divided into eight fanlike sections. Each individual section
is configured to be composed of three rows of resonators in the
radial direction. Each row consists of four Helmholtz cavities
and a straight pipe. The phase of the outgoing sound wave
can be expressed as φout = φin + k(ef f )l, where k(ef f ) is the
equivalent wave number and varies with the ratio of the heights
of the pipe and resonator h1/h and l is the structural thickness.
It can be shown that when k(ef f ) depends on azimuth change,
the output phase also changes with azimuth. The cascade
of four Helmholtz cavities in each row can flexibly control
k(ef f ) from −k to k over the whole azimuth, ie, achieving
an exit phase from 0 to 2π, while the combination of the
Helmholtz cavity and the straight pipe can get a very high
transmittance. The phase difference between adjacent sections
of the metamaterial is π/4, so that planar acoustic waves are
converted into vortex sound waves with the topological charge
of (cid:96) = 1. This metamaterial structure has the advantages of
being highly efficient with a small size and a planar shape.
However, the proposed structure can only be used for the
generation of acoustic vortices at specific frequencies in the
air.
Table III compares the generation methods of acoustic
OAM. Due to the dimension of the structure, SPPs are
more suitable for high frequency sound fields and UCTAs
are more suitable for low frequency sound fields. Because
of their flexibility, UCTAs are more common in acoustic
communications and particle manipulation. Both spiral diffrac-
tion gratings and metamaterials are characterized by sub-
wavelength dimensions. However, passive diffraction gratings
have low efficiency because a large amount of incident sound
field energy is blocked and lost, and active diffraction gratings
are more expensive. Therefore, metamaterials with small size,
low cost, and high efficiency have attracted a lot of attention
recently.
Fig. 13: (a) Schematic of the assembled planar acoustic reso-
nance layer consisting of eight fanlike sections. The thickness
of the layer is 0.5λ. (b) An individual section consisting of
three rows of resonators with fixed height h in the radial r
direction, sided by pipes of varying height h1 to produce the
needed effective wave number [64].
spiral type [62], and logarithmic spiral type [59] shown in
Fig. 12. The first two types can only be used for a specific
frequency, while the logarithmic spiral type is for a wider
frequency range. In polar coordinates, when m = 1 the two
concentric logarithmic spirals are denoted as r1 = a1 exp(bθ)
and r2 = a2 exp(bθ), respectively, where θ is the angular
coordinate, r1 and r2 are the radial coordinates, a1 and a2 are
the inner radii of the two spirals, b varies with the number of
arms, determining the growth rate of the spiral and the width of
the spiral grating is d = r2 − r1. As long as the wavelength of
the incident acoustic wave is greater than 2dmin and less than
2dmax, effective diffraction can occur to obtain the desired
acoustic OAM. Since the rigid material blocks a large amount
of incident power, such a passive spiral diffraction grating has
a low efficiency. Recently, an active spiral grating fabricated
using a cellular ferroelectret film has been proposed to improve
conversion efficiency [63].
In addition to optical OAM and radio OAM, metamaterial
can also be used to generate acoustic OAM [64] -- [66]. Usually,
to generate acoustic OAM a metasurface with sub-wavelength
thickness is equally divided into N regions. By changing
structural parameters in each region, the phase of the outgoing
10
TABLE IV: Summary of OAM Generation Methods.
Methods
Cylindrical lenses
SPPs
Introduction and Features
HG modes without OAM can be converted to LG modes with OAM.
High conversion efficiency; high OAM mode purity; a special incident field
is required.
With spiral thickness h = (cid:96)λθ/[2π(n − n0)] in optics and electromagnetics
and h = (cid:96)θ/[2πf (c
Simple structure; high precision; high conversion efficiency; specific working
frequency.
0 − c−1)] in acoustics.
−1
Remarks
Optical OAM
Optical OAM
Radio OAM
Acoustic OAM
Holographic gratings
Interference patterns of plane waves and vortex waves, such as (cid:96)-fold forked
diffraction gratings, whose transmittance function is related to exp(i(cid:96)θ).
Simple and fast; low efficiency.
Optical OAM
Radio OAM
SLMs
UCAAs
UCTAs
A pixelated liquid crystal device loaded with a phase hologram whose
transmittance function related to exp(i(cid:96)θ).
High cost; flexible control.
Optical OAM
N elements uniformly distributed on the circumference with the excitation
phase difference 2π(cid:96)/N can generate radio OAM with (cid:96) < N/2.
Flexible control; a complicated feed network is required.
Similar to UCAAs.
Radio OAM
Acoustic OAM
Radio OAM
Acoustic OAM
Spiral reflectors
Spiral diffraction grat-
ings
The reflection surfaces are divided into N discrete regions and can generate
OAM with (cid:96) = 2µ0(N + 1)/(λN ).
Simple structure; beam convergence effect; it is difficult to determine topo-
logical charge values; specific working frequency.
Archimedes spiral type, Fresnel spiral type or logarithmic spiral type gratings
depicted on sub-wavelength rigid materials.
Low efficiency; the topological charge is changed by controlling the number
of arms of the grating.
Metamaterials
Q-plates
It is composed of sub-wavelength constitutional units and can flexibly manip-
ulate wavefront phase, amplitude and polarization.
Low cost; small size; easy integration; high conversion efficiency; specific
working frequency.
Optical OAM
Radio OAM
Acoustic OAM
A uniform birefringent liquid crystal or dielectric plate that can generate OAM
as a result of optical spin-to-orbital angular momentum conversion.
High conversion efficiency; fast response; high OAM mode purity; simple
chiral control of OAM beams.
Optical OAM
Radio OAM
D. Summary and Open Challenges
Table IV summarizes the most important methods for gener-
ating OAM. Optical OAM generation is a field that has been
developed relatively earlier with respect to radio OAM and
acoustic OAM generation. Moreover many methods, such as
SPPs and metamaterials, are proposed by analogy with optical
OAM for radio and acoustic bands. Nevertheless, based on
their respective characteristics, specific methods have been
proposed to generate radio OAM and acoustic OAM, such
as UCAAs and UCTAs.
The methods of generating optical, radio, and acoustic OAM
have been compared in their respective subsections. In general,
SLMs are relatively optimal in optical OAM generation. SPPs
and UCAAs have also proven to be good at generating radio
OAM in practice. For acoustic OAM generation, UCTAs are
flexible in both air and water. However, the inherent size of the
transducer makes UCTAs not suitable for many miniaturized
applications. The metamaterials with small size proposed so
far can only generate OAM in the air. Nevertheless, underwater
acoustic communication is a research area of great interest and
so, finding a new material or structure with low cost, small size
and high efficiency to generate a stable acoustic vortex field
underwater is an important research direction. In addition, the
capacity of focusing the beam for long distance transmissions
is essential. How to generate convergent OAM beams is a
challenging topic in optical, radio and acoustic research.
IV. DETECTION OF OAM WAVES
In this section, we discuss the various methods for detecting
OAM waves. At the receiver of a communication system, the
different OAM modes can be separated easily by exploiting
the orthogonality of the helical phase fronts. A variety of
methods for detecting OAM have been proposed for light and
radio waves. In the acoustic frequency range, OAM detection
11
Fig. 14: Taxonomy diagram of detection of OAM waves.
Fig. 15: A Dove prism interferometer using a Mach-Zehnder
interferometer with a Dove prism placed in each arm, where
α/2 is the relative angle between the Dove prisms. If α/2 =
π/2, it sorts photons with even-(cid:96) into port A1 and photons
with odd-(cid:96) into port B1 [67].
is still in its early days. A taxonomy diagram of the detection
methods discussed here is shown in Fig. 14.
A. Optical OAM
Being the detection of optical OAM the inverse process of
generating OAM, OAM modes can be detected by using an
inverse SPP designed for the specific mode to be detected
[7]. Moreover, a vortex beam with the topological charge of (cid:96)
is converted into planar Gaussian light after being irradiated
through a holographic grating with a transmittance function
related to the anti-helical phase factor exp(−i(cid:96)θ) [68]. Using
spatial mode filters, Gaussian light can be separated from other
OAM beams, detecting only one specific OAM mode at the
time. For multiplexed OAM modes, a series of holographic
gratings are required to sequentially detect each single OAM
mode, or, alternatively, beam splitter can be combined for
parallel detection. In [69] and [70] a two-dimensional forked
holographic grating, which is the superposition of a vertical
grating and a horizontal grating, is designed to simultaneously
detect 8 OAM modes. Two-dimensional Dammann gratings,
combining the characteristics of conventional gratings and
Dammann gratings, can achieve equal energy distribution on
the designed diffraction order, expanding the detection range
Fig. 16: The first three stages of a general sorting scheme,
where the gray boxes represent Dove prism interferometers.
The first stage introduces a phase shift of α = π to sort even
(cid:96)s into port A1 and odd (cid:96)s into port B1. The second stage
and the third stage introduce phase shifts of α = π/2 and
α = π/4 to sort (cid:96) = 4n/(cid:96) = 4n + 2 and (cid:96) = 8n/(cid:96) = 8n + 4,
respectively. It should be noted that (cid:52)(cid:96) = 1 and (cid:52)(cid:96) = 2
holograms are required before sorting in the second and third
stages [67].
of the grating [71], [72]. At present, Dammann holographic
gratings are widely used at the receiving end of optical com-
munication systems since they can perform parallel detection
on multiple OAM states. These holographic gratings require
high precision and are normally loaded onto SLMs or recorded
using advanced grating fabrication techniques. At the same
time, their conversion efficiency is not higher than 1/N, where
N is the total number of detected modes. Recently, [73] has
proposed a novel phase hologram, designed by modifying the
Lin's algorithm [74], which can be used to achieve almost
complete concentration of incident energy to a specified target
OAM mode that can be detected more effectively than with
conventional forked gratings.
A vortex wave with the topological charge of (cid:96) can inter-
fere with a plane wave to obtain (cid:96) spiral stripes or (cid:96)-fold
forked stripes. As such, there are many methods for detecting
optical OAM using interference, like the Young's double-slit
interference [75], [76], the multipoint interference [77], and
the annular aperture interference [78]. However, with these
methods it is difficult to correctly detect the intended OAM
mode because of the complexity of the received interference
pattern. A simpler method is to use a Mach-Zehnder inter-
ferometer with a Dove prism placed in each arm to detect
single photon OAM [67], called Dove prism interferometer,
as shown in Fig. 15. The angle of rotation between the two
Dove prisms is α/2, and the rotation angle of the passing
beam is α. Correspondingly, the phase difference between the
two arms is (cid:96)α. If α/2 = π/2, by correctly adjusting the
path length of the interferometer, all beams with even (cid:96) have
constructive interference at one output port and cancellation at
the other output port, while for all beams with odd (cid:96) it is just
the opposite. Therefore, such an interferometer can achieve
Holographic gratings [68]-[72]Dove prism interferometers[67], [79]Mode sorters [80]-[82]Interference phase detection [34]Single-point estimation method [89]Phase gradient method [89], [90]Uniform circular antenna arrays (UCAAs) [92]Partial aperture sampling receiving (PASR) [91]Metamaterials [95]Optical OAMRadio OAMAcoustic OAMDetection of OAM wavesInner product operation [94]Inverse SPPs [7]Inverse SPPs [88](cid:28595)Α1Β1Inputα/2∆l=2 ∆l=12nd stageα=π/23rd stageα=π/41st stageα=πA1B1A2B2C2D2A3B3C3D3E3F3G3H3 12
TABLE V: Comparison of Optical OAM Detection Methods.
Inverse SPPs Holographic gratings
Features
Cost
Low
Conversion efficiency High
Low
Low, no higher than 1/N
Dove prism interferometers
Low
High, near 100%
OAM mode
Single mode
detection
Composite mode detection;
The number is limited by
the size of the grating
Composite mode sort;
The more the number, the more
complicated the device
Flexibility
System complexity
Market readyness
Low
Low
Yes
High
Low
Yes
Low
High
Yes
Mode sorters
Low
Relatively high
Composite mode sort;
Adjacent OAM mode
cannot be effectively
separated
High
Low
Yes (If implemented
on SLMs)
Thus, a ring beam with a helical phase front on the input plane
is mapped to a rectangular beam with a tilted wavefront on
the output plane. The rectangular beams with different lateral
phase gradients are focused by the lens on different lateral
positions on the focal plane, effectively separating the different
OAM modes. In [81] an OAM mode sorter composed of
two custom refractive optical elements is designed to achieve
higher conversion efficiency. Nevertheless, there is a limit to
the mode sorter due to the overlap between the focal points of
two adjacent OAM modes. In [82] it is shown that by using
a series of unitary optical transformations, which generate
multiple copies of the rectangular beams, it is possible to
achieve a larger separation between two OAM modes, trading
a greater efficacy with higher system complexity.
In addition to the above methods, there are other techniques
for separating OAM modes. In [83] it is proposed a method
based on the use of q-plates. The radius ratio method can be
used to detect certain special optical OAM [84], [85]. Photonic
integrated circuits are well compatible with single-mode fibers
while achieving optical OAM multiplexing and demultiplexing
[86], [87].
Four common methods for detecting optical OAM are
compared in Table V. As mentioned above, also in this
case each method has its own advantages and disadvantages.
From the perspective of conversion efficiency, the Dove prism
interferometer is the best method, but it is bad in terms of
flexibility and system complexity, so that it is a not suitable
technology for detecting a large number of OAM modes.
Inverse SPPs and mode sorters perform better in both respects.
A drawback of inverse SPPs is that they can only detect a
specific OAM mode, and it is necessary to combine beam
splitters and multiple inverse SPPs to achieve composite mode
detection. Mode sorters enable composite mode sort, but they
can not effectively separate two adjacent OAM modes. The
holographic grating has a low conversion efficiency, but it
can be easily loaded onto a SLM. In current optical OAM
experimental systems, the holographic grating loaded on a
SLM is a relatively common method. In addition, inverse phase
holograms are also frequently used in practice. Some novel
phase holograms have also been tried to detect optical OAM
with high efficiency.
Fig. 17: Principle of the OAM mode sorter. By using geomet-
ric transformations, ring beams with helical wavefronts are
mapped to rectangular beams with tilted wavefronts [7].
OAM mode odd-(cid:96) and even-(cid:96) separation at the single photon
level. As shown in [79], the improved interferometer system
is capable of detecting a single photon OAM and SAM. By
cascading multiple interferometers as shown in Fig. 16, it is
possible to generalize this architecture to detect any number
of OAM modes, and the theoretical efficiency is close to
100%. In principle, detecting N OAM modes requires N − 1
interferometers. Therefore, such an interferometer device is
too complex for practical applications with a large number of
OAM modes.
After passing through a lens, a plane wave converges into
a point, whose lateral position on the focal plane depends on
the lateral phase gradient of the plane wave. Note that the
two plane waves need at least a phase difference of 2π to
make the distance between the two focal points larger than
the Rayleigh resolution limit. Using this idea, [80] proposes a
mode conversion method for sorting OAM modes. The mode
sorter consists of a converter and a corrector, both imple-
mented on the SLM. The converter implements the conversion
from a Cartesian coordinate system to a log-polar coordinate
system, so that a point (x, y) on the input plane is converted
into a point (u, v) on the output plane, where the conversion
law is u = −a ln((cid:112)x2 + y2/b) and v = a arctan(y/x),
where a and b are two independently chosen parameters. The
corrector is used to compensate the phase distortion caused by
the optical path length during the coordinate transformation.
13
Fig. 19: Illustration of the phase gradient method. The OAM
mode can be estimated by computing the phase difference of
the electric field components at two sampling points separated
by an angle β, i.e. (cid:96) = (φ1 − φ2)/β [89].
Fig. 18: Sketch of the famous Venice experiment of radio
OAM communications, in which two receive antennas A and
B are aligned with the phase singularity [34].
B. Radio OAM
As in optics, the detection of OAM at radio frequencies can
be achieved by employing carefully designed inverse SPPs
[88]. It is also possible to detect the radio OAM mode by
observing the interference pattern of a vortex wave and a plane
wave. Nevertheless, many detection methods in optics are no
longer applicable at radio frequencies and some new methods
have been proposed.
In [34] the interference phase detection is used to separate
two electromagnetic beams with topological charges of (cid:96) = 0
and (cid:96) = 1. In the experiment set-up, two Yagi antennas A and
B are connected by a 180◦ phase-shift cable to form a phase
interferometer, designed to reduce background interference in
a real environment. As shown in Fig. 18, if the device is fully
aligned, the vortex electromagnetic beam with (cid:96) = 1 will
produce a phase difference of 180◦ between receive antennas
A and B. This phase difference is compensated by the cable
line phase delay, so that the received interference intensity
is maximized. For the beam with (cid:96) = 0, there is only the
phase delay of the cable, and the intensity is minimal. In this
way, it is possible to separate beams with different wavefront
characteristics. When the interference phase detection method
is used for more radio OAM modes, the receiving device
becomes very complicated.
In 2010, Mohammadi et al. proposed two methods to
estimate far-field OAM modes in radio beams, the single-
point estimation method and the phase gradient method [89].
The single-point estimation method uses an approximation for
OAM in the far-field to estimate the OAM from the mea-
surements at a single point of the vertical and the transverse
components of the electric field. This method is efficient for
the detection of OAM with low mode numbers. Compared
with the single-point estimation method, the phase gradient
method is more intuitive and effective. As shown in Fig. 19,
the measurement set-up consists of two test points separated
Fig. 20: The PASR scheme with M antennas uniformly
distributed on a 1/P arc [91].
by an angle β on a circumference at the receiving end. If the
measured phases are φ1 and φ2, respectively, the OAM mode
will be (cid:96) = (φ1 − φ2)/β. In order to accurately detect the
OAM mode, the angle β between the two test points should
satisfy the relationship β < π/(cid:96). The phase gradient method
requires that the center of the receiving circle is aligned with
the center axis of the vortex beam, otherwise there will be
large errors. Generally, it can only be used to detect single-
mode OAM beams, but with some modifications, it can also
achieve accurate measurement of two mixed modes [90].
UCAAs are used not only for radio OAM generation but
also for OAM detection [92]. The condition for detection
is that the two arrays are aligned, i.e, the cross section of
the receiving circular array C should be perpendicular to
the beam axis and its center should coincide with the axis.
UCAAs can demultiplex several radio OAM modes. The
whole process of OAM demultiplexing can be seen as some
sort of spectral analysis of the received vortex electromagnetic
waves. When the mode number (cid:96)(cid:48) of the circular array C is
different from one of the beam modes, the result of the spectral
analysis tends to 0, conversely, it tends to 1. By selecting
different (cid:96)(cid:48) values, different
information can be extracted
from the OAM multiplexed beam. When multiple topological
charge values are selected at the same time, composite mode
detection can be realized. In general, the detection process
needs to receive information of the entire wavefront. Due
to the divergence of OAM beams, a large receiving array is
required to capture the effective power of the OAM beam
in the far-field, but this might be difficult to achieve. An
answer to the problem of OAM detection in the far-field
(cid:3)(cid:3)(cid:3)(cid:3)(cid:3) 14
TABLE VI: Comparison of Radio OAM Detection Methods.
Features
Inverse SPPs
Interference
phase detection
Cost
Low
Low
Single-point
estimation
method
Low
Phase gradient
method
Low
Sampling range
Entire wavefront
Two points
One point
Two points
UCAAs
PASR
High
Entire wave-
front
Composite
mode
High
Partial wavefront
Composite mode;
There is a restri-
ctionon the use
of OAM mode
OAM mode
Single mode
System comple-
xity
Computational
complexity
Low
Low
Composite mode;
The higher the n-
umber, the more
complicated the
device
Single mode;
It is more su-
itable for low
mode
Single mode;
There is a restri-
ction on the use
of OAM mode
High
Low
Low
High
Low
Low
High
Low
High
Low
is partial aperture sampling receiving (PASR) [91], which
uses partial wavefront information to detect and distinguish
different radio OAM modes, as shown in Fig. 20, and has
been verified by computational simulation. The PASR, which
is a combination of the partial angular receiving aperture
method for OAM demultiplexing in optics [93] and a sampling
receiving scheme, has a receiving arc which is 1/P of a
circumference and employs M antennas evenly distributed on
the arc as signal sampling points. The angle between adjacent
antennas is 2π/M P . To ensure strict orthogonality and realize
non-crosstalk separation of the two OAM modes (cid:96)n1 and
(cid:96)n2, they must satisfy simultaneously the two conditions:
mod ((cid:96)n1 − (cid:96)n2, P ) = 0 and mod ((cid:96)n1 − (cid:96)n2, M P ) (cid:54)= 0.
Although the PASR has certain limitations,
it can greatly
simplify the scale of the receiving end and is robust to non-
ideal OAM beams.
Table VI compares the common detection methods for
radio OAM. Unlike the case of optical OAM detection, these
methods often require a heavy computational load so that also
computational complexity is one of the parameters evaluated
in Table VI. For multiple OAM modes detection, the system
complexity of interference phase detection is very high. The
single-point estimation method extrapolates the OAM mode
in the far field from the measurements of the two components
of the electric field, so its computational complexity is rather
high. Therefore, these two methods are not recommended for
practical implementations. In term of complexity, the phase
gradient method is probably the best method, but
it can
only detect single mode OAM. Although UCAAs need to
receive the entire wavefront information, it can be used for
composite OAM mode detection. The PASR can be seen as an
improvement of UCAAs. It has proved its advantages through
computational simulation, but further experimental verification
is required. In radio OAM communications, inverse SPPs are
widely used to detect OAM modes. In the particular case of
line-of-sight (LOS) wireless communications, UCAAs have
proved to be more suitable for demultiplexing composite OAM
modes.
C. Acoustic OAM
As mentioned above, research on OAM detection at acoustic
frequencies is still in an early stage. Due to the orthogonality
between the OAM eigenstates, Shi et al. use the inner product
operation to separate the multiplexed OAM modes received
by a transducer array [94]. Recently, Jiang et al. used passive
metamaterials to detect acoustic OAM [95]. The structure of
such passive acoustic metamaterials has been described in
Section III. After that an acoustic beam is passed through
a metamaterial with the topological charge of (cid:96) = −1, the
mode number of all multiplexed OAM modes will be reduced
by 1. Therefore, m metamaterials with (cid:96) = −1 can convert
the component with the mode number (cid:96) = m into a plane
wave and a non-zero value can be detected at the "dark" core
point of the OAM beam. This non-zero value contains the
information carried by the OAM with (cid:96) = m. This method can
realize real-time demodulation of OAM beams and has high
conversion efficiency but low flexibility and is not suitable for
detection of high-mode OAM.
D. Summary and Open Challenges
Table VII summarizes the most common methods employed
for the detection of OAM modes. As seen from the table, for
optical OAM and radio OAM, many detection methods have
been proposed, while further research, investigation and testing
is needed for developing effective acoustic OAM detection.
In general, using the inverse SPP is the easiest way to detect
optical OAM and radio OAM. In optical OAM communi-
cations, holographic gratings or phase holograms loaded on
SLMs, which can detect multiple OAM modes simultaneously,
are used in most cases. In radio OAM communications, in
addition to SPPs, UCAAs have the potential to demultiplex
composite OAM modes. However, OAM beam divergence
increases the aperture of the receiving UCAA. Reducing the
aperture and scale of the far field receiving array is extremely
challenging. The PASR and the phase gradient method are
promising solutions to reduce the scale of the receiver. But
15
TABLE VII: Summary of OAM Detection Methods.
Methods
Holographic gratings
Introduction and Features
Holographic gratings with exp(−i(cid:96)θ) phase factor can convert vortex beams into
plane beams, and Daman gratings can realize parallel detection of multiple OAM
modes.
Simple and fast; low conversion efficiency, no higher than 1/N, where N is the total
number of detected modes.
Remarks
Optical OAM
Dove prism interferom-
eters
Mode sorters
Inverse SPPs
By adjusting the rotation angle of the Dove prism, the odd-even sort of OAM modes
can be achieved on a single photon level.
High conversion efficiency, close to 100%; any number of OAM modes can be
detected by cascading multiple interferometers, while increasing the complexity of
the device.
The geometric transformation is used to map the Cartesian coordinate to the log-polar
coordinate, and ring beams with helical wavefronts are mapped to rectangular beams
with tilted wavefronts, which are focused by lens in different lateral positions.
High conversion efficiency; adjacent OAM mode cannot be effectively separated.
SPPs with the topological charge of −(cid:96).
Simple structure; single mode detection; composite mode detection can be realized
by combining multiple inverse SPPs and beam splitters.
Interference phase de-
tection
The phase shift cable is used to compensate the phase difference between the receiving
antennas, so that the specific OAM mode has high intensity and the other modes have
low intensity.
Intuitive; the receiving device is very complicated when detecting multimodal OAM.
Single-point estimation
method
It uses an approximation for OAM in the far-field to estimate the OAM from the
measurements at a single point of the vertical and the transverse components of the
electric field.
It is efficient for the detection of OAM with low mode numbers.
Phase gradient method
UCAAs
PASR
Metamaterials
Estimate the OAM mode by computing the phase difference of the electric field
components at two test points separated by an angle β, i.e. (cid:96) = (φ1 − φ2)/β.
Low complexity; small scale; the angle β between the test points should satisfy β <
π/((cid:96)).
A receiving circular array C with (cid:96) = (cid:96)(cid:48) can extract the information carried by the
beam with (cid:96) = (cid:96)(cid:48) from the multiplexed OAM beam.
The detection process needs to receive information of the entire wavefront.
Demultiplexing is implemented by utilizing partial wavefront information and the
orthogonality between OAM modes.
Small scale; any two OAM modes (cid:96)n1 and (cid:96)n2 must satisfy mod ((cid:96)n1−(cid:96)n2, P ) =
0 and mod ((cid:96)n1 − (cid:96)n2, M P ) (cid:54)= 0.
M metamaterials with (cid:96) = −1 can convert the component with the mode number
(cid:96) = m into a plane wave.
Realtime; high conversion efficiency; low flexibility; specific working frequency; it is
not suitable for detection of high-mode OAM.
Optical OAM
Optical OAM
Optical OAM
Radio OAM
Radio OAM
Radio OAM
Radio OAM
Radio OAM
Radio OAM
Acoustic OAM
how to apply them to OAM mode demultiplexing requires
further research in the coming years.
V. APPLICATIONS OF OAM IN COMMUNICATIONS
This section focuses on the applications of OAM in com-
munications. Since OAM modes with different values of (cid:96) are
mutually orthogonal, vortex beams carrying different OAM
modes can provide independent communication channels for
efficient information transmission. As shown in Fig. 21, there
are two main strategies to transfer information with OAM:
OAM shift keying (OAM-SK), where the information is en-
coded in the value (cid:96) of the OAM beam, and OAM division mul-
tiplexing (OAM-DM), where the information is multiplexed in
multimodal OAM beams. OAM-DM uses the vortex beams
as the carrier of information so that data is loaded onto
different OAM modes and then multiplexed and transmitted
coaxially through a single aperture. At the receiver, the beams
are collected by another aperture and then demultiplexed and
detected for data recovery. Compared with OAM-SK, OAM-
DM system has higher spectral efficiency and exhibits lower
bit error rates.
OAM-DM as a mode-division multiplexing (MDM)
is
a subset of space-division multiplexing (SDM) and can
be compatible with different modulation formats, such as
16
Fig. 21: Schematic diagram of OAM-SK and OAM-DM communications: in OAM-SK communication (top) the information
symbols are encoded in the topological charges of OAM beams, while in OAM-DM communication (bottom) the information
is multiplexed in multimodal OAM beams.
M-ary amplitude-shift keying (M-ASK), M-ary phase-shift
keying (M-PSK) and M-ary quadrature amplitude modula-
tion (M-QAM), as well as other multiplexing techniques
such as frequency-division/wavelength-division multiplexing
(FDM/WDM) and polarization-division multiplexing (PDM),
thereby further improving the communication system capacity
from another dimension. In the OAM-DM system, N OAM
waves carrying information are multiplexed, and the obtained
field can be expressed as
N(cid:88)
UM U X (r, θ, t) =
Sp(t)Ap(r)ei(cid:96)pθ,
(7)
p=1
where Sp(t) is the modulated data signal on the pth OAM
mode and Ap(r) is the complex electric field amplitude of the
pth OAM mode. Then, at the receiving end, the multiplexed
OAM wave is multiplied by an anti-helical phase factor
exp(−i(cid:96)qθ), which is achieved by using an inverse SPP or
an antenna array. The OAM wave with (cid:96) = (cid:96)q is converted
into a plane wave and can be easily separated from other
vortex waves with (cid:96) = (cid:96)p − (cid:96)q, thus achieving demultiplexing.
Eventually, the data information carried by the OAM wave
with (cid:96) = (cid:96)q can be obtained.
OAM communications still face many challenges and tech-
nical issues, such as misalignment and OAM beam divergence.
OAM waves are required to be transmitted and received coaxi-
ally in OAM-DM system, so that precise alignment is required
between the transmitter and receiver, i.e. the center of the
receiver coincides with the center of the transmitted beam and
the receiver is perpendicular to the line connecting the centers.
Moreover, since optical components usually have a limited
aperture size, OAM beam divergence during propagation can
result in received power loss in the far field, limiting the
link achievable distance. The atmospheric turbulence, mode
coupling and multipath effects should to be considered in free-
space optical links, fiber links and radio communication links,
respectively. For acoustic OAM communications, it is clear
that underwater creatures and turbulence have a great impact
on links, although there is currently no research to discuss
them. A taxonomy diagram of the applications discussed in
this section, together with main challenges and solutions for
each application, is shown in Fig. 22.
A. Free-Space Optical OAM Communications
to the
charge, which belongs
In 2004, Gibson et al. [69] demonstrated for the first
time that OAM-SK modulation can be used for
free-
space optical communications with good results. The trans-
mitter maps the information data on the OAM beams'
set L =
topological
{−16,−12,−8,−4, +4, +8, +12, +16}. At the receiving end,
two vertically stacked forked gratings are used to detect the
transmitted OAM modes. To compensate for small perturba-
tions in alignment, a Gaussian beam with (cid:96) = 0 is used as
a reference signal. In [96] 16 superimposed OAM modes are
employed to transmit information over a 3 km intra-city link.
At the receiver the beam topological charge is recovered by
a non-coherent detection scheme aided by an artificial neural
network. This experimental scheme has recently been extended
to a 143 km free-space link between the two Canary Islands
of La Palma and Tenerife [97].
If an OAM-SK system with L different topological charges
can transmit up to log2(L) bits per beam, an OAM-DM system
that multiplexes L modes can transmit up to L bits per beam.
For example, consider the case with L = {(cid:96)0, (cid:96)1}, i.e., L = 2.
With OAM-SK it is possible to transmit one bit per beam (0 if
(cid:96) = (cid:96)0 and 1 if (cid:96) = (cid:96)1). With OAM-DM, the on and off states
of each OAM mode can represent either a "1" or a "0", so
that the two-bit set of {00, 01, 10, 11} can be mapped on the
multiplexed modes {00, 0(cid:96)1, (cid:96)00, (cid:96)0(cid:96)1}. Using this idea, [99]
(cid:17932)(cid:3434)(cid:1376)(cid:16849)(cid:8965)(cid:9083)(cid:18431)(cid:11109)(cid:3434)(cid:11444)(cid:1108)(cid:1114)(cid:2102)(cid:18431)(cid:1376)(cid:16849)(cid:12017)(cid:3434)(cid:2102)(cid:18431)(cid:712)(cid:16849)(cid:12743)(cid:5575)(cid:2090)(cid:8273)(cid:5439)(cid:2586)(cid:17970)(cid:11096)(cid:1214)(cid:11805)(cid:11109)(cid:1702)(cid:7601)(cid:4480)(cid:3929)(cid:13551)(cid:1239)(cid:11087)(cid:11444)(cid:4649)(cid:12320)(cid:9169)(cid:7163)(cid:11109)(cid:12017)(cid:3434)(cid:712)(cid:4649)(cid:1406)(cid:8273)(cid:5681)(cid:11444)(cid:17932)(cid:3434)(cid:7920)(cid:9083)(cid:7032)(cid:7628)(cid:7460)(cid:4013)(cid:11109)(cid:12017)(cid:11560)(cid:1405)(cid:7903)(cid:5334)(cid:8965)(cid:17994)(cid:17911)(cid:9083)(cid:18431)(cid:1108)(cid:9961)(cid:19492)(cid:11444)(cid:11560)(cid:1405)(cid:7903)(cid:5334)(cid:7573)(cid:1376)(cid:16849)(cid:8273)(cid:5439)(cid:712)(cid:142)(cid:3404)(cid:4666)(cid:2038)(cid:2869)(cid:3398)(cid:2038)(cid:2870)(cid:4667)(cid:512)(cid:574)(cid:3407)(cid:590)(cid:512)(cid:513)(cid:142)(cid:513)(cid:19581)(cid:1406)(cid:1206)(cid:6613)(cid:7014)(cid:12575)(cid:11444)(cid:3901)(cid:7538)(cid:5334)(cid:712)(cid:1398)(cid:1072)(cid:14428)(cid:2586)(cid:14125)(cid:11096)(cid:7573)(cid:7920)(cid:9083)(cid:2437)(cid:8273)(cid:5681)(cid:8978)(cid:7567)(cid:712)(cid:2620)(cid:7206)(cid:9083)(cid:18431)(cid:9961)(cid:19492)(cid:11444)(cid:3945)(cid:16386)(cid:574)(cid:5316)(cid:9489)(cid:17379)(cid:574)(cid:3407)(cid:590)(cid:512)(cid:513)(cid:142)(cid:513)(cid:3)(cid:11109)(cid:12017)(cid:3382)(cid:5522)(cid:3929)(cid:13551)(cid:19557)(cid:2119)(cid:6613)(cid:7014)(cid:8965)(cid:8273)(cid:5439)(cid:7072)(cid:1130)(cid:1864)(cid:4593)(cid:11444)(cid:6613)(cid:7014)(cid:3382)(cid:19557)(cid:14125)(cid:3919)(cid:1278)(cid:3901)(cid:11096)(cid:8978)(cid:7567)(cid:1117)(cid:6656)(cid:2566)(cid:2090)(cid:1864)(cid:3404)(cid:1864)(cid:4593)(cid:8978)(cid:7567)(cid:6762)(cid:5206)(cid:11444)(cid:1553)(cid:5791)(cid:1072)(cid:8529)(cid:2586)(cid:14125)(cid:7920)(cid:9083)(cid:1072)(cid:12285)(cid:10409)(cid:4554)(cid:11444)(cid:5681)(cid:712)(cid:19760)(cid:16305)(cid:6613)(cid:7014)(cid:7076)(cid:1114)(cid:8978)(cid:19557)(cid:19858)(cid:11444)(cid:1553)(cid:5791)(cid:712)(cid:17932)(cid:3434)(cid:5316)(cid:11096)(cid:2567)(cid:19584)(cid:11109)(cid:12017)(cid:18200)(cid:2102)(cid:4484)(cid:5556)(cid:18423)(cid:7783)(cid:6613)(cid:7014)(cid:8965)(cid:2137)(cid:11096)(cid:18200)(cid:2102)(cid:8978)(cid:19557)(cid:19858)(cid:1553)(cid:5791)(cid:2748)(cid:8273)(cid:5439)(cid:19492)(cid:11444)(cid:8595)(cid:1236)(cid:5719)(cid:7573)(cid:4558)(cid:10720)(cid:16403)(cid:3901)(cid:11096)(cid:712)(cid:3901)(cid:11096)(cid:11444)(cid:8273)(cid:5439)(cid:2748)(cid:19760)(cid:16305)(cid:2620)(cid:7206)(cid:9489)(cid:17379)(cid:513)(cid:1864)(cid:3041)(cid:2869)(cid:3398)(cid:1864)(cid:3041)(cid:2870)(cid:513)(cid:143)(cid:145)(cid:134)(cid:19)(cid:3404)(cid:882)(cid:2748)(cid:513)(cid:1864)(cid:3041)(cid:2869)(cid:3398)(cid:1864)(cid:3041)(cid:2870)(cid:513)(cid:143)(cid:145)(cid:134)(cid:16)(cid:19)(cid:3405)(cid:882)(cid:12720)(cid:2374)(cid:1206)(cid:6613)(cid:7014)(cid:12575)(cid:11444)(cid:16372)(cid:8273)(cid:712)(cid:4649)(cid:1214)(cid:19854)(cid:10806)(cid:5923)(cid:11444)(cid:50)(cid:36)(cid:48)(cid:8978)(cid:7567)(cid:1959)(cid:7481)(cid:5560)(cid:5482)(cid:11444)(cid:21169)(cid:7938)(cid:5719)(cid:11109)(cid:12017)(cid:17333)(cid:16024)(cid:19858)(cid:1114)(cid:142)(cid:3404)(cid:3398)(cid:883)(cid:11444)(cid:17333)(cid:16024)(cid:19858)(cid:14125)(cid:3919)(cid:4662)(cid:8273)(cid:5439)(cid:7072)(cid:1130)(cid:11444)(cid:9169)(cid:7163)(cid:8978)(cid:17820)(cid:2374)(cid:1130)(cid:5283)(cid:19858)(cid:8978)(cid:4558)(cid:7206)(cid:7920)(cid:9083)(cid:712)(cid:17820)(cid:6546)(cid:7032)(cid:10679)(cid:20744)(cid:712)(cid:1398)(cid:5141)(cid:1420)(cid:7032)(cid:10679)(cid:1406)(cid:712)(cid:1085)(cid:17970)(cid:2616)(cid:20744)(cid:8273)(cid:5681)(cid:11444)(cid:7920)(cid:9083)(cid:712)(cid:1072)(cid:14428)(cid:1959)(cid:7481)(cid:10409)(cid:4554)(cid:11444)(cid:5141)(cid:1420)(cid:20161)(cid:9961)(cid:3872)(cid:3416)(cid:17994)(cid:1553)(cid:1117)(cid:5316)(cid:11096)(cid:7530)(cid:2150)(cid:1131)(cid:16305)(cid:2102)(cid:1130)(cid:1108)(cid:12285)(cid:726)(cid:19294)(cid:6615)(cid:712)(cid:2748)(cid:3914)(cid:17439)(cid:3901)(cid:11096)(cid:712)(cid:452)(cid:7263)(cid:4662)(cid:9169)(cid:7163)(cid:8978)(cid:7567)(cid:1420)(cid:1130)(cid:1553)(cid:5791)(cid:13638)(cid:11825)(cid:11444)(cid:17837)(cid:1411)(cid:712)(cid:5418)(cid:12539)(cid:7072)(cid:4487)(cid:1553)(cid:2599)(cid:1086)(cid:5681)(cid:11444)(cid:7248)(cid:4660)(cid:1955)(cid:13099)(cid:452)(cid:5284)(cid:712)(cid:12665)(cid:1258)(cid:16881)(cid:7230)(cid:1206)(cid:13638)(cid:11825)(cid:1553)(cid:5791)(cid:11444)(cid:1913)(cid:7567)(cid:14125)(cid:3919)(cid:11096)(cid:1214)(cid:14362)(cid:11105)(cid:12458)(cid:19492)(cid:1553)(cid:5791)(cid:1360)(cid:17859)(cid:452)(cid:3344)(cid:1130)(cid:1085)(cid:2620)(cid:7076)(cid:7072)(cid:1644)(cid:11444)(cid:5681)(cid:19492)(cid:7263)(cid:11560)(cid:1218)(cid:8595)(cid:1236)(cid:11444)(cid:712)(cid:6256)(cid:1301)(cid:6762)(cid:5206)(cid:1085)(cid:2620)(cid:11444)(cid:9169)(cid:7163)(cid:8978)(cid:7567)(cid:14125)(cid:3919)(cid:6656)(cid:1483)(cid:10524)(cid:12539)(cid:11444)(cid:17994)(cid:1553)(cid:17994)(cid:18051)(cid:712)(cid:4558)(cid:10720)(cid:20744)(cid:7032)(cid:1553)(cid:5791)(cid:1360)(cid:17859)(cid:452)(cid:4705)(cid:7263)(cid:4662)(cid:9169)(cid:7163)(cid:8978)(cid:7567)(cid:1420)(cid:1130)(cid:1553)(cid:5791)(cid:16947)(cid:2150)(cid:11444)(cid:17837)(cid:1411)(cid:712)(cid:7072)(cid:6558)(cid:16091)(cid:2256)(cid:17837)(cid:2144)(cid:1085)(cid:2620)(cid:11444)(cid:5681)(cid:1082)(cid:1981)(cid:17931)(cid:15996)(cid:3901)(cid:11096)(cid:1360)(cid:17859)(cid:452)(cid:3914)(cid:17439)(cid:3901)(cid:11096)(cid:7263)(cid:12458)(cid:2102)(cid:3901)(cid:11096)(cid:11444)(cid:1072)(cid:1114)(cid:4480)(cid:19702)(cid:712)(cid:10806)(cid:16874)(cid:1082)(cid:14125)(cid:3919)(cid:1086)(cid:1958)(cid:1286)(cid:3901)(cid:11096)(cid:6320)(cid:7519)(cid:11560)(cid:1964)(cid:4585)(cid:712)(cid:4018)(cid:20161)(cid:2102)(cid:3901)(cid:11096)(cid:8978)(cid:2102)(cid:3901)(cid:11096)(cid:451)(cid:1663)(cid:2102)(cid:3901)(cid:11096)(cid:452)(cid:3416)(cid:11560)(cid:2620)(cid:11444)(cid:20161)(cid:10679)(cid:8978)(cid:19375)(cid:6214)(cid:1663)(cid:6495)(cid:5681)(cid:1082)(cid:1455)(cid:11096)(cid:1085)(cid:2620)(cid:11444)(cid:1553)(cid:18051)(cid:7573)(cid:1360)(cid:17859)(cid:1553)(cid:5791)(cid:712)(cid:2591)(cid:1301)(cid:6208)(cid:1597)(cid:3424)(cid:6656)(cid:20744)(cid:17994)(cid:1553)(cid:13099)(cid:13583)(cid:4585)(cid:18431)(cid:452)(cid:11560)(cid:17843)(cid:1214)(cid:712)(cid:13099)(cid:13583)(cid:1553)(cid:5791)(cid:1360)(cid:17859)(cid:17999)(cid:5334)(cid:5659)(cid:451)(cid:20161)(cid:16993)(cid:7032)(cid:10679)(cid:20744)(cid:451)(cid:16927)(cid:11825)(cid:10679)(cid:1406)(cid:452) (cid:3374)18 OAM-SK(cid:2748)OAM-DM(cid:17994)(cid:1553)(cid:13099)(cid:13583)(cid:12138)(cid:5951)(cid:3374)(cid:3398) (cid:5284)(cid:712)(cid:12665)(cid:1258)(cid:6656)(cid:2090)(cid:1072)(cid:12285)(cid:3626)(cid:1214)(cid:11444)(cid:17994)(cid:1553)(cid:7145)(cid:7800)(cid:712)(cid:20422)(cid:8529)(cid:4662)(cid:5316)(cid:11096)(cid:2144)(cid:1913)(cid:17994)(cid:1553)(cid:13099)(cid:13583)(cid:1117)(cid:452)(cid:1286)(cid:1308)(cid:17977)(cid:11096)(cid:451)(cid:451)(cid:451)(cid:451)(cid:451)(cid:451)(cid:451)(cid:11444)(cid:9169)(cid:7163)(cid:1913)(cid:7567)(cid:712)(cid:8703)(cid:1114)(cid:5681)(cid:1299)(cid:16024)(cid:1072)(cid:1114)(cid:7072)(cid:6558)(cid:12630)(cid:2599)(cid:712)(cid:1485)(cid:8529)(cid:17931)(cid:15996)(cid:1360)(cid:17859)(cid:712)(cid:3416)(cid:6613)(cid:7014)(cid:12575)(cid:2121)(cid:2137)(cid:11096)(cid:1108)(cid:1114)(cid:3506)(cid:11556)(cid:2576)(cid:2256)(cid:11444)(cid:2553)(cid:3515)(cid:1913)(cid:7733)(cid:7573)(cid:9083)(cid:18431)(cid:5681)(cid:452)(cid:1130)(cid:1206)(cid:1549)(cid:16881)(cid:13099)(cid:13583)(cid:8595)(cid:11934)(cid:4649)(cid:2038)(cid:712)(cid:11444)(cid:20744)(cid:7135)(cid:1913)(cid:7567)(cid:16091)(cid:11096)(cid:7573)(cid:1420)(cid:1130)(cid:2546)(cid:13875)(cid:1553)(cid:2599)(cid:452)(cid:5284)(cid:712)(cid:2748)(cid:1286)(cid:11444)(cid:2620)(cid:1211)(cid:1308)(cid:1455)(cid:11096)(cid:2748)(cid:1114)(cid:8595)(cid:17231)(cid:2576)(cid:2256)(cid:11444)(cid:5681)(cid:2539)(cid:13638)(cid:11825)(cid:1212)(cid:13604)(cid:9888)(cid:5334)(cid:3374)(cid:1791)(cid:708)(cid:2102)(cid:2139)(cid:4649)(cid:5316)(cid:8703)(cid:1791)(cid:13136)(cid:8708)(cid:10409)(cid:2748)(cid:8703)(cid:1791)(cid:13136)(cid:8708)(cid:10409)(cid:709)(cid:712)OAM with OAM with OAM with receiver with multiplexing receiver with receiver with OAM coding OAM decoding symbol =2=2=2 =1=1=1 =2=2=2 =3=3=3 0, 1, 2 symbol 0, 1, 2 demultiplexing =1=1 =2=2 =3=3 data 1 data 2 data 3 data 1 data 2 data 3 17
Fig. 22: Taxonomy diagram of applications of OAM in communications.
the modulation information data 1, data 2 and data 3, are
multiplexed into one multimodal beam. A second OAM beam
with the same set of modes L carrying the data streams data
4, data 5 and data 6 is multiplexed with the first one em-
ploying orthogonal polarizations. Finally, three polarization-
and-OAM-multiplexed beams carrying three different sets of
information streams, data 1-6, data 7-12 and data 13-18, are
transmitted at three different wavelengths λ1, λ2 and λ3. In
a recent laboratory experiment [101], a transmission rate of
2.56 Tbps and a spectral efficiency of 95.7 bps/Hz have been
achieved by using 20 × 4 Gbps 16-QAM signals on 8 OAM
modes, 2 polarization states, and two sets of concentric rings.
In [98] a transmission rate of 100.8 Tbps is achieved by
transmitting 100 Gbps quadrature phase-shift keying (QPSK)
signals on 12 OAM modes, 2 polarization states, and 42
wavelengths. In a similar fashion, in [102] a transmission rate
of 1.036 Pbps with a spectral efficiency of 112.6 bps/Hz has
been achieved by multiplexing 54.139 Gbps OFDM-8QAM
signals over 368 wavelengths, 2 polarization states and 26
OAM modes. In addition to these lab-scale high-rate data
experiments, a 400 Gbps transmission rate has been achieved
over an outdoor link of about 120m by using 100 Gbps QPSK
signals on 4 OAM modes [103].
Multimodal LG beams with different topological charges are
often employed to provide a set of orthogonal OAM modes
for information transmission. Most of the literature on the LG
beam consider azimuthal index (cid:96) > 0 and radial index p = 0.
Because the radial index p can be used as a radial degree of
freedom just as (cid:96) can provide an azimuthal degree of freedom,
LG beams with p > 0 have received some attention [104],
[105] recently. Indeed, LG beams with different p and (cid:96) form
a complete set of orthogonal mode bases. By multiplexing LG
beams with different p, the system capacity and transmission
rate can be improved. In addition to LG beams, Bessel beams
with non-diffractive properties, because of their self-healing
properties after encountering an obstruction, also have a great
application potential
in free-space optical communications
[106] -- [108]. Perfect vortex beams, obtained by the Fourier
Fig. 23: Concept of using three-dimensional multiplexing
to increase the multiplexed data channels. (a), (b), and (c)
are performed successively to achieve OAM-DM, PDM, and
WDM, respectively [98].
transmits the information in free space by multiplexing four
OAM modes employing a phase hologram loaded on a SLM.
Subsequently, this method is used to implement the encoding
of two-dimensional images [73], [100].
In order to meet the ever-increasing demands for higher
data rates, OAM multiplexing can be combined with different
modulation formats and different multiplexing techniques to
achieve high-speed communication in multiple dimension. A
data link multiplexing the signal in three dimensions is shown
in Fig. 23. Three OAM beams with L = {(cid:96)1, (cid:96)2, (cid:96)3}, carrying
OAM-SK [69], [96], [97], [107]OAM-DM [73], [98]-[106], [108], [109]PDM [98], [101], [102]WDM [98], [102], [104]Free-space optical OAM communicationsOptical OAM fiber communicationsRadio OAM communicationsApplication of OAM in communicationsAcoustic OAM communicationsHigh data rate communicationsChallenges and solutionsOAM beam divergence and misalignment [110]Atmospheric turbulence effects [111]-[116]Adaptive optics (AO) compensation [116]-[124]Signal processing-based mitigation [117], [125], [127], [128]OAM-DM [135]WDM [135]High data rate communicationsChallenges and solutionsMode coulpingDigital signal processing (DSP) algorithmNovel fibers [132]-[139]OAM-DM [88], [145]-[148]OAM-DM combined with MIMO [154]-[157]High data rate communicationsChallenges and solutionsOAM beam divergence and misalignment [159]Multipath effects [161], [162]OAM-OFDM transceiver [163]OAM-DM [94], [95]High data rate communicationsDigital signal processing (DSP) algorithmBeam steering [160]PDM [88], [145]OAM-SK [144](cid:28595) 18
Fig. 24: Concept of OAM beam divergence in free-space
optical communications. Tx, transmitted; Rx, receiver [110].
transformation of Bessel beams, have the attractive feature
that
the beam radius is independent of the OAM mode
and have recently used with success in a free-space optical
communication link [109].
Due to the unique wavefront of OAM beams, there are some
challenges in designing an OAM-based communication link in
free space, such as beam divergence and misalignement. Based
on diffraction theory, it is known that divergence occurs when
a collimated OAM beam propagates in free space. Divergence
increases as OAM mode number increases. Fig. 24 shows the
concept of OAM beam divergence in a free space commu-
nication link. It can be seen that divergence leads to system
power loss, especially when the size of the receiving aperture
is limited. A suitable transmitted beam size can be designed
for a specific OAM mode number and transmitted distance to
achieve the minimum received beam diameter [110]. However,
in long-distance propagation, a larger receiving aperture is
still expected to capture more received power. Besides beam
divergence, the misalignment between transmitter and receiver,
which can result in both power loss and mode crosstalk, also
needs to be considered. Misalignment errors generally include
lateral displacement and receiver angle errors [110], as shown
in Fig. 25. It can be found through simulation that large
lateral displacements and large receiver angle errors cause
a high power leakage into other modes, resulting in severe
crosstalk. Misalignment effects can be mitigated by increasing
OAM mode spacing, but this will result in higher received
power loss due to beam divergence. The trade-off between
system power loss and crosstalk needs to be considered in
link design. When the system power loss dominates (i.e. small
lateral displacement and receiver angular error), small mode
spacing is used, and when crosstalk dominates (i.e. large lateral
displacement and receiver angular error), large mode spacing
is used.
For a free-space optical OAM link another critical challenge
is atmospheric turbulence, caused by changes of the atmo-
sphere refractive index due to temperature and pressure non-
uniformities. Atmospheric turbulence can destroy the helical
phase front of vortex beams. Fig. 26 shows its effects on
an OAM beam, including received energy fluctuations and
crosstalk between OAM channels. The effects have been quan-
titatively analyzed using the Kolmogorov spectral statistical
model [111], [112] and experimentally verified by simulating
turbulence in laboratory [113], [114]. In [113], a thin phase
Fig. 25: Alignment between the transmitter and the receiver
for (a) a perfectly aligned system, (b) a system with lateral
displacement d, (c) a system with a receiver angular error
ϕ, and (d) a system with a transmitter pointing error θ. Tx,
transmitted; Rx, receiver [110].
Fig. 26: Concept diagram of the effects of atmospheric tur-
bulence on an OAM beam. A distorted OAM mode can be
decomposed into multiple OAM modes [113].
screen plate mounted on a rotating stage and placed in the
middle of the optical path is used as a turbulence emulator.
The pseudorandom phase distribution due to the thin plate
obeys the Kolmogorov spectral statistics and is characterized
by a parameter r0. The strength of the simulated turbulence
can be varied by using a plate with a different r0 or by
adjusting the size of the beam incident on the plate or the
number of passes through the plate. It is found that, as the
turbulence strength increases, the power of the transmitted
OAM mode leaks into neighboring modes and for strong
turbulence tends to be equally distributed among modes, which
will result in severe crosstalk at the receiver. The effects of
atmospheric turbulence on different vortex beams, including
LG and Bessel beams, have also been numerically calculated
by analyzing their OAM spectra [115] and it has been shown
that Bessel beams suffer more than LG beams in passing
through atmosphere turbulence. The non-diffractive property
of Bessel beams is affected by a strong turbulence [116].
Many approaches for mitigating the turbulence effects have
been proposed and are divided into two main categories:
adaptive optics (AO) compensation and signal processing-
based mitigation [117]. AO compensation corrects the dis-
PowerOAM Model1Pure OAM ModeDistorted OAM ModeOAM Model1 Powerl2 l3 l5 l4 10.8Atmospheric Turbulence 19
Fig. 27: (a) Schematic of an AO compensation system for OAM beams using a Gaussian probe beam for wavefront sensing,
and (b) detailed implementation of the AO system [117].
torted OAM wavefronts in the optical domain, and signal
processing-based mitigation utilizes digital signal processing
(DSP) algorithms, such as MIMO equalization, to reduce the
signal degradation in the electrical domain.
AO compensation systems usually work in a closed-loop
configuration. A typical working iteration is to first sense the
wavefront of the distorted OAM beam and then, on the base
of the received feedback, generate an error correction pattern
and apply it to the beam to undo the distortion. Based on
whether there is a wavefront sensor (WFS) to measure the
distorted OAM wavefront, AO compensation can be further
divided into WFS-based and non-WFS types.
The main challenge for the WFS-based AO compensation
is the difficulty to correctly measure the wavefront of the
helical phase front of OAM beams. Recently, a modified
AO system that employs a Gaussian probe beam has been
proposed to overcome this problem [118]. As shown in Fig.
27, the Gaussian probe and the multiplexed OAM beams are
coaxially propagated through the atmospheric turbulence so
that they all suffer the same distortion. At the receiver, the
distorted Gaussian probe is separated and sent to the WFS
for the estimation of wavefront distortion and the retrieval
of the required correction pattern. At the transmitter the two
wavefront controllers are updated with the fedback correction
pattern. It should be noted that, for efficient separation, the
Gaussian probe should occupy a separate orthogonal channel,
which can be an orthogonal polarization [118] or a separate
wavelength [119]. However, the use of an orthogonal polar-
ization channel sacrifices the polarization degree of freedom
for multiplexing. It has been shown that using a separate
wavelength channel, the compensation performance degrades
slowly with the increase in the probe's wavelength offset from
the OAM beams [119]. Moreover, the AO system shown in
Fig. 27 can be used not only to post-compensate but also
to pre-compensate the distorted multiplexed OAM beams in
the bidirectional free-space optical communication link [120].
The AO compensation has been experimentally demonstrated
for multiple Bessel beams through atmospheric turbulence and
obstructions. Fortunately, the self-healing property of Bessel
beams can be recuperated after turbulence compensation [116].
In addition to WFS sensing, the distortion wavefronts can
also be retrieved from measured intensity profiles by using
phase retrieval algorithms, such as the Gerchberg-Saxton (GS)
algorithm [121] -- [123] and stochastic-parallel-gradient-descent
(SPGD) algorithm [124]. The GS algorithm, a well-known
iterative algorithm, can be used to analyze the original and
distorted probe Gaussian intensity patterns to obtain a correc-
tion phase pattern, which can be used for pre-compensation of
the distorted OAM beam through the same turbulence [122].
If the complex amplitude of the transmitted vortex beam,
including the "doughnut" amplitude and the helical phase (or
mixed helical phase), is known, this method can be generalized
to pre-turbulence compensation without
the addition of a
Gaussian probe [123]. Moreover, the phase correction pattern
can also be derived by using the distorted OAM beam intensity
pattern with the SPGD algorithm [124].
Signal processing-based algorithms at the receiver can also
help mitigate atmospheric turbulence effects and partially
shift the complexity of the optical subsystem to the elec-
trical domain, providing a complementary approach to AO
compensation systems. A 4 × 4 adaptive MIMO equalization
system has been implemented in a 4-channel OAM multi-
plexed free-space optical link to reduce crosstalk caused by
weak turbulence [125]. The used MIMO DSP is similar to
the one that has been used in few-mode and multi-mode fiber
multiplexing systems for mitigating the mode coupling effects
[126]. Experimental results show that all data channels can
be recovered and the system power penalty is reduced after
MIMO equalization. However, MIMO equalization is not uni-
versally useful. Under strong turbulence distortion, outage may
occur because crosstalk between channels exceeds a certain
threshold or severe power attenuation causes some channels to
be barely detectable, in which case MIMO equalization will
be ineffective [127]. A modified method is proposed in [128]
to improve system performance under strong turbulence. The
proof-of-concept experiment shows that, in an OAM-based
spatial diversity free-space optical link, the OAM channel can
be recovered under strong turbulence distortion by using a
diversity reception strategy assisted with MIMO equalization
[128].
B. Optical OAM Fiber Communications
As mentioned above, OAM free-space optical communi-
cation links are greatly affected by beam divergence and
20
atmospheric turbulence. In recent years, research has focused
on OAM-based optical fiber communications, which do not
need to take these effects into account. Conventional optical
fiber communications utilize SDM by using multicore fibers
(MCFs) [129], [130] and few-mode fibers (FMFs) [131] to
increase system capacity and spectral efficiency. Generally,
multicore fibers require complex manufacturing, while multi-
mode fibers face the problem of mode coupling caused by
random perturbations or other nonidealities. OAM modes, just
like linear polarization (LP) modes in fibers, can be used as
an orthogonal basis for data transmission in FMFs. Like LP
modes, also OAM modes face the challenge of mode coupling,
which leads to channel crosstalk. One possible solution is to
use DSP algorithms based on complex MIMO equalization.
The other is to use a specially designed FMF, which is called a
vortex fiber. The vortex fiber lifts the near-degeneracy between
OAM modes and the parasitic TE01 and TM01 by modifying
the fiber refractive index profile, minimizing mode crosstalk
[132]. Therefore, by using vortex fibers, OAM multiplexing
technology has the potential to increase the throughput of
optical fiber communication systems with low complexity DSP
algorithms or even without any DSP algorithm.
Using a vortex fiber with a characteristic high-index ring
around the fiber core, Bozinovic et al. have performed exper-
iments of OAM mode transmission in fibers 20 m and 900
m long [133], [134]. In the experiments, a microbend grating
is used before the vortex fiber to achieve the conversion from
the fundamental modes to the desired HE21 modes. The linear
combination of the odd and even modes of HE21 with a ±π/2
phase shift between them results in OAM modes with (cid:96) = ±1.
In a later experiment, two OAM modes with (cid:96) = ±1 and two
LP modes, each mode carrying a 50 Gbaud QPSK signal, are
simultaneously propagated to achieve a transmission rate of
400 Gbps in a 1.1 km long vortex fiber [135]. At the output
of the fiber, the measured mode crosstalk between two OAM
modes is approximately −20dB. In addition, WDM has also
been added to further increase the capacity of the system. By
using two OAM modes over 10 wavelengths in a vortex fiber,
20 channels, each transmitting a 20 Gbaud 16-QAM signal,
are established and a transmission rate of 1.6 Tbps is achieved.
These experiments show that OAM can provide an additional
degree of freedom for data multiplexing in fiber networks.
At present, research on OAM-based optical fiber commu-
nications mainly focuses on the design of fibers that sup-
ports stable transmission of multiple OAM modes. In 2012,
Birnbaum et al. designed a ring fiber with 0.05 up-doping
to support up to 10 OAM modes, while maintaining radial
single-mode conditions [136]. In 2013, Li et al. designed a
multi-OAM-mode multi-ring fiber (MOMRF) that supports
multi-mode OAM transmissions [137]. As shown in Fig. 28,
the fiber consists of seven rings, each supporting 18 OAM
modes. Mode crosstalk and inter-ring crosstalk are reduced
by increasing the effective refractive index and the distance
between rings. In addition, it is compatible with WDM and ad-
vanced multilevel amplitude/phase modulation formats, which
makes it possible to realize a total transmission capacity in the
range of the petabits-per-second and hundredbits-per-second-
per-hertz aggregate spectral efficiency. In addition to these ring
Fig. 28: (a) 3D structure and cross-section of multi-OAM-
mode multi-ring fiber (MOMRF), in which the ring-to-ring
distance is Λ. (b) Index profile of single ring (black) and mode
profile of TE01 mode (red) in the ring [137].
fibers, some new types of microstructured fibers have also been
proposed for multi-mode OAM transmission [138], [139].
Recently, some studies have explored the potential of the
conversion between LP modes and OAM modes in fibers. This
can be used as a complement to the OAM generation method,
and it is easier to couple than the methods mentioned above.
In [140], Zeng et al. designed a novel all-fiber OAM generator,
which is cascaded by a mode-selective coupler and a few
mode-polarization maintaining fiber, to convert LP01 mode to
OAM mode. In [141] a series of LP11 modes with microphase
difference distribution are generated by twisting a few-mode
fiber long period grating (FMF-LPG), these LP11 modes are
then superimposed in a fiber to generate OAM modes. In [142]
Li et al. proposed and demonstrated a controllable broadband
fiber-based OAM converter. As shown in Fig. 29, the converter
consists of a single-mode fiber (SMF), a two-mode fiber
(TMF) with specific offsets and tilt angles, two polarization
controllers (PCs) and a polarizer. The input end of TMF is
stuck to a standard SMF. By adjusting the two PCs, an input
fundamental mode LP01 can be selectively converted to high
order LP11 modes or OAM modes with (cid:96) = ±1. The purity
of the OAM mode is ensured by adjusting the state of the
two PCs and the polarizer. The experimental results show that
the extinction ratio (used to evaluate the mode purity) of the
generated OAM mode and other modes is greater than 20 dB
in a wide wavelength range (1480 nm to 1640 nm).
C. Radio OAM Communications
The application of OAM to radio communications is ex-
pected to be a possible solution to the problem of spectrum
scarcity and it has, accordingly, received widespread attention.
The first OAM-based wireless communication experiment that
successfully separated two radio signals at the same frequency
was presented in [34]. A Yagi antenna and a spiral parabolic
antenna were used to transmit a plane electromagnetic wave
and a vortex electromagnetic wave, respectively, at a distance
of 442 meters. Subsequently, a 4 Gbps uncompressed video
transmission link over a 60 GHz OAM radio channel has
been implemented [143]. In the RF or millimeter wave bands,
provided that there is a LOS link so to guarantee the correct
transmitter-receiver alignment, OAM can be used to encode
21
Fig. 29: Controllable broadband fiber-based OAM converter. The converter consists of a SMF, a TMF wih specific offsets and
tilt angles, two PCs and a polarizer, and the input end of TMF is stuck to a standard SMF [142].
receivers have a lower complexity because, unlike traditional
MIMO systems,
the inherent orthogonality between OAM
modes can mitigate inter-channel interference [149]. A sketch
of a system employing OAM-MDM in a LOS MIMO channel
is shown in Fig. 30, where phase shifters networks (PSNs)
are used to multiplex OAM modes at the transmitter and
demultiplex OAM modes with a reduced computational load at
the receiver. Moreover, as described in Section III, radio OAM
can be generated not only by employing antenna arrays, but
also by using SPPs or spiral parabolic antennas or other single
transmit antennas. Combining OAM with traditional MIMO
technology can result in higher capacity gains [154], [155] or
provide a more flexible system design [156], [157]. A 2 × 2
antenna aperture architecture, where each aperture multiplexes
two OAM modes, is implemented in the 28 GHz band to
achieve a 16 Gbit/s transmission rate. At the receiving end,
MIMO signal processing is employed to mitigate inter-channel
interference [157]. Such a system architecture incorporating
OAM multiplexing and MIMO technology can also be used
for free-space optical communication links [158].
As it happens for optical OAM, radio OAM communication
systems are affected by beam divergence and misalignment
between the transmitter and the receiver. Since wavelengths
at radio frequencies are much larger than at light frequen-
cies, the effects of atmospheric turbulence on radio OAM
decrease and can be usually neglected. Unfortunately, for
the same reason the problem of beam divergence becomes
more important and represents a bigger challenge in wireless
communication systems than in free-space optical commu-
nication systems, limiting the achievable distance of OAM
radio links. Like in free-space optical communications, radio
OAM communication systems also require perfect alignment
between the transmitter and the receiver [159]. To prevent large
performance degradation due to misalignment, one possible
solution in UCAA-based OAM links is to use a beam steering
approach [160]. By adding additional phases, the UCAA can
transmit/receive OAM beams in the desired steering direction.
In the non-parallel case receive beam steering can compensate
the phase change caused by oblique angle at
the receive
UCAA. The off-axis case, where the transmitter and receiver
are parallel but not around the same axis, can be decomposed
into two non-parallel cases by introducing a virtual UCAA
Fig. 30: System model for a UCAA-based OAM-MDM in
a LOS MIMO channel. The OAM-MDM system uses PSN1
and PSN2 for OAM modes multiplexing and demultiplexing,
respectively. The vectors xl and yl are the complex transmitted
and received data vectors, respectively, and the matrix H is the
LOS MIMO channel matrix [149].
information [144], and can also be multiplexed and combined
with other technologies [88], [145], [146]. In [88] are reported
the results of an experiment in the 28 GHz band that has
achieved a 32 Gbit/s capacity and about 16 bit/s/Hz spectral
efficiency by combining 4 OAM modes each carrying 4
Gbit/s 16-QAM modulated signals with 2 polarization states.
In recent experiments, multiple radio OAM beams carrying
multiple data streams have been generated by employing some
specially designed structures, such as thin metamaterial plates
based on rectangular apertures [147] and dual OAM mode
antennas based on ring resonators [148].
It
is well known that MIMO is a key technology that
can greatly increase the capacity of wireless communication
systems by using multiple transmitting and receiving antennas.
The relationship between MIMO SDM and OAM multiplexing
has caused a heated debate in the academia. It has been pointed
out that radio OAM systems based on UCAA are a subset of
MIMO systems and do not really provide additional capacity
gain [150], [151]. But in practice there are differences be-
tween the two approaches. For example, in a keyhole channel
traditional MIMO systems have very poor performance, while
UCAA-based OAM benefits of the fact that the keyhole does
not change the helical phase structure of OAM [152]. In terms
of capacity over a LOS link, a UCAA-based OAM system is
equivalent to a traditional MIMO system from the perspec-
tive of channel spatial multiplexing [153]. However, OAM
22
that is perpendicular to and in the middle of the connection
between the transmit and receive UCAA centers. Therefore, in
the off-axis or the more general misalignment cases, transmit
and receive beam steering need to be used simultaneously.
The beam steering approach has been demonstrated through
simulation [160] and the results show that after applying beam
steering in the non-parallel and off-axis case, the misaligned
OAM channel capacity shows almost no loss with respect to
the case of perfect alignment.
It is well known that multipath effects caused by beam
spreading and reflections of surrounding objects must be
considered in wireless communication systems. In OAM-based
wireless communications,
the multipath effects caused by
specular reflection from a plane parallel to the propagation
path have been discussed in [161] and [162]. Both simulation
and experimental results show that OAM channels with larger
(cid:96) have stronger intra- and inter-channel interference due to
OAM beam divergence. DSP equalization algorithms have
the potential to mitigate multipath effects. Another potential
solution is to use orthogonal frequency-division multiplexing
(OFDM) technology in radio OAM systems. Recently, a
transceiver architecture for broadband OAM-OFDM wireless
communication systems has been proposed in [163]. The trans-
mitter uses a baseband digital two-dimensional Fast Fourier
Transform (2-D FFT) algorithm and a UCAA to generate
OAM-OFDM signals. After multipath channel propagation,
the signal is collected by another UCAA and processed using
an inverse FFT (IFFT) algorithm. By using the 2-D FFT/IFFT
algorithms, system implementation complexity can be reduced
and multipath effects are mitigated.
D. Acoustic OAM Communications
In the field of acoustics, OAM-based communication tech-
nology is still in its infancy. In 2017, Shi et al. [94] used
8 acoustic OAM modes to transmit the letters of the word
'Berkly' in ASCII binary protocol, achieving a high spectral
efficiency of 8.0 ± 0.4 bit/s/Hz. In details, each OAM mode
carries 1-bit information of the eight bits that map each letter,
as shown in Fig. 31 (A). These 8 OAM orthogonal bases are
multiplexed and transmitted by a single transducer array. At
the receiver, another transducer array receives the signal and
then demultiplexes it by exploiting the orthogonality between
OAM modes. The recovered signal is shown in Fig. 31 (B).
Instead of using OAM modes to encode data, Jiang et al.
directly loaded the data onto the acoustic OAM channels and
established a real-time information transmission system based
on passive metamaterials [95]. These studies have shown that
in acoustic communications, OAM can be a viable alternative
to more traditional technologies.
Since microwaves and mid- and far-infrared radiations are
strongly absorbed underwater, and light is easily obstructed
and scattered by small particles in the oceans, sound waves
are the only information carrier for underwater long-distance
(more than 200 meters) communications. Accordingly, OAM
can be applied to underwater acoustic communications to fur-
ther increase system capacity and spectrum efficiency. In ad-
dressing link design for underwater acoustic communications,
Fig. 31: (A) Encoding of the letters of the word 'Berkly'.
Each letter is mapped on 1 byte (8 bits) of information, and
each byte contains the same amount of total amplitude and
this amplitude amount is equally distributed into the vortex
beams. (B) Decoding after transmission with 8 OAM modes
[94].
underwater creatures and turbulence effects should also be
considered in addition to the inherent OAM beam divergence
and misalignment problems. However, very little research and
experimental data are available in current scientific literature.
Nevertheless, considered the potential capacity gains of OAM,
theoretical analysis and experimental verification are expected
to be developed in the near future.
E. Discussion and Lessons Learned
OAM has shown great potential in the communications
community due to the inherent orthogonality between modes.
Integrating OAM into existing communication systems is
expected to further improve spectral efficiency and meet ever-
increasing data rate demands. Moreover, the typical circular
symmetry of OAM waves makes them very easy to be included
in many communication system components. However, OAM
integration presents some challenges and technical issues, such
as beam divergence, misalignment, and atmospheric turbulence
effects in free-space optical links, mode coupling in fiber links,
and multipath effects in radio communication links. These
technical issues and their solutions must be considered when
building an OAM-based communication system.
Analyzing OAM implementation on the basis of the fre-
quency bands used, we can summarize the most important
characteristics of OAM systems:
• For free-space optical communication systems, SLMs
loaded with phase holograms are recommended in trans-
mitters and receivers to flexibly multiplex and demulti-
plex OAM beams. The effects of atmospheric turbulence
(cid:3)(cid:3) on free-space optical OAM links must be considered in
practical environments. Therefore, AO compensation and
DSP-based mitigation techniques play a significant role
in improving system performance. Optical OAM fiber
communication systems can use common optical OAM
generation and detection methods, such as SPPs and
SLMs. Mode conversion in fibers can also be used to
generate OAM. It should be noted that common OAM
generators and detectors are required to be well compat-
ible with fibers. Mode coupling is a major challenge in
optical OAM fiber links. Novel fibers that support stable
transmission of multiple OAM modes and reduce mode
crosstalk should be developed in the future.
• In radio and acoustic communication systems, the flex-
ibility of UCAAs and UCTAs makes them perform
well
in generating and detecting OAM waves. How-
ever, for generating high frequency OAM waves, SPPs
are more suitable than UCAAs and UCTAs. In radio
OAM communication systems, OAM beam divergence
and misalignment between the transmitter and receiver
will greatly limit link achievable distance and degrade
system performance. OAM beam convergence and system
misalignment compensation schemes are required for
long-distance transmission. Acoustic OAM communica-
tion system is still in an initial stage and the feasibility
of acoustic OAM communications, especially underwater,
need further experimental verification. The impact of the
underwater environment on OAM links also needs further
research.
After considering all possible challenges and technical is-
sues, we are at a stage where experiments and research on
OAM systems should be more oriented to practice testing
than to develop proof of concept. At present, most OAM
experiments are performed in the controlled environments of
laboratories. Testing beyond laboratory distances and practical
deployment of OAM systems should be the focus of future
research activities. Another important issue is that, since all
communication systems keep moving towards low costs, small
sizes and high data rates, all OAM system components must
achieve a reduction in cost and size in the near future.
VI. APPLICATION OF OAM IN PARTICLE MANIPULATION
AND IMAGING
23
the total angular momentum is either ∼ 0.06 or ∼ 2.06 per
photon depending on the spin direction of the beam, resulting
in the stopping or the rotation of the particles. Similarly, using
the same method but with a different beam with (cid:96) = 3, two
different rotational velocities of the particles can be observed
[167]. In the above work, the size of the particles is larger
than the beam, and the particles are considered to be trapped
on or near the axis. When the size is smaller than the ring
beam, the particles are generally trapped in the off-axis area.
In this case, the spin of the trapped particles depends on the
SAM of the ring beam, and the ring beam with the helical
phase structure associated with OAM imposes an azimuthal
scattering force on the particle to rotate the particle around
the beam axis [168], [169].
When interacting with the absorbing particles, OAM carried
by acoustic vortices can also be transferred to the particles,
exerting a torque on them. This mechanical effect can be
applied in the form of acoustic tweezers and acoustic spanners.
Acoustic vortices can manipulate larger size particles than
optical vortices due to larger wavelength, and have great
application prospects in the field of ultrasonic medicine.
Courtney et al. implemented acoustic tweezers using first-order
and superimposed high-order Bessel acoustic vortices [170],
[171]. Their experiment uses a transducer array to generate
the required Bessel acoustic vortex field and can control the
movement of the vortex center by adjusting the drive signal
of the array element, achieving particle capture at different
locations. Unlike optical vortices, acoustic vortices cannot
carry SAM and, when vortex beams act on particles, only
OAM transfer occurs. In [56] it is experimentally demonstrated
that acoustic vortices in free space can transmit acoustic
OAM to an object and make it rotate. The experiment uses
a torsion pendulum to measure the angular momentum of
acoustic vortices and compares the effective acoustic torque
obtained with different topological charges. Moreover, in [172]
the amount of OAM transmitted by an acoustic vortex to
an absorbing disk in a viscous liquid has been quantitatively
measured.
From the perspective of simplicity and flexibility, optical
OAM for particle manipulation is usually generated using
SPPs and SLMs, while acoustic OAM for particle manipu-
lation are generated using UCTAs.
A. Optical and Acoustical Particle Manipulation
B. Optical and Radar Imaging
laureate Arthur Ashkin,
OAM has found applications also outside the field of optical,
radio and acoustic communications. Thanks to the work of
Nobel
light has since long been
used to trap small particles in what we call optical tweezers.
Optical tweezers use a single tightly-focused beam with a large
enough gradient force around the focus to overcome the linear
momentum of the light, attracting the particles toward the
center [164]. In 1995, [165] has shown that, by transferring
the OAM carried by photons to small particles, lasers with
OAM can be employed to achieve the rotation of the trapped
particles, transforming the tweezers into optical spanners that
cause the objects to spin [166]. When a circularly polarized
LG beam with (cid:96) = 1 is used to interact with the particles,
OAM applied to an optical or radar imaging system can
break the limits of resolution or sensitivity. The stimulated
emission depletion (STED) microscope uses a ring beam with
a helical phase distribution to suppress the fluorescence of
particles around the scanning point, achieving a resolution
beyond the diffraction limit [173]. Applying OAM to diffrac-
tion tomography can also make a significant breakthrough
with respect to the diffraction limits associated with traditional
techniques [174].
Inserting a spiral phase mask into the Fourier plane of
an optical imaging system can effectively convert the point
spread function of the system into an annular intensity cross
section with an exp(i(cid:96)θ) phase distribution [175]. This can
24
Fig. 32: Schematic diagram of radar target detection based on
vortex electromagnetic waves. The target position P is denoted
by (r, θ, ϕ). In the multiple-input multiple-output mode, all
antenna in the UCAA (black arrows) are used to receive the
echo signal, while only a single antenna (red arrow) is used
in the multiple-input single-output mode [182].
be used to observe the light around bright objects [176] or
to achieve spiral phase contrast imaging that is often used to
observe bright edges of phase objects [177], [178]. In [179]
this method is used to suppress the brightness of stars and
make orbital planets more visible. In addition, when a helical
phase beam is used as reference wave of an interference
system, the interference pattern is a chiral spiral stripe, so
that the protrusions and depressions of the sample surface can
be distinguished by observing the direction of rotation of the
spiral stripe [180].
In 2013, [181] applied OAM to the field of radar imaging,
and proposed the idea that vortex electromagnetic waves have
the potential for radar target imaging. The paper developed
an echo model of the ideal point scattering target under
vortex electromagnetic wave illumination and implemented
radar imaging using the back projection algorithm and the
filtered FFT based algorithm. Afterwards, [182] established
echo signal models for multiple-input multiple-output and
multiple-input single-output systems. The OAM-based radar
imaging system model is shown in Fig. 32. Assume that the
target is made up of M ideal scattering points, denoted by
Pm(rm, θm, ϕm) and corresponding radar cross-section σm,
m = 1,··· , M. In the multiple-input single-output mode, a
single antenna is set at the original point to receive the echo
signal. The normalized echo signal is expressed as
M(cid:88)
m=1
s(k, (cid:96)) =
σm
r2
m
ei2krmei(cid:96)ϕmJ(cid:96)(ka sin θm),
(8)
where k = 2πf /c is the wave number, J(cid:96)(ka sin θ) is the
Bessel function of the first kind, and a is the radius of the
UCAA. As can be seen from (8), the OAM mode number
(cid:96) and the azimuth ϕ satisfy the dual relationship, and so do
the frequency k and target range r. Thus, the 2-D FFT can be
used in estimating the range and azimuth of targets. This shows
that the OAM-based radar imaging system has the prospect of
acquiring the azimuth information of target.
OAM-based radar detection provides a new set of ideas
and solutions for the development of accurate target imaging.
Fig. 33: Comparisons of imaging results between the tradi-
tional array imaging and the electromagnetic vortex imaging:
(a) traditional array imaging of two reflectors, (b) electromag-
netic vortex imaging of two reflectors, (c) traditional array
imaging of three reflectors, and (d) electromagnetic vortex
imaging of three reflectors [185].
Due to its helical phase structure, a vortex electromagnetic
wave carrying OAM can be regarded as a traditional plane
wave illuminating a target from multiple consecutive angles,
which is equivalent to realizing continuous sampling in a two-
dimensional space in a short time, obtaining a good degree
of spatial diversity. Therefore, applying multimodal OAM
to radar imaging enables azimuthal imaging without relative
motion [183]. Moreover, when vortex electromagnetic waves
with different helical phase structures hit the target in space,
the phase differences associated to the various OAM modes
make the backscattering characteristics different and improve
the radar cross-section diversity gain [184]. In conventional
radar imaging, the azimuth resolution is generally enhanced
by increasing the aperture size, while vortex electromagnetic
waves can provide a high resolution that is not limited by
the array aperture, as it has been verified by a proof of
concept experiment in [185]. The comparison of the results of
traditional array imaging and electromagnetic vortex imaging
is shown in Fig. 33, where it is shown that the imaging
resolution is clearly improved by using vortex electromagnetic
waves. Similarly, it has been demonstrated that applying vortex
electromagnetic waves to synthetic aperture radar (SAR) imag-
ing can also achieve higher azimuthal resolution than planar
electromagnetic waves [186].
Recently, research on OAM-based radar imaging has mainly
focused on azimuth imaging algorithms. Due to the existence
of the Rayleigh limit, traditional radar imaging algorithms can
only provide limited azimuth resolution. Some efficient algo-
rithms have been proposed for OAM-based super-resolution
radar imaging, such as UCAA-based autoregressive models
and power spectral density (PSD) estimation algorithm [187],
(cid:26)(cid:20)(cid:21)(cid:44)(cid:40)(cid:40)(cid:40)(cid:36)(cid:49)(cid:55)(cid:40)(cid:49)(cid:49)(cid:36)(cid:54)(cid:36)(cid:49)(cid:39)(cid:58)(cid:44)(cid:53)(cid:40)(cid:47)(cid:40)(cid:54)(cid:54)(cid:51)(cid:53)(cid:50)(cid:51)(cid:36)(cid:42)(cid:36)(cid:55)(cid:44)(cid:50)(cid:49)(cid:47)(cid:40)(cid:55)(cid:55)(cid:40)(cid:53)(cid:54)(cid:15)(cid:57)(cid:50)(cid:47)(cid:17)(cid:20)(cid:23)(cid:15)(cid:21)(cid:19)(cid:20)(cid:24)(cid:41)(cid:76)(cid:74)(cid:17)(cid:21)(cid:17)(cid:51)(cid:75)(cid:68)(cid:86)(cid:72)(cid:16)(cid:73)(cid:85)(cid:82)(cid:81)(cid:87)(cid:71)(cid:76)(cid:86)(cid:87)(cid:85)(cid:76)(cid:69)(cid:88)(cid:87)(cid:76)(cid:82)(cid:81)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:72)(cid:79)(cid:72)(cid:70)(cid:87)(cid:85)(cid:82)(cid:80)(cid:68)(cid:74)(cid:81)(cid:72)(cid:87)(cid:76)(cid:70)(cid:89)(cid:82)(cid:85)(cid:87)(cid:72)(cid:91)(cid:90)(cid:68)(cid:89)(cid:72)(cid:17)(cid:11)(cid:68)(cid:12)(cid:17)(cid:11)(cid:69)(cid:12)(cid:17)(cid:41)(cid:76)(cid:74)(cid:17)(cid:22)(cid:17)(cid:51)(cid:75)(cid:68)(cid:86)(cid:72)(cid:71)(cid:76)(cid:86)(cid:87)(cid:85)(cid:76)(cid:69)(cid:88)(cid:87)(cid:76)(cid:82)(cid:81)(cid:73)(cid:82)(cid:85)(cid:72)(cid:79)(cid:72)(cid:70)(cid:87)(cid:85)(cid:82)(cid:80)(cid:68)(cid:74)(cid:81)(cid:72)(cid:87)(cid:76)(cid:70)(cid:89)(cid:82)(cid:85)(cid:87)(cid:72)(cid:91)(cid:90)(cid:68)(cid:89)(cid:72)(cid:86)(cid:90)(cid:76)(cid:87)(cid:75)(cid:71)(cid:76)(cid:73)(cid:73)(cid:72)(cid:85)(cid:72)(cid:81)(cid:87)(cid:73)(cid:85)(cid:72)(cid:16)(cid:84)(cid:88)(cid:72)(cid:81)(cid:70)(cid:76)(cid:72)(cid:86)(cid:17)(cid:11)(cid:68)(cid:12)(cid:42)(cid:43)(cid:93)(cid:17)(cid:11)(cid:69)(cid:12)(cid:42)(cid:43)(cid:93)(cid:17)(cid:11)(cid:70)(cid:12)(cid:42)(cid:43)(cid:93)(cid:17)(cid:41)(cid:76)(cid:74)(cid:17)(cid:23)(cid:17)(cid:51)(cid:75)(cid:68)(cid:86)(cid:72)(cid:71)(cid:76)(cid:86)(cid:87)(cid:85)(cid:76)(cid:69)(cid:88)(cid:87)(cid:76)(cid:82)(cid:81)(cid:73)(cid:82)(cid:85)(cid:72)(cid:79)(cid:72)(cid:70)(cid:87)(cid:85)(cid:82)(cid:80)(cid:68)(cid:74)(cid:81)(cid:72)(cid:87)(cid:76)(cid:70)(cid:89)(cid:82)(cid:85)(cid:87)(cid:72)(cid:91)(cid:90)(cid:68)(cid:89)(cid:72)(cid:86)(cid:90)(cid:76)(cid:87)(cid:75)(cid:71)(cid:76)(cid:73)(cid:73)(cid:72)(cid:85)(cid:72)(cid:81)(cid:87)(cid:86)(cid:72)(cid:87)(cid:16)(cid:87)(cid:76)(cid:81)(cid:74)(cid:86)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:85)(cid:68)(cid:71)(cid:76)(cid:88)(cid:86)(cid:17)(cid:11)(cid:68)(cid:12)(cid:17)(cid:11)(cid:69)(cid:12)(cid:17)(cid:11)(cid:70)(cid:12)(cid:17)(cid:90)(cid:75)(cid:72)(cid:85)(cid:72)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:70)(cid:82)(cid:81)(cid:86)(cid:87)(cid:68)(cid:81)(cid:87)(cid:70)(cid:88)(cid:85)(cid:85)(cid:72)(cid:81)(cid:87)(cid:71)(cid:72)(cid:81)(cid:86)(cid:76)(cid:87)(cid:92)(cid:89)(cid:72)(cid:70)(cid:87)(cid:82)(cid:85)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:71)(cid:76)(cid:83)(cid:82)(cid:79)(cid:72)(cid:15)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:72)(cid:79)(cid:72)(cid:70)(cid:87)(cid:85)(cid:76)(cid:70)(cid:71)(cid:76)(cid:83)(cid:82)(cid:79)(cid:72)(cid:79)(cid:72)(cid:81)(cid:74)(cid:87)(cid:75)(cid:15)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:80)(cid:68)(cid:74)(cid:81)(cid:72)(cid:87)(cid:76)(cid:70)(cid:70)(cid:82)(cid:81)(cid:71)(cid:88)(cid:70)(cid:87)(cid:76)(cid:89)(cid:76)(cid:87)(cid:92)(cid:76)(cid:81)(cid:87)(cid:75)(cid:72)(cid:89)(cid:68)(cid:70)(cid:88)(cid:88)(cid:80)(cid:15)(cid:68)(cid:81)(cid:71)(cid:68)(cid:81)(cid:71)(cid:68)(cid:85)(cid:72)(cid:87)(cid:75)(cid:72)(cid:70)(cid:76)(cid:85)(cid:70)(cid:88)(cid:79)(cid:68)(cid:85)(cid:73)(cid:85)(cid:72)(cid:84)(cid:88)(cid:72)(cid:81)(cid:70)(cid:92)(cid:68)(cid:81)(cid:71)(cid:90)(cid:68)(cid:89)(cid:72)(cid:89)(cid:72)(cid:70)(cid:87)(cid:82)(cid:85)(cid:15)(cid:85)(cid:72)(cid:86)(cid:83)(cid:72)(cid:70)(cid:87)(cid:76)(cid:89)(cid:72)(cid:79)(cid:92)(cid:17)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:37)(cid:72)(cid:86)(cid:86)(cid:72)(cid:79)(cid:73)(cid:88)(cid:81)(cid:70)(cid:87)(cid:76)(cid:82)(cid:81)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:191)(cid:85)(cid:86)(cid:87)(cid:78)(cid:76)(cid:81)(cid:71)(cid:15)(cid:15)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:68)(cid:81)(cid:74)(cid:79)(cid:72)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:72)(cid:79)(cid:72)(cid:80)(cid:72)(cid:81)(cid:87)(cid:83)(cid:82)(cid:86)(cid:76)(cid:87)(cid:76)(cid:82)(cid:81)(cid:15)(cid:68)(cid:81)(cid:71)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:85)(cid:68)(cid:71)(cid:76)(cid:88)(cid:86)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:17)(cid:55)(cid:75)(cid:72)(cid:83)(cid:75)(cid:68)(cid:86)(cid:72)(cid:16)(cid:73)(cid:85)(cid:82)(cid:81)(cid:87)(cid:71)(cid:76)(cid:86)(cid:87)(cid:85)(cid:76)(cid:69)(cid:88)(cid:87)(cid:76)(cid:82)(cid:81)(cid:82)(cid:69)(cid:87)(cid:68)(cid:76)(cid:81)(cid:72)(cid:71)(cid:69)(cid:92)(cid:81)(cid:88)(cid:80)(cid:72)(cid:85)(cid:76)(cid:70)(cid:68)(cid:79)(cid:86)(cid:76)(cid:80)(cid:88)(cid:79)(cid:68)(cid:16)(cid:87)(cid:76)(cid:82)(cid:81)(cid:86)(cid:76)(cid:86)(cid:86)(cid:75)(cid:82)(cid:90)(cid:81)(cid:76)(cid:81)(cid:41)(cid:76)(cid:74)(cid:17)(cid:21)(cid:17)(cid:55)(cid:75)(cid:72)(cid:73)(cid:85)(cid:72)(cid:84)(cid:88)(cid:72)(cid:81)(cid:70)(cid:92)(cid:76)(cid:86)(cid:86)(cid:72)(cid:87)(cid:68)(cid:86)(cid:42)(cid:43)(cid:93)(cid:15)(cid:68)(cid:81)(cid:71)(cid:87)(cid:75)(cid:72)(cid:85)(cid:68)(cid:71)(cid:76)(cid:88)(cid:86)(cid:76)(cid:86)(cid:15)(cid:90)(cid:75)(cid:72)(cid:85)(cid:72)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:90)(cid:68)(cid:89)(cid:72)(cid:79)(cid:72)(cid:81)(cid:74)(cid:87)(cid:75)(cid:17)(cid:55)(cid:75)(cid:72)(cid:69)(cid:72)(cid:68)(cid:80)(cid:86)(cid:68)(cid:85)(cid:72)(cid:86)(cid:68)(cid:80)(cid:83)(cid:79)(cid:72)(cid:71)(cid:68)(cid:87)(cid:68)(cid:71)(cid:76)(cid:86)(cid:87)(cid:68)(cid:81)(cid:70)(cid:72)(cid:73)(cid:85)(cid:82)(cid:80)(cid:87)(cid:75)(cid:72)(cid:87)(cid:85)(cid:68)(cid:81)(cid:86)(cid:80)(cid:76)(cid:87)(cid:87)(cid:76)(cid:81)(cid:74)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:15)(cid:68)(cid:81)(cid:71)(cid:87)(cid:75)(cid:72)(cid:90)(cid:76)(cid:81)(cid:71)(cid:82)(cid:90)(cid:76)(cid:86)(cid:90)(cid:76)(cid:71)(cid:72)(cid:76)(cid:81)(cid:69)(cid:82)(cid:87)(cid:75)(cid:16)(cid:68)(cid:81)(cid:71)(cid:16)(cid:71)(cid:76)(cid:85)(cid:72)(cid:70)(cid:87)(cid:76)(cid:82)(cid:81)(cid:86)(cid:17)(cid:36)(cid:70)(cid:75)(cid:68)(cid:81)(cid:74)(cid:72)(cid:76)(cid:81)(cid:70)(cid:82)(cid:79)(cid:82)(cid:85)(cid:73)(cid:85)(cid:82)(cid:80)(cid:69)(cid:79)(cid:88)(cid:72)(cid:87)(cid:82)(cid:85)(cid:72)(cid:71)(cid:68)(cid:81)(cid:71)(cid:69)(cid:68)(cid:70)(cid:78)(cid:87)(cid:82)(cid:69)(cid:79)(cid:88)(cid:72)(cid:68)(cid:74)(cid:68)(cid:76)(cid:81)(cid:70)(cid:82)(cid:85)(cid:85)(cid:72)(cid:86)(cid:83)(cid:82)(cid:81)(cid:71)(cid:86)(cid:87)(cid:82)(cid:68)(cid:70)(cid:75)(cid:68)(cid:81)(cid:74)(cid:72)(cid:76)(cid:81)(cid:83)(cid:75)(cid:68)(cid:86)(cid:72)(cid:82)(cid:73)(cid:17)(cid:53)(cid:72)(cid:86)(cid:88)(cid:79)(cid:87)(cid:86)(cid:86)(cid:75)(cid:82)(cid:90)(cid:87)(cid:75)(cid:68)(cid:87)(cid:87)(cid:75)(cid:72)(cid:40)(cid:48)(cid:89)(cid:82)(cid:85)(cid:87)(cid:72)(cid:91)(cid:90)(cid:68)(cid:89)(cid:72)(cid:75)(cid:68)(cid:86)(cid:68)(cid:75)(cid:72)(cid:79)(cid:76)(cid:70)(cid:68)(cid:79)(cid:83)(cid:75)(cid:68)(cid:86)(cid:72)(cid:73)(cid:85)(cid:82)(cid:81)(cid:87)(cid:15)(cid:87)(cid:75)(cid:72)(cid:81)(cid:88)(cid:80)(cid:69)(cid:72)(cid:85)(cid:82)(cid:73)(cid:69)(cid:85)(cid:68)(cid:81)(cid:70)(cid:75)(cid:72)(cid:86)(cid:72)(cid:84)(cid:88)(cid:68)(cid:79)(cid:86)(cid:15)(cid:68)(cid:81)(cid:71)(cid:87)(cid:75)(cid:72)(cid:74)(cid:72)(cid:81)(cid:72)(cid:85)(cid:68)(cid:87)(cid:72)(cid:71)(cid:50)(cid:36)(cid:48)(cid:80)(cid:82)(cid:71)(cid:72)(cid:81)(cid:88)(cid:80)(cid:69)(cid:72)(cid:85)(cid:86)(cid:68)(cid:87)(cid:76)(cid:86)(cid:191)(cid:72)(cid:86)(cid:87)(cid:75)(cid:68)(cid:87)(cid:62)(cid:24)(cid:64)(cid:17)(cid:41)(cid:82)(cid:85)(cid:68)(cid:83)(cid:68)(cid:76)(cid:85)(cid:82)(cid:73)(cid:50)(cid:36)(cid:48)(cid:80)(cid:82)(cid:71)(cid:72)(cid:81)(cid:88)(cid:80)(cid:69)(cid:72)(cid:85)(cid:86)(cid:15)(cid:87)(cid:75)(cid:72)(cid:83)(cid:82)(cid:86)(cid:76)(cid:87)(cid:76)(cid:89)(cid:72)(cid:68)(cid:81)(cid:71)(cid:81)(cid:72)(cid:74)(cid:68)(cid:87)(cid:76)(cid:89)(cid:72)(cid:89)(cid:68)(cid:79)(cid:88)(cid:72)(cid:86)(cid:70)(cid:82)(cid:85)(cid:85)(cid:72)(cid:86)(cid:83)(cid:82)(cid:81)(cid:71)(cid:87)(cid:82)(cid:79)(cid:72)(cid:73)(cid:87)(cid:16)(cid:75)(cid:68)(cid:81)(cid:71)(cid:68)(cid:81)(cid:71)(cid:85)(cid:76)(cid:74)(cid:75)(cid:87)(cid:16)(cid:75)(cid:68)(cid:81)(cid:71)(cid:89)(cid:82)(cid:85)(cid:87)(cid:72)(cid:91)(cid:80)(cid:82)(cid:71)(cid:72)(cid:86)(cid:15)(cid:85)(cid:72)(cid:86)(cid:83)(cid:72)(cid:70)(cid:87)(cid:76)(cid:89)(cid:72)(cid:79)(cid:92)(cid:17)(cid:55)(cid:75)(cid:72)(cid:73)(cid:85)(cid:72)(cid:84)(cid:88)(cid:72)(cid:81)(cid:70)(cid:92)(cid:76)(cid:81)(cid:192)(cid:88)(cid:72)(cid:81)(cid:70)(cid:72)(cid:82)(cid:81)(cid:87)(cid:75)(cid:72)(cid:83)(cid:75)(cid:68)(cid:86)(cid:72)(cid:71)(cid:76)(cid:86)(cid:87)(cid:85)(cid:76)(cid:69)(cid:88)(cid:87)(cid:76)(cid:82)(cid:81)(cid:76)(cid:86)(cid:86)(cid:75)(cid:82)(cid:90)(cid:81)(cid:76)(cid:81)(cid:41)(cid:76)(cid:74)(cid:17)(cid:22)(cid:17)(cid:44)(cid:87)(cid:76)(cid:86)(cid:82)(cid:69)(cid:89)(cid:76)(cid:82)(cid:88)(cid:86)(cid:87)(cid:75)(cid:68)(cid:87)(cid:87)(cid:75)(cid:72)(cid:85)(cid:68)(cid:71)(cid:76)(cid:88)(cid:86)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:89)(cid:82)(cid:85)(cid:87)(cid:72)(cid:91)(cid:69)(cid:72)(cid:70)(cid:82)(cid:80)(cid:72)(cid:86)(cid:86)(cid:80)(cid:68)(cid:79)(cid:79)(cid:72)(cid:85)(cid:90)(cid:76)(cid:87)(cid:75)(cid:87)(cid:75)(cid:72)(cid:76)(cid:81)(cid:70)(cid:85)(cid:72)(cid:68)(cid:86)(cid:72)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:73)(cid:85)(cid:72)(cid:84)(cid:88)(cid:72)(cid:81)(cid:70)(cid:92)(cid:15)(cid:69)(cid:88)(cid:87)(cid:87)(cid:75)(cid:72)(cid:50)(cid:36)(cid:48)(cid:80)(cid:82)(cid:71)(cid:72)(cid:81)(cid:88)(cid:80)(cid:69)(cid:72)(cid:85)(cid:78)(cid:72)(cid:72)(cid:83)(cid:86)(cid:76)(cid:81)(cid:89)(cid:68)(cid:85)(cid:76)(cid:68)(cid:81)(cid:87)(cid:17)(cid:55)(cid:75)(cid:72)(cid:76)(cid:81)(cid:192)(cid:88)(cid:72)(cid:81)(cid:70)(cid:72)(cid:82)(cid:73)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:85)(cid:68)(cid:71)(cid:76)(cid:88)(cid:86)(cid:82)(cid:81)(cid:87)(cid:75)(cid:72)(cid:83)(cid:75)(cid:68)(cid:86)(cid:72)(cid:71)(cid:76)(cid:86)(cid:87)(cid:85)(cid:76)(cid:69)(cid:88)(cid:87)(cid:76)(cid:82)(cid:81)(cid:76)(cid:86)(cid:86)(cid:75)(cid:82)(cid:90)(cid:81)(cid:76)(cid:81)(cid:41)(cid:76)(cid:74)(cid:17)(cid:23)(cid:17)(cid:55)(cid:75)(cid:72)(cid:69)(cid:72)(cid:68)(cid:80)(cid:86)(cid:68)(cid:85)(cid:72)(cid:86)(cid:68)(cid:80)(cid:83)(cid:79)(cid:72)(cid:71)(cid:68)(cid:87)(cid:68)(cid:71)(cid:76)(cid:86)(cid:87)(cid:68)(cid:81)(cid:70)(cid:72)(cid:73)(cid:85)(cid:82)(cid:80)(cid:87)(cid:75)(cid:72)(cid:87)(cid:85)(cid:68)(cid:81)(cid:86)(cid:80)(cid:76)(cid:87)(cid:87)(cid:76)(cid:81)(cid:74)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:15)(cid:68)(cid:81)(cid:71)(cid:87)(cid:75)(cid:72)(cid:90)(cid:76)(cid:81)(cid:71)(cid:82)(cid:90)(cid:76)(cid:86)(cid:90)(cid:76)(cid:71)(cid:72)(cid:76)(cid:81)(cid:69)(cid:82)(cid:87)(cid:75)(cid:16)(cid:68)(cid:81)(cid:71)(cid:16)(cid:71)(cid:76)(cid:85)(cid:72)(cid:70)(cid:87)(cid:76)(cid:82)(cid:81)(cid:86)(cid:17)(cid:36)(cid:70)(cid:70)(cid:82)(cid:85)(cid:71)(cid:76)(cid:81)(cid:74)(cid:87)(cid:82)(cid:62)(cid:20)(cid:23)(cid:64)(cid:15)(cid:87)(cid:75)(cid:72)(cid:86)(cid:76)(cid:71)(cid:72)(cid:79)(cid:82)(cid:69)(cid:72)(cid:86)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:85)(cid:68)(cid:71)(cid:76)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:83)(cid:68)(cid:87)(cid:87)(cid:72)(cid:85)(cid:81)(cid:76)(cid:81)(cid:70)(cid:85)(cid:72)(cid:68)(cid:86)(cid:72)(cid:68)(cid:86)(cid:87)(cid:75)(cid:72)(cid:85)(cid:68)(cid:71)(cid:76)(cid:88)(cid:86)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:68)(cid:81)(cid:87)(cid:72)(cid:81)(cid:81)(cid:68)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:76)(cid:81)(cid:70)(cid:85)(cid:72)(cid:68)(cid:86)(cid:72)(cid:86)(cid:17)(cid:43)(cid:72)(cid:81)(cid:70)(cid:72)(cid:15)(cid:87)(cid:75)(cid:72)(cid:40)(cid:48)(cid:89)(cid:82)(cid:85)(cid:87)(cid:72)(cid:91)(cid:90)(cid:68)(cid:89)(cid:72)(cid:76)(cid:86)(cid:81)(cid:82)(cid:87)(cid:76)(cid:81)(cid:68)(cid:83)(cid:88)(cid:85)(cid:72)(cid:50)(cid:36)(cid:48)(cid:86)(cid:87)(cid:68)(cid:87)(cid:72)(cid:15)(cid:69)(cid:88)(cid:87)(cid:68)(cid:86)(cid:88)(cid:83)(cid:72)(cid:85)(cid:83)(cid:82)(cid:86)(cid:76)(cid:87)(cid:76)(cid:82)(cid:81)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:80)(cid:68)(cid:76)(cid:81)(cid:79)(cid:82)(cid:69)(cid:72)(cid:68)(cid:81)(cid:71)(cid:86)(cid:76)(cid:71)(cid:72)(cid:79)(cid:82)(cid:69)(cid:72)(cid:86)(cid:41)(cid:76)(cid:74)(cid:17)(cid:24)(cid:17)(cid:53)(cid:68)(cid:71)(cid:68)(cid:85)(cid:76)(cid:80)(cid:68)(cid:74)(cid:76)(cid:81)(cid:74)(cid:70)(cid:82)(cid:82)(cid:85)(cid:71)(cid:76)(cid:81)(cid:68)(cid:87)(cid:72)(cid:73)(cid:85)(cid:68)(cid:80)(cid:72)(cid:17)(cid:90)(cid:76)(cid:87)(cid:75)(cid:71)(cid:76)(cid:73)(cid:73)(cid:72)(cid:85)(cid:72)(cid:81)(cid:87)(cid:50)(cid:36)(cid:48)(cid:86)(cid:87)(cid:68)(cid:87)(cid:72)(cid:86)(cid:17)(cid:36)(cid:79)(cid:86)(cid:82)(cid:15)(cid:87)(cid:75)(cid:72)(cid:86)(cid:88)(cid:83)(cid:72)(cid:85)(cid:83)(cid:82)(cid:86)(cid:76)(cid:87)(cid:76)(cid:82)(cid:81)(cid:83)(cid:75)(cid:72)(cid:81)(cid:82)(cid:80)(cid:16)(cid:72)(cid:81)(cid:82)(cid:81)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:83)(cid:75)(cid:68)(cid:86)(cid:72)(cid:71)(cid:76)(cid:86)(cid:87)(cid:85)(cid:76)(cid:69)(cid:88)(cid:87)(cid:76)(cid:82)(cid:81)(cid:82)(cid:70)(cid:70)(cid:88)(cid:85)(cid:86)(cid:17)(cid:55)(cid:75)(cid:72)(cid:68)(cid:81)(cid:68)(cid:79)(cid:92)(cid:86)(cid:72)(cid:86)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:83)(cid:75)(cid:68)(cid:86)(cid:72)(cid:71)(cid:76)(cid:86)(cid:87)(cid:85)(cid:76)(cid:69)(cid:88)(cid:87)(cid:76)(cid:82)(cid:81)(cid:68)(cid:85)(cid:72)(cid:69)(cid:72)(cid:81)(cid:72)(cid:191)(cid:70)(cid:76)(cid:68)(cid:79)(cid:87)(cid:82)(cid:87)(cid:75)(cid:72)(cid:86)(cid:76)(cid:74)(cid:81)(cid:68)(cid:79)(cid:71)(cid:72)(cid:86)(cid:76)(cid:74)(cid:81)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:82)(cid:85)(cid:69)(cid:76)(cid:87)(cid:68)(cid:79)(cid:16)(cid:68)(cid:81)(cid:74)(cid:88)(cid:79)(cid:68)(cid:85)(cid:16)(cid:80)(cid:82)(cid:80)(cid:72)(cid:81)(cid:87)(cid:88)(cid:80)(cid:16)(cid:69)(cid:68)(cid:86)(cid:72)(cid:71)(cid:85)(cid:68)(cid:71)(cid:68)(cid:85)(cid:68)(cid:81)(cid:71)(cid:87)(cid:75)(cid:72)(cid:85)(cid:72)(cid:86)(cid:72)(cid:68)(cid:85)(cid:70)(cid:75)(cid:82)(cid:81)(cid:87)(cid:75)(cid:72)(cid:76)(cid:81)(cid:87)(cid:72)(cid:85)(cid:68)(cid:70)(cid:87)(cid:76)(cid:82)(cid:81)(cid:80)(cid:72)(cid:70)(cid:75)(cid:68)(cid:81)(cid:76)(cid:86)(cid:80)(cid:82)(cid:73)(cid:85)(cid:68)(cid:71)(cid:68)(cid:85)(cid:68)(cid:81)(cid:71)(cid:87)(cid:75)(cid:72)(cid:87)(cid:68)(cid:85)(cid:74)(cid:72)(cid:87)(cid:17)(cid:44)(cid:44)(cid:44)(cid:17)(cid:50)(cid:53)(cid:37)(cid:44)(cid:55)(cid:36)(cid:47)(cid:16)(cid:36)(cid:49)(cid:42)(cid:56)(cid:47)(cid:36)(cid:53)(cid:16)(cid:48)(cid:50)(cid:48)(cid:40)(cid:49)(cid:55)(cid:56)(cid:48)(cid:16)(cid:37)(cid:36)(cid:54)(cid:40)(cid:39)(cid:53)(cid:36)(cid:39)(cid:36)(cid:53)(cid:44)(cid:48)(cid:36)(cid:42)(cid:44)(cid:49)(cid:42)(cid:44)(cid:81)(cid:87)(cid:75)(cid:76)(cid:86)(cid:86)(cid:72)(cid:70)(cid:87)(cid:76)(cid:82)(cid:81)(cid:15)(cid:87)(cid:75)(cid:72)(cid:76)(cid:80)(cid:68)(cid:74)(cid:76)(cid:81)(cid:74)(cid:80)(cid:72)(cid:87)(cid:75)(cid:82)(cid:71)(cid:86)(cid:82)(cid:73)(cid:87)(cid:90)(cid:82)(cid:80)(cid:82)(cid:71)(cid:72)(cid:86)(cid:15)(cid:85)(cid:72)(cid:73)(cid:72)(cid:85)(cid:85)(cid:72)(cid:71)(cid:87)(cid:82)(cid:68)(cid:86)(cid:80)(cid:88)(cid:79)(cid:87)(cid:76)(cid:83)(cid:79)(cid:72)(cid:16)(cid:76)(cid:81)(cid:177)(cid:80)(cid:88)(cid:79)(cid:87)(cid:76)(cid:83)(cid:79)(cid:72)(cid:16)(cid:82)(cid:88)(cid:87)(cid:68)(cid:81)(cid:71)(cid:80)(cid:88)(cid:79)(cid:87)(cid:76)(cid:83)(cid:79)(cid:72)(cid:16)(cid:76)(cid:81)(cid:177)(cid:86)(cid:76)(cid:81)(cid:74)(cid:79)(cid:72)(cid:16)(cid:82)(cid:88)(cid:87)(cid:15)(cid:68)(cid:85)(cid:72)(cid:83)(cid:85)(cid:82)(cid:89)(cid:76)(cid:71)(cid:72)(cid:71)(cid:17)(cid:41)(cid:76)(cid:74)(cid:17)(cid:24)(cid:86)(cid:75)(cid:82)(cid:90)(cid:86)(cid:87)(cid:75)(cid:72)(cid:85)(cid:68)(cid:71)(cid:68)(cid:85)(cid:76)(cid:80)(cid:68)(cid:74)(cid:76)(cid:81)(cid:74)(cid:70)(cid:82)(cid:82)(cid:85)(cid:71)(cid:76)(cid:81)(cid:68)(cid:87)(cid:72)(cid:73)(cid:85)(cid:68)(cid:80)(cid:72)(cid:17)(cid:55)(cid:75)(cid:72)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:70)(cid:82)(cid:81)(cid:191)(cid:74)(cid:88)(cid:85)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:86)(cid:68)(cid:80)(cid:72)(cid:68)(cid:86)(cid:87)(cid:75)(cid:68)(cid:87)(cid:76)(cid:81)(cid:41)(cid:76)(cid:74)(cid:17)(cid:20)(cid:17)(cid:44)(cid:81)(cid:87)(cid:75)(cid:72)(cid:80)(cid:88)(cid:79)(cid:87)(cid:76)(cid:83)(cid:79)(cid:72)(cid:16)(cid:76)(cid:81)(cid:177)(cid:80)(cid:88)(cid:79)(cid:87)(cid:76)(cid:83)(cid:79)(cid:72)(cid:16)(cid:82)(cid:88)(cid:87)(cid:80)(cid:82)(cid:71)(cid:72)(cid:15)(cid:68)(cid:79)(cid:79)(cid:68)(cid:81)(cid:87)(cid:72)(cid:81)(cid:81)(cid:68)(cid:86)(cid:76)(cid:81)(cid:87)(cid:75)(cid:72)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:68)(cid:85)(cid:72)(cid:88)(cid:86)(cid:72)(cid:71)(cid:87)(cid:82)(cid:85)(cid:72)(cid:70)(cid:72)(cid:76)(cid:89)(cid:72)(cid:87)(cid:75)(cid:72)(cid:72)(cid:70)(cid:75)(cid:82)(cid:86)(cid:76)(cid:74)(cid:81)(cid:68)(cid:79)(cid:15)(cid:68)(cid:81)(cid:71)(cid:87)(cid:75)(cid:72)(cid:85)(cid:72)(cid:70)(cid:72)(cid:76)(cid:89)(cid:76)(cid:81)(cid:74)(cid:80)(cid:82)(cid:71)(cid:72)(cid:76)(cid:86)(cid:76)(cid:81)(cid:68)(cid:70)(cid:70)(cid:82)(cid:85)(cid:71)(cid:90)(cid:76)(cid:87)(cid:75)(cid:87)(cid:75)(cid:68)(cid:87)(cid:76)(cid:81)(cid:87)(cid:75)(cid:72)(cid:72)(cid:80)(cid:76)(cid:86)(cid:86)(cid:76)(cid:82)(cid:81)(cid:83)(cid:85)(cid:82)(cid:70)(cid:72)(cid:86)(cid:86)(cid:17)(cid:55)(cid:75)(cid:72)(cid:86)(cid:76)(cid:74)(cid:81)(cid:68)(cid:79)(cid:85)(cid:72)(cid:70)(cid:72)(cid:76)(cid:89)(cid:72)(cid:71)(cid:69)(cid:92)(cid:87)(cid:75)(cid:72)(cid:87)(cid:75)(cid:72)(cid:79)(cid:72)(cid:80)(cid:72)(cid:81)(cid:87)(cid:76)(cid:86)(cid:80)(cid:88)(cid:79)(cid:87)(cid:76)(cid:83)(cid:79)(cid:76)(cid:72)(cid:71)(cid:69)(cid:92)(cid:68)(cid:83)(cid:75)(cid:68)(cid:86)(cid:72)(cid:15)(cid:75)(cid:72)(cid:81)(cid:70)(cid:72)(cid:87)(cid:75)(cid:72)(cid:87)(cid:82)(cid:87)(cid:68)(cid:79)(cid:72)(cid:70)(cid:75)(cid:82)(cid:68)(cid:73)(cid:87)(cid:72)(cid:85)(cid:86)(cid:88)(cid:80)(cid:80)(cid:76)(cid:81)(cid:74)(cid:82)(cid:89)(cid:72)(cid:85)(cid:87)(cid:75)(cid:72)(cid:86)(cid:76)(cid:74)(cid:81)(cid:68)(cid:79)(cid:86)(cid:85)(cid:72)(cid:70)(cid:72)(cid:76)(cid:89)(cid:72)(cid:71)(cid:69)(cid:92)(cid:68)(cid:79)(cid:79)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:68)(cid:81)(cid:87)(cid:72)(cid:81)(cid:81)(cid:68)(cid:86)(cid:76)(cid:86)(cid:11)(cid:21)(cid:12)(cid:90)(cid:75)(cid:72)(cid:85)(cid:72)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:85)(cid:68)(cid:71)(cid:68)(cid:85)(cid:70)(cid:85)(cid:82)(cid:86)(cid:86)(cid:86)(cid:72)(cid:70)(cid:87)(cid:76)(cid:82)(cid:81)(cid:11)(cid:53)(cid:38)(cid:54)(cid:12)(cid:82)(cid:73)(cid:87)(cid:75)(cid:72)(cid:87)(cid:68)(cid:85)(cid:74)(cid:72)(cid:87)(cid:17)(cid:36)(cid:86)(cid:86)(cid:88)(cid:80)(cid:72)(cid:87)(cid:75)(cid:68)(cid:87)(cid:87)(cid:75)(cid:72)(cid:87)(cid:68)(cid:85)(cid:74)(cid:72)(cid:87)(cid:76)(cid:86)(cid:80)(cid:68)(cid:71)(cid:72)(cid:88)(cid:83)(cid:82)(cid:73)(cid:80)(cid:88)(cid:79)(cid:87)(cid:76)(cid:83)(cid:79)(cid:72)(cid:76)(cid:71)(cid:72)(cid:68)(cid:79)(cid:86)(cid:70)(cid:68)(cid:87)(cid:87)(cid:72)(cid:85)(cid:76)(cid:81)(cid:74)(cid:83)(cid:82)(cid:76)(cid:81)(cid:87)(cid:86)(cid:15)(cid:71)(cid:72)(cid:81)(cid:82)(cid:87)(cid:72)(cid:71)(cid:69)(cid:92)(cid:68)(cid:81)(cid:71)(cid:87)(cid:75)(cid:72)(cid:70)(cid:82)(cid:85)(cid:85)(cid:72)(cid:86)(cid:83)(cid:82)(cid:81)(cid:71)(cid:76)(cid:81)(cid:74)(cid:53)(cid:38)(cid:54)(cid:17)(cid:55)(cid:75)(cid:72)(cid:72)(cid:70)(cid:75)(cid:82)(cid:76)(cid:86)(cid:87)(cid:75)(cid:72)(cid:81)(cid:72)(cid:91)(cid:83)(cid:85)(cid:72)(cid:86)(cid:86)(cid:72)(cid:71)(cid:68)(cid:86)(cid:11)(cid:22)(cid:12)(cid:58)(cid:75)(cid:72)(cid:81)(cid:15)(cid:87)(cid:75)(cid:72)(cid:37)(cid:72)(cid:86)(cid:86)(cid:72)(cid:79)(cid:73)(cid:88)(cid:81)(cid:70)(cid:87)(cid:76)(cid:82)(cid:81)(cid:70)(cid:68)(cid:81)(cid:69)(cid:72)(cid:68)(cid:83)(cid:83)(cid:85)(cid:82)(cid:91)(cid:76)(cid:80)(cid:68)(cid:87)(cid:72)(cid:71)(cid:68)(cid:86)(cid:73)(cid:82)(cid:79)(cid:79)(cid:82)(cid:90)(cid:86)(cid:29)(cid:11)(cid:23)(cid:12)(cid:55)(cid:75)(cid:72)(cid:85)(cid:72)(cid:73)(cid:82)(cid:85)(cid:72)(cid:15)(cid:87)(cid:75)(cid:72)(cid:81)(cid:82)(cid:85)(cid:80)(cid:68)(cid:79)(cid:76)(cid:93)(cid:72)(cid:71)(cid:72)(cid:70)(cid:75)(cid:82)(cid:86)(cid:76)(cid:74)(cid:81)(cid:68)(cid:79)(cid:70)(cid:68)(cid:81)(cid:69)(cid:72)(cid:74)(cid:76)(cid:89)(cid:72)(cid:81)(cid:69)(cid:92)(cid:11)(cid:24)(cid:12) UCAA-based echo models and multiple signal classification
(MUSIC) algorithm [188], [189], estimating signal parameter
via rotational invariance techniques (ESPRIT) algorithm [190]
and least squares algorithms [191], [192].
There are also many challenges in OAM-based radar imag-
ing. Because the center of vortex electromagnetic beams is
null and the angular direction of the main lobe varies with the
OAM modes, it is difficult to simultaneously illuminate the
target with vortex electromagnetic beams of different modes,
which results in limited echo energy. The difference of echo
intensity will lead to amplitude modulation of the echo signal,
resulting in a deterioration of the imaging capability. These
effects can be mitigated by carefully designing the radius
and excitation of the UCAA to adjust
the main lobes of
the vortex electromagnetic waves with different modes [183],
[187], [191]. In addition, detecting dynamic targets is also
challenging, which requires more in-depth research.
Optical OAM beams for imaging is generated using SPPs
and SLMs, while radio OAM beams for radar imaging are
generated using UCAAs. OAM beam divergence and beam
steering must be considered in order to make OAM waves
with multiple modes illuminate the same target. More in-depth
research on topics such as high-precision azimuth imaging and
dynamic target detection are the object of future research.
VII. CONCLUSIONS AND PERSPECTIVES
In recent years, OAM has been extensively studied in many
fields of application, especially in the field of communications.
A large number of studies have shown that it is possible
to multiplex a set of orthogonal OAM modes on the same
frequency channel so to achieve a high spectral efficiency.
In this paper, we provide a comprehensive overview of the
generation and detection methods of optical, radio, and acous-
tic OAM, together with their applications in communications,
particle manipulation and imaging. Particular attention has
been devoted to the issues that still obstruct or limit
the
application of OAM in practice.
OAM technology is on the brink of been deployed in many
communication systems but there are still some problems that
need to be solved. For example, many of the experiments
related to OAM multiplexing still use bulky and expensive
components, which are not suitable for practical implemen-
tation and large-scale use. How to generate and detect high-
order and multi-mode OAM beams with a finite-size aperture
is challenging. Future success of OAM heavily relies on
the development of system components such as transmitters,
multiplexers, demultiplexers and receivers. These components
are required to provide reductions in cost and size and to be
compatible with existing technologies. Moreover, the state of
the propagation channel has a large impact on OAM-based
links and the problems arising in some complex and severe
channel conditions need to be addressed and solved. In spite
of the many challenges in applying OAM to communication
systems, on the base of the research progress surveyed in
this paper, we remain optimistic that these challenges will be
gradually resolved in the future and that OAM will help to
make breakthroughs in future communications and radar target
detection systems.
25
REFERENCES
[1] L. Allen, M. W. Beijersbergen, R. J. Spreeuw, and J. P. Woerdman, "Or-
bital angular momentum of light and the transformation of Laguerre-
Gaussian laser modes," Phys. Rev. A: At. Mol. Opt. Phys., vol. 45,
no. 11, pp. 8185 -- 8189, 1992.
[2] B. Thid´e, H. Then, J. Sjoholm, K. Palmer, J. Bergman, T. D. Carozzi,
Y. N. Istomin, N. H. Ibragimov, and R. Khamitova, "Utilization of
photon orbital angular momentum in the low-frequency radio domain,"
Phys. Rev. Lett., vol. 99, no. 8, p. 087701, 2007.
[3] E. G. Broadbent and D. W. Moore, "Acoustic destabilization of
vortices," Philos. Trans. R. Soc. London, vol. 290, no. 1372, pp. 353 --
371, 1979.
[4] B. T. Hefner and P. L. Marston, "An acoustical helicoidal wave trans-
ducer with applications for the alignment of ultrasonic and underwater
systems," J. Acoust. Soc. Am., vol. 106, no. 6, pp. 3313 -- 3316, 1999.
[5] A. M. Yao and M. J. Padgett, "Orbital angular momentum: origins,
behavior and applications," Adv. Opt. Photonics, vol. 3, no. 2, pp. 161 --
204, 2011.
[6] A. E. Willner, Y. Ren, G. Xie, Y. Yan, L. Li, Z. Zhao, J. Wang,
M. Tur, A. F. Molisch, and S. Ashrafi, "Recent advances in high-
capacity free-space optical and radio-frequency communications using
orbital angular momentum multiplexing," Phil. Trans. R. Soc. A, vol.
375, no. 2087, p. 20150439, 2017.
[7] A. E. Willner, H. Huang, Y. Yan, Y. Ren, N. Ahmed, G. Xie,
C. Bao, L. Li, Y. Cao, Z. Zhao, J. Wang, M. P. J. Lavery, M. Tur,
S. Ramachandran, A. F. Molisch, N. Ashrafi, and S. Ashrafi, "Optical
communications using orbital angular momentum beams," Adv. Opt.
Photonics, vol. 7, no. 1, pp. 66 -- 106, 2015.
[8] S. M. Mohammadi, L. K. S. Daldorff, J. E. S. Bergman, R. L. Karlsson,
B. Thid´e, K. Forozesh, T. D. Carozzi, and B. Isham, "Orbital angular
momentum in radio -- a system study," IEEE Trans. Antennas Propag.,
vol. 58, no. 2, pp. 565 -- 572, 2010.
[9] J. M. Vaughan and D. V. Willetts, "Temporal and interference fringe
analysis of TEM01* laser modes," J. Opt. Soc. Am., vol. 73, no. 8, pp.
1018 -- 1021, 1983.
[10] K. Kano, Y. Kozawa, and S. Sato, "Generation of a purely single
transverse mode vortex beam from a He-Ne laser cavity with a spot-
defect mirror," Int. J. Opt., vol. 2012, pp. 1 -- 6, 2012.
[11] M. W. Beijersbergen, L. Allen, H. van der Veen, and J. P. Woerdman,
"Astigmatic laser mode converters and transfer of orbital angular
momentum," Opt. Commun., vol. 96, no. 1, pp. 123 -- 132, 1993.
[12] M. W. Beijersbergen, R. P. C. Coerwinkel, M. Kristensen, and J. P.
Woerdman, "Helical-wavefront laser beams produced with a spiral
phaseplate," Opt. Commun., vol. 112, no. 5, pp. 321 -- 327, 1994.
[13] V. Y. Bazhenov, M. V. Vasnetsov, and M. S. Soskin, "Laser beams with
screw dislocations in their wavefronts," JETP Lett., vol. 52, no. 8, pp.
429 -- 431, 1990.
[14] V. Y. Bazhenov, M. S. Soskin, and M. V. Vasnetsov, "Screw dislocations
in light wavefronts," J. Mod. Opt., vol. 39, no. 5, pp. 985 -- 990, 1992.
[15] N. R. Heckenberg, R. Mcduff, C. P. Smith, and A. G. White, "Gener-
ation of optical phase singularities by computer-generated holograms,"
Opt. Lett., vol. 17, no. 3, pp. 221 -- 223, 1992.
[16] J. E. Curtis, B. A. Koss, and D. G. Grier, "Dynamic holographic optical
tweezers," Opt. Commun., vol. 207, no. 1, pp. 169 -- 175, 2002.
[17] C. Maurer, A. Jesacher, S. Furhapter, S. Bernet, and M. Ritsch-Marte,
"Tailoring of arbitrary optical vector beams," New J. Phys., vol. 9,
no. 3, p. 78, 2007.
[18] M. Mirhosseini, O. S. Magana Loaiza, C. Chen, B. Rodenburg,
M. Malik, and R. W. Boyd, "Rapid generation of light beams carrying
orbital angular momentum," Opt. Express, vol. 21, no. 25, pp. 30 196 --
30 203, 2013.
[19] K. J. Mitchell, S. Turtaev, M. J. Padgett, T. Cizm´ar, and D. B. Phillips,
"High-speed spatial control of the intensity, phase and polarisation
of vector beams using a digital micro-mirror device," Opt. Express,
vol. 24, no. 25, pp. 29 269 -- 29 282, 2016.
[20] N. Yu, P. Genevet, M. A. Kats, F. Aieta, J. P. Tetienne, F. Capasso, and
Z. Gaburro, "Light propagation with phase discontinuities: generalized
laws of reflection and refraction," Science, vol. 334, no. 6054, pp. 333 --
337, 2011.
[21] P. Genevet, N. Yu, F. Aieta, J. Lin, M. A. Kats, R. Blanchard, M. O.
Scully, Z. Gaburro, and F. Capasso, "Ultra-thin plasmonic optical
vortex plate based on phase discontinuities," Appl. Phys. Lett., vol.
100, no. 1, p. 013101, 2012.
[22] E. Karimi, S. A. Schulz, I. D. Leon, H. Qassim, J. Upham, and
R. W. Boyd, "Generating optical orbital angular momentum at visible
26
wavelengths using a plasmonic metasurface," Light: Sci. Appl., vol. 3,
p. e167, 2014.
[23] Z. Zhao, J. Wang, S. Li, and A. E. Willner, "Metamaterials-based
broadband generation of orbital angular momentum carrying vector
beams," Opt. Lett., vol. 38, no. 6, pp. 932 -- 934, 2013.
[24] W. Wang, Y. Li, Z. Guo, R. Li, J. Zhang, A. Zhang, and S. Qu, "Ultra-
thin optical vortex phase plate based on the metasurface and the angular
momentum transformation," J. Opt., vol. 17, no. 4, p. 045102, 2015.
[25] L. Marrucci, E. Karimi, S. Slussarenko, B. Piccirillo, E. Santamato,
E. Nagali, and F. Sciarrino, "Spin-to-orbital conversion of the angular
momentum of light and its classical and quantum applications," J. Opt.,
vol. 13, no. 6, p. 064001, 2011.
[26] L. Marrucci, C. Manzo, and D. Paparo, "Optical spin-to-orbital angular
momentum conversion in inhomogeneous anisotropic media," Phys.
Rev. Lett., vol. 96, no. 16, p. 163905, 2006.
[27] Z. Bomzon, G. Biener, V. Kleiner, and E. Hasman, "Space-variant
Pancharatnam-Berry phase optical elements with computer-generated
subwavelength gratings," Opt. Lett., vol. 27, no. 13, pp. 1141 -- 1143,
2002.
[28] G. Biener, A. Niv, V. Kleiner, and E. Hasman, "Formation of helical
beams by use of Pancharatnam-Berry phase optical elements," Opt.
Lett., vol. 27, no. 21, pp. 1875 -- 1877, 2002.
[29] L. Zhu, X. Wei, J. Wang, Z. Zhang, Z. Li, H. Zhang, S. Li, K. Wang,
and J. Liu, "Experimental demonstration of basic functionalities for
0.1-THz orbital angular momentum (OAM) communications," in Proc.
Opt. Fiber Commun. Conf. Expo., 2014, p. M3K.7.
[30] A. Bennis, R. Niemiec, C. Brousseau, K. Mahdjoubi, and O. Emile,
"Flat plate for OAM generation in the millimeter band," in Proc. 7th
Eur. Conf. Antennas Propag., 2013, pp. 3203 -- 3207.
[31] L. Cheng, W. Hong, and Z.-C. Hao, "Generation of electromagnetic
waves with arbitrary orbital angular momentum modes," Sci. Rep.,
vol. 4, p. 4814, 2014.
[32] F. E. Mahmouli and S. Walker, "Orbital angular momentum generation
in a 60GHz wireless radio channel," in Proc. 20th Telecommun. Forum,
2012, pp. 315 -- 318.
[33] F. Tamburini, E. Mari, B. Thid´e, C. Barbieri, and F. Romanato,
"Experimental verification of photon angular momentum and vorticity
with radio techniques," Appl. Phys. Lett., vol. 99, no. 20, p. 204102,
2011.
[34] F. Tamburini, E. Mari, A. Sponselli, B. Thid´e, A. Bianchini, and
F. Romanato, "Encoding many channels in the same frequency through
radio vorticity: first experimental test," New J. Phys., vol. 14, no. 3, p.
033001, 2012.
[35] G. A. Turnbull, D. A. Robertson, G. M. Smith, L. Allen, and M. J.
Padgett, "The generation of free-space Laguerre-Gaussian modes at
millimetre-wave frequencies by use of a spiral phaseplate," Opt.
Commun., vol. 127, no. 4, pp. 183 -- 188, 1996.
[36] Y. Yan, G. Xie, H. Huang, M. J. Lavery, N. Ahemd, C. Bao, Y. Ren,
A. F. Molisch, M. Tur, M. Padgett, and A. E. Willner, "Demonstration
of 8-mode 32-Gbit/s millimeter-wave free-space communication link
using 4 orbital-angular-momentum modes on 2 polarizations," in Proc.
IEEE Int. Conf. Commun., 2014, pp. 4850 -- 4855.
[37] C. Deng, W. Chen, Z. Zhang, Y. Li, and Z. Feng, "Generation of
OAM radio waves using circular Vivaldi antenna array," Int. J. Antennas
Propag., vol. 2013, no. 2, pp. 607 -- 610, 2013.
[38] X. Bai, X. Liang, R. Jin, and J. Geng, "Generation of OAM radio
waves with three polarizations using circular horn antenna array," in
Proc. 9th Eur. Conf. Antennas Propag., 2015.
[39] R. Gaffoglio, A. Cagliero, A. D. Vita, and B. Sacco, "OAM multiple
transmission using uniform circular arrays: Numerical modeling and
experimental verification with two digital television signals," Radio
Sci., vol. 51, no. 6, pp. 645 -- 658, 2016.
[40] Q. Bai, A. Tennant, B. Allen, and M. U. Rehman, "Generation of orbital
angular momentum (OAM) radio beams with phased patch array," in
Proc. Loughborough Antennas Propag. Conf., 2013, pp. 410 -- 413.
[41] X. Sun, Y. Du, Y. Fan, and M. Sun, "The design of array antenna
based on multi-modal OAM vortex electromagnetic wave," in Proc.
Prog. Electromagn. Res. Symp., 2016, pp. 2786 -- 2791.
[42] M. Lin, Y. Gao, P. Liu, and J. Liu, "Theoretical analyses and design
of circular array to generate orbital angular momentum," IEEE Trans.
Antennas Propag., vol. 65, no. 7, pp. 3510 -- 3519, 2017.
[43] A. Tennant and B. Allen, "Generation of radio frequency OAM radi-
ation modes using circular time-switched and phased array antennas,"
in Proc. Loughborough Antennas Propag. Conf., 2012.
[44] S. Yu, L. Li, G. Shi, C. Zhu, X. Zhou, and Y. Shi, "Design, fabrication,
and measurement of reflective metasurface for orbital angular momen-
tum vortex wave in radio frequency domain," Appl. Phys. Lett., vol.
108, no. 12, p. 121903, 2016.
[45] M. L. N. Chen, L. J. Jiang, and W. E. I. Sha, "Artificial perfect
electric conductor-perfect magnetic conductor anisotropic metasurface
for generating orbital angular momentum of microwave with nearly
perfect conversion efficiency," J. Appl. Phys., vol. 119, no. 6, p. 064506,
2016.
[46] S. Yu, L. Li, G. Shi, C. Zhu, and Y. Shi, "Generating multiple
orbital angular momentum vortex beams using a metasurface in radio
frequency domain," Appl. Phys. Lett., vol. 108, no. 24, p. 241901, 2016.
[47] S. Yu, L. Li, and G. Shi, "Dual-polarization and dual-mode orbital
angular momentum radio vortex beam generated by using reflective
metasurface," Appl. Phys. Express, vol. 9, no. 8, p. 082202, 2016.
[48] Y. Zhang, L. Yang, H. Wang, X. Zhang, and X. Jin, "Transforming
surface wave to propagating OAM vortex wave via flat dispersive
metasurface in radio frequency," IEEE Antennas Wireless Propag. Lett.,
vol. 17, no. 1, pp. 172 -- 175, 2018.
[49] N. Kou, S. Yu, and L. Li, "Generation of high-order Bessel vortex beam
carrying orbital angular momentum using multilayer amplitude-phase-
modulated surfaces in radiofrequency domain," Appl. Phys. Express,
vol. 10, no. 1, p. 016701, 2016.
[50] S. Maccalli, G. Pisano, S. Colafrancesco, B. Maffei, M. W. Ng, and
M. Gray, "Q-plate for millimeter-wave orbital angular momentum
manipulation," Appl. Opt., vol. 52, no. 4, pp. 635 -- 639, 2013.
[51] R. Wunenburger, J. I. V. Lozano, and E. Brasselet, "Acoustic orbital
angular momentum transfer to matter by chiral scattering," New J.
Phys., vol. 17, no. 10, p. 103022, 2015.
[52] S. Gspan, A. Meyer, S. Bernet, and M. Ritsch-Marte, "Optoacoustic
generation of a helicoidal ultrasonic beam," J. Acoust. Soc. Am., vol.
115, no. 3, pp. 1142 -- 1146, 2004.
[53] B. T. Hefner and P. L. Marston, "Acoustical helicoidal waves and
Laguerre-Gaussian beams: Applications to scattering and to angular
momentum transport," J. Acoust. Soc. Am., vol. 103, no. 5, pp. 2971 --
2971, 1998.
[54] J. L. Ealo, J. C. Prieto, and F. Seco, "Airborne ultrasonic vortex gen-
eration using flexible ferroelectrets," IEEE T. Ultrason. Ferr., vol. 58,
no. 8, pp. 1651 -- 1657, 2011.
[55] R. Marchiano and J. L. Thomas, "Synthesis and analysis of linear and
nonlinear acoustical vortices," Phys. Rev. E, vol. 71, no. 6, p. 066616,
2005.
[56] K. Volke-Sep´ulveda, A. O. Santill´an, and R. R. Boullosa, "Transfer of
angular momentum to matter from acoustical vortices in free space,"
Phys. Rev. Lett., vol. 100, no. 2, p. 024302, 2008.
[57] L. Yang, Q. Ma, J. Tu, and D. Zhang, "Phase-coded approach for
controllable generation of acoustical vortices," J. Appl. Phys., vol. 113,
no. 15, p. 154904, 2013.
[58] Y. Li, G. Guo, Q. Ma, J. Tu, and D. Zhang, "Deep-level stereoscopic
multiple traps of acoustic vortices," J. Appl. Phys., vol. 121, no. 16, p.
164901, 2017.
[59] X. Jiang, J. Zhao, S. Liu, B. Liang, X. Zou, J. Yang, C. Qiu, and
J. Cheng, "Broadband and stable acoustic vortex emitter with multi-
arm coiling slits," Appl. Phys. Lett., vol. 108, no. 20, p. 203501, 2016.
[60] T. Wang, M. Ke, W. Li, Q. Yang, C. Qiu, and Z. Liu, "Particle
manipulation with acoustic vortex beam induced by a brass plate with
spiral shape structure," Appl. Phys. Lett., vol. 109, no. 12, p. 123506,
2016.
[61] N. Jim´enez, R. Pic´o, V. S´anchez-Morcillo, V. Romero-Garc´ıa, L. M.
Garc´ıa-Raffi, and K. Staliunas, "Formation of high-order acoustic
Bessel beams by spiral diffraction gratings," Phys. Rev. E, vol. 94,
no. 5, p. 053004, 2016.
[62] N. Jim´enez, V. Romero-Garc´ıa, L. M. Garc´ıa-Raffi, F. Camarena, and
K. Staliunas, "Sharp acoustic vortex focusing by Fresnel-spiral zone
plates," Appl. Phys. Lett., vol. 112, no. 20, p. 204101, 2018.
[63] R. D. Muelas-Hurtado, J. L. Ealo, J. F. Pazos-Ospina, and K. Volke-
Sep´ulveda, "Generation of multiple vortex beam by means of active
diffraction gratings," Appl. Phys. Lett., vol. 112, no. 8, p. 084101, 2018.
[64] J. Xue, L. Yong, L. Bin, C. Jianchun, and Z. Likun, "Convert acoustic
resonances to orbital angular momentum," Phys. Rev. Lett., vol. 117,
no. 3, p. 034301, 2016.
[65] L. Ye, C. Qiu, J. Lu, K. Tang, H. Jia, M. Ke, S. Peng, and Z. Liu,
"Making sound vortices by metasurfaces," AIP Advances, vol. 6, no. 8,
p. 085007, 2016.
[66] H. Esfahlani, H. Lissek, and J. R. Mosig, "Generation of acoustic
helical wavefronts using metasurfaces," Phys. Rev. B, vol. 95, no. 2, p.
024312, 2017.
[67] J. Leach, M. J. Padgett, S. M. Barnett, S. Franke-Arnold, and J. Cour-
tial, "Measuring the orbital angular momentum of a single photon,"
Phys. Rev. Lett., vol. 88, no. 25, p. 257901, 2002.
[68] A. Mair, A. Vaziri, G. Weihs, and A. Zeilinger, "Entanglement of the
orbital angular momentum states of photons," Nature, vol. 412, no.
6844, pp. 313 -- 316, 2001.
[69] G. Gibson, J. Courtial, M. Vasnetsov, M. J. Padgett, S. Franke-Arnold,
S. M. Barnett, and V. PasKo, "Free-space information transfer using
light beams carrying orbital angular momentum," Opt. Express, vol. 12,
no. 22, pp. 5448 -- 5456, 2004.
[70] G. Gibson, J. Courtial, M. Vasnetsov, S. Barnett, S. Franke-Arnold,
and M. Padgett, "Increasing the data density of free-space optical
communications using orbital angular momentum," in Proc. SPIE,
2004, pp. 367 -- 373.
[71] N. Zhang, X. C. Yuan, and R. E. Burge, "Extending the detection range
of optical vortices by dammann vortex gratings," Opt. Lett., vol. 35,
no. 20, pp. 3495 -- 3497, 2010.
[72] C. Gao, S. Zhang, S. Fu, and T. Wang, "Integrating 5×5 dammann
gratings to detect orbital angular momentum states of beams with the
range of −24 to +24," Appl. Opt., vol. 55, no. 7, pp. 1514 -- 1517, 2016.
[73] C. Kai, P. Huang, F. Shen, H. Zhou, and Z. Guo, "Orbital angular
momentum shift keying based optical communication system," IEEE
Photonics J., vol. 9, no. 2, pp. 1 -- 10, 2017.
[74] J. Lin, X. Yuan, S. H. Tao, and R. E. Burge, "Synthesis of multiple
collinear helical modes generated by a phase-only element," J. Opt.
Soc. Am. A, vol. 23, no. 5, pp. 1214 -- 1218, 2006.
[75] H. I. Sztul and R. R. Alfano, "Double-slit interference with Laguerre-
Gaussian beams," Opt. Lett., vol. 31, no. 7, pp. 999 -- 1001, 2006.
[76] H. Zhou, L. Shi, X. Zhang, and J. Dong, "Dynamic interferometry
measurement of orbital angular momentum of light," Opt. Lett., vol. 39,
no. 20, pp. 6058 -- 6061, 2014.
[77] G. C. Berkhout and M. W. Beijersbergen, "Method for probing the
orbital angular momentum of optical vortices in electromagnetic waves
from astronomical objects," Phys. Rev. Lett., vol. 101, no. 10, p.
100801, 2008.
[78] G. Chengshan, L. Leilei, and W. Huitian, "Characterizing topological
charge of optical vortices by using an annular aperture," Opt. Lett.,
vol. 34, no. 23, pp. 3686 -- 3688, 2009.
[79] J. Leach, J. Courtial, K. Skeldon, S. M. Barnett, S. Franke-Arnold, and
M. J. Padgett, "Interferometric methods to measure orbital and spin,
or the total angular momentum of a single photon," Phys. Rev. Lett.,
vol. 92, no. 1, p. 013601, 2004.
[80] G. C. Berkhout, M. P. Lavery, J. Courtial, M. W. Beijersbergen, and
M. J. Padgett, "Efficient sorting of orbital angular momentum states of
light," Phys. Rev. Lett., vol. 105, no. 15, p. 153601, 2010.
[81] M. P. J. Lavery, D. J. Robertson, G. C. G. Berkhout, G. D. Love, M. J.
Padgett, and J. Courtial, "Refractive elements for the measurement
of the orbital angular momentum of a single photon," Opt. Express,
vol. 20, no. 3, pp. 2110 -- 2115, 2012.
[82] M. Mirhosseini, M. Malik, Z. Shi, and R. W. Boyd, "Efficient sepa-
ration of the orbital angular momentum eigenstates of light," Nature
Commun., vol. 4, p. 2781, 2013.
[83] E. Karimi, B. Piccirillo, E. Nagali, L. Marrucci, and E. Santamato,
"Efficient generation and sorting of orbital angular momentum eigen-
modes of light by thermally tuned q-plates," Appl. Phys. Lett., vol. 94,
no. 23, p. 231124, 2009.
[84] P. Jia, Y. Yang, C. J. Min, H. Fang, and X.-C. Yuan, "Sidelobe-
modulated optical vortices for free-space communication," Opt. Lett.,
vol. 38, no. 4, pp. 588 -- 590, 2013.
[85] J. Long, R. Liu, F. Wang, Y. Wang, P. Zhang, H. Gao, and F. Li,
"Evaluating Laguerre-Gaussian beams with an invariant parameter,"
Opt. Lett., vol. 38, no. 16, pp. 3047 -- 3049, 2013.
[86] N. K. Fontaine, C. R. Doerr, and L. L. Buhl, "Efficient multiplexing and
demultiplexing of free-space orbital angular momentum using photonic
integrated circuits," in Proc. Opt. Fiber Commun. Conf. Expo., 2012,
p. OTu1I.2.
[87] T. Su, R. P. Scott, S. S. Djordjevic, N. K. Fontaine, D. J. Geisler,
X. Cai, and S. J. B. Yoo, "Demonstration of free space coherent
optical communication using integrated silicon photonic orbital angular
momentum devices," Opt. Express, vol. 20, no. 9, pp. 9396 -- 9402, 2012.
[88] Y. Yan, G. Xie, M. P. J. Lavery, H. Huang, N. Ahmed, C. Bao, Y. Ren,
Y. Cao, L. Li, Z. Zhao, A. F. Molisch, M. Tur, M. J. Padgett, and A. E.
Willner, "High-capacity millimetre-wave communications with orbital
angular momentum multiplexing," Nature Commun., vol. 5, p. 4876,
2014.
[89] S. M. Mohammadi, L. K. S. Daldorff, K. Forozesh, B. Thid´e, J. E. S.
Bergman, B. Isham, R. Karlsson, and T. D. Carozzi, "Orbital angular
27
momentum in radio: Measurement methods," Radio Sci., vol. 45, no. 4,
pp. 1 -- 14, 2010.
[90] M. Xie, X. Gao, M. Zhao, W. Zhai, W. Xu, J. Qian, M. Lei, and
S. Huang, "Mode measurement of a dual-mode radio frequency orbital
angular momentum beam by circular phase gradient method," IEEE
Antennas Wireless Propag. Lett., vol. 16, pp. 1143 -- 1146, 2017.
[91] Y. Hu, S. Zheng, Z. Zhang, H. Chi, X. Jin, and X. Zhang, "Simulation
of orbital angular momentum radio communication systems based
on partial aperture sampling receiving scheme," IET Microwaves,
Antennas Propag., vol. 10, no. 10, pp. 1043 -- 1047, 2016.
[92] H. Wu, Y. Yuan, Z. Zhang, and J. Cang, "UCA-based orbital angular
momentum radio beam generation and reception under different array
configurations," in Proc. 6th Int. Conf. Wireless Commun. Signal
Process., 2014.
[93] S. Zheng, X. Hui, J. Zhu, H. Chi, X. Jin, S. Yu, and X. Zhang, "Orbital
angular momentum mode-demultiplexing scheme with partial angular
receiving aperture," Opt. Express, vol. 23, no. 9, pp. 12 251 -- 12 257,
2015.
[94] C. Shi, M. Dubois, Y. Wang, and X. Zhang, "High-speed acoustic
communication by multiplexing orbital angular momentum," Proc.
Natl. Acad. Sci., vol. 114, no. 28, pp. 7250 -- 7253, 2017.
[95] X. Jiang, B. Liang, J. Cheng, and C. Qiu, "Twisted acoustics:
Metasurface-enabled multiplexing and demultiplexing," Adv. Mater.,
vol. 30, no. 18, p. 1800257, 2018.
[96] M. Krenn, R. Fickler, M. Fink, J. Handsteiner, M. Malik, T. Scheidl,
R. Ursin, and A. Zeilinger, "Communication with spatially modulated
light through turbulent air across vienna," New J. Phys., vol. 16, no. 11,
p. 113028, 2014.
[97] M. Krenn, J. Handsteiner, M. Fink, R. Fickler, R. Ursin, M. Malik,
and A. Zeilinger, "Twisted light transmission over 143 km," Proc. Natl.
Acad. Sci., vol. 113, no. 48, pp. 13 648 -- 13 653, 2016.
[98] H. Huang, G. Xie, Y. Yan, N. Ahmed, Y. Ren, Y. Yue, D. Rogawski,
M. J. Willner, B. I. Erkmen, and K. M. Birnbaum, "100 Tbit/s free-
space data link enabled by three-dimensional multiplexing of orbital
angular momentum, polarization, and wavelength," Opt. Lett., vol. 39,
no. 2, pp. 197 -- 200, 2014.
[99] J. Lin, X. Yuan, S. H. Tao, and R. E. Burge, "Multiplexing free-space
optical signals using superimposed collinear orbital angular momentum
states," Appl. Opt., vol. 46, no. 21, pp. 4680 -- 4685, 2007.
[100] L. Zou, X. Gu, and L. Wang, "High-dimensional free-space optical
communications based on orbital angular momentum coding," Opt.
Commun., vol. 410, pp. 333 -- 337, 2018.
[101] J. Wang, J.-Y. Yang, I. M. Fazal, N. Ahmed, Y. Yan, H. Huang, Y. Ren,
Y. Yue, S. Dolinar, M. Tur, and A. E. Willner, "Terabit free-space data
transmission employing orbital angular momentum multiplexing," Nat.
Photonics, vol. 6, no. 7, pp. 488 -- 496, 2012.
[102] J. Wang, S. Li, M. Luo, J. Liu, L. Zhu, C. Li, D. Xie, Q. Yang,
S. Yu, J. Sun, X. Zhang, W. Shieh, and A. E. Willner, "N-dimentional
multiplexing link with 1.036-Pbit/s transmission capacity and 112.6-
bit/s/Hz spectral efficiency using OFDM-8QAM signals over 368
WDM pol-muxed 26 OAM modes," in Proc. Eur. Conf. Opt. Commun.,
2014, pp. 1 -- 3.
[103] Y. Ren, Z. Wang, P. Liao, L. Li, G. Xie, H. Huang, Z. Zhao, and
Y. Yan, "Experimental characterization of a 400 Gbit/s orbital angular
momentum multiplexed free-space optical link over 120 m," Opt. Lett.,
vol. 41, no. 3, pp. 622 -- 625, 2016.
[104] A. Trichili, C. Rosales-Guzm´an, A. Dudley, B. Ndagano, A. B. Salem,
M. Zghal, and A. Forbes, "Optical communication beyond orbital
angular momentum," Sci. Rep., vol. 6, p. 27674, 2016.
[105] G. Xie, Y. Ren, Y. Yan, H. Huang, N. Ahmed, L. Li, Z. Zhao, C. Bao,
M. Tur, S. Ashrafi, and A. E. Willner, "Experimental demonstration of a
200-Gbit/s free-space optical link by multiplexing Laguerre-Gaussian
beams with different radial indices," Opt. Lett., vol. 41, no. 15, pp.
3447 -- 3450, 2016.
[106] A. Gatto, M. Tacca, P. Martelli, P. Boffi, and M. Martinelli, "Free-space
orbital angular momentum division multiplexing with Bessel beams,"
J. Opt., vol. 13, no. 6, p. 064018, 2011.
[107] J. Du and J. Wang, "High-dimensional structured light coding/decoding
for free-space optical communications free of obstructions," Opt. Lett.,
vol. 40, no. 21, pp. 4827 -- 4830, 2015.
[108] N. Ahmed, Z. Zhao, L. Li, H. Huang, M. P. J. Lavery, P. Liao, Y. Yan,
Z. Wang, G. Xie, Y. Ren, A. Almaiman, A. J. Willner, S. Ashrafi,
A. F. Molisch, M. Tur, and A. E. Willner, "Mode-division-multiplexing
of multiple Bessel-Gaussian beams carrying orbital-angular-momentum
for obstruction-tolerant free-space optical and millimetre-wave commu-
nication links," Sci. Rep., vol. 6, p. 22082, 2016.
28
[109] F. Zhu, S. Huang, W. Shao, J. Zhang, M. Chen, W. Zhang, and J. Zeng,
"Free-space optical communication link using perfect vortex beams
carrying orbital angular momentum (OAM)," Opt. Commun., vol. 396,
pp. 50 -- 57, 2017.
[110] G. Xie, L. Li, Y. Ren, H. Huang, Y. Yan, N. Ahmed, Z. Zhao, M. P. J.
Lavery, N. Ashrafi, S. Ashrafi, R. Bock, M. Tur, A. F. Molisch, and
A. E. Willner, "Performance metrics and design considerations for a
free-space optical orbital-angular-momentum-multiplexed communica-
tion link," Optica, vol. 2, no. 4, pp. 357 -- 365, 2015.
[111] C. Paterson, "Atmospheric turbulence and orbital angular momentum
of single photons for optical communication," Phys. Rev. Lett., vol. 94,
no. 15, p. 153901, 2005.
[112] G. A. Tyler and R. W. Boyd, "Influence of atmospheric turbulence
on the propagation of quantum states of light carrying orbital angular
momentum," Opt. Lett., vol. 34, no. 2, pp. 142 -- 144, 2009.
[113] Y. Ren, H. Huang, G. Xie, N. Ahmed, Y. Yan, B. I. Erkmen,
N. Chandrasekaran, M. P. J. Lavery, N. K. Steinhoff, M. Tur, S. Dolinar,
M. Neifeld, M. J. Padgett, R. W. Boyd, J. H. Shapiro, and A. E. Willner,
"Atmospheric turbulence effects on the performance of a free space
optical link employing orbital angular momentum multiplexing," Opt.
Lett., vol. 38, no. 20, pp. 4062 -- 4065, 2013.
[114] B. Rodenburg, M. P. J. Lavery, M. Malik, M. N. O'Sullivan, M. Mirhos-
seini, D. J. Robertson, M. Padgett, and R. W. Boyd, "Influence of
atmospheric turbulence on states of light carrying orbital angular
momentum," Opt. Lett., vol. 37, no. 17, pp. 3735 -- 3737, 2012.
[115] S. Fu and C. Gao, "Influences of atmospheric turbulence effects on the
orbital angular momentum spectra of vortex beams," Photonics Res.,
vol. 4, no. 5, pp. B1 -- B4, 2016.
[116] S. Li and J. Wang, "Adaptive free-space optical communications
through turbulence using self-healing Bessel beams," Sci. Rep., vol. 7,
p. 43233, 2017.
[117] S. Li, S. Chen, C. Gao, A. E. Willner, and J. Wang, "Atmospheric tur-
bulence compensation in orbital angular momentum communications:
Advances and perspectives," Opt. Commun., vol. 408, pp. 68 -- 81, 2018.
[118] Y. Ren, G. Xie, H. Huang, C. Bao, Y. Yan, N. Ahmed, M. P. J. Lavery,
B. I. Erkmen, S. Dolinar, M. Tur, M. A. Neifeld, M. J. Padgett, R. W.
Boyd, J. H. Shapiro, and A. E. Willner, "Adaptive optics compensation
of multiple orbital angular momentum beams propagating through
emulated atmospheric turbulence," Opt. Lett., vol. 39, no. 10, pp. 2845 --
2848, 2014.
[119] Y. Ren, G. Xie, H. Huang, L. Li, N. Ahmed, Y. Yan, M. P. J. Lavery,
R. Bock, M. Tur, M. A. Neifeld, R. W. Boyd, J. H. Shapiro, and A. E.
Willner, "Turbulence compensation of an orbital angular momentum
and polarization-multiplexed link using a data-carrying beacon on a
separate wavelength," Opt. Lett., vol. 40, no. 10, pp. 2249 -- 2252, 2015.
[120] Y. Ren, G. Xie, H. Huang, N. Ahmed, Y. Yan, L. Li, C. Bao, M. P. J.
Lavery, M. Tur, M. A. Neifeld, R. W. Boyd, J. H. Shapiro, and
A. E. Willner, "Adaptive-optics-based simultaneous pre- and post-
turbulence compensation of multiple orbital-angular-momentum beams
in a bidirectional free-space optical link," Optica, vol. 1, no. 6, pp.
376 -- 382, 2014.
[121] Y. Ren, H. Huang, J.-Y. Yang, Y. Yan, N. Ahmed, Y. Yue, A. E. Willner,
K. Birnbaum, J. Choi, B. Erkmen, and S. Dolinar, "Correction of phase
distortion of an OAM mode using GS algorithm based phase retrieval,"
in Proc. Conf. Lasers Electro-Opt., 2012, p. CF3I.4.
[122] S. Fu, S. Zhang, T. Wang, and C. Gao, "Pre-turbulence compensation of
orbital angular momentum beams based on a probe and the Gerchberg-
Saxton algorithm," Opt. Lett., vol. 41, no. 14, pp. 3185 -- 3188, 2016.
[123] S. Fu, T. Wang, S. Zhang, Z. Zhang, Y. Zhai, and C. Gao, "Non-probe
compensation of optical vortices carrying orbital angular momentum,"
Photonics Res., vol. 5, no. 3, pp. 251 -- 255, 2017.
[124] G. Xie, Y. Ren, H. Huang, M. P. J. Lavery, N. Ahmed, Y. Yan, C. Bao,
L. Li, Z. Zhao, Y. Cao, M. Willner, M. Tur, S. J. Dolinar, R. W. Boyd,
J. H. Shapiro, and A. E. Willner, "Phase correction for a distorted
orbital angular momentum beam using a Zernike polynomials-based
stochastic-parallel-gradient-descent algorithm," Opt. Lett., vol. 40,
no. 7, pp. 1197 -- 1200, 2015.
[125] H. Huang, Y. Cao, G. Xie, Y. Ren, Y. Yan, C. Bao, N. Ahmed, M. A.
Neifeld, S. J. Dolinar, and A. E. Willner, "Crosstalk mitigation in a
free-space orbital angular momentum multiplexed communication link
using 4× 4 MIMO equalization," Opt. Lett., vol. 39, no. 15, pp. 4360 --
4363, 2014.
[126] D. J. Richardson, J. M. Fini, and L. E. Nelson, "Space-division
multiplexing in optical fibres," Nat. Photonics, vol. 7, no. 5, pp. 354 --
362, 2013.
[127] P. J. Winzer and G. J. Foschini, "MIMO capacities and outage
probabilities in spatially multiplexed optical transport systems," Opt.
Express, vol. 19, no. 17, pp. 16 680 -- 16 696, 2011.
[128] Y. Ren, Z. Wang, G. Xie, L. Li, A. J. Willner, Y. Cao, Z. Zhao,
Y. Yan, N. Ahmed, N. Ashrafi, S. Ashrafi, R. Bock, M. Tur, and A. E.
Willner, "Atmospheric turbulence mitigation in an OAM-based MIMO
free-space optical link using spatial diversity combined with MIMO
equalization," Opt. Lett., vol. 41, no. 11, pp. 2406 -- 2409, 2016.
[129] B. Zhu, T. F. Taunay, M. Fishteyn, X. Liu, S. Chandrasekhar, M. F.
Yan, J. M. Fini, E. M. Monberg, and F. V. Dimarcello, "112-Tb/s space-
division multiplexed DWDM transmission with 14-b/s/Hz aggregate
spectral efficiency over a 76.8-km seven-core fiber," Opt. Express,
vol. 19, no. 17, pp. 16 665 -- 16 671, 2011.
[130] J. Sakaguchi, Y. Awaji, N. Wada, A. Kanno, T. Kawanishi, T. Hayashi,
T. Taru, T. Kobayashi, and M. Watanabe, "Space division multiplexed
transmission of 109-Tb/s data signals using homogeneous seven-core
fiber," J. Lightw. Technol., vol. 30, no. 4, pp. 658 -- 665, 2012.
[131] S. Randel, R. Ryf, A. Sierra, P. J. Winzer, A. H. Gnauck, C. A.
Bolle, R. J. Essiambre, D. W. Peckham, A. Mccurdy, and R. Lingle,
"6×56-Gb/s mode-division multiplexed transmission over 33-km few-
mode fiber enabled by 6×6 MIMO equalization," Opt. Express, vol. 19,
no. 17, pp. 16 697 -- 16 707, 2011.
[132] S. Ramachandran, P. Kristensen, and M. F. Yan, "Generation and
propagation of radially polarized beams in optical fibers," Opt. Lett.,
vol. 34, no. 16, pp. 2525 -- 2527, 2009.
[133] N. Bozinovic, P. Kristensen, S. Ramachandran, and S. Golowich,
"Control of orbital angular momentum of light with optical fibers,"
Opt. Lett., vol. 37, no. 13, pp. 2451 -- 2453, 2012.
[134] N. Bozinovic, P. Kristensen, and S. Ramachandran, "Long-range fiber-
transmission of photons with orbital angular momentum," in Proc.
Conf. Lasers Electro-Opt., 2011, p. CTuB1.
[135] N. Bozinovic, Y. Yue, Y. Ren, M. Tur, P. Kristensen, H. Huang,
A. E. Willner, and S. Ramachandran, "Terabit-scale orbital angular
momentum mode division multiplexing in fibers," Science, vol. 340,
no. 6140, pp. 1545 -- 1548, 2013.
[136] Y. Yue, Y. Yan, N. Ahmed, J.-Y. Yang, L. Zhang, Y. Ren, H. Huang,
K. M. Birnbaum, B. I. Erkmen, S. Dolinar, M. Tur, and A. E. Willner,
"Mode properties and propagation effects of optical orbital angular
momentum (OAM) modes in a ring fiber," IEEE Photonics J., vol. 4,
no. 2, pp. 535 -- 543, 2012.
[137] S. Li and J. Wang, "Multi-orbital-angular-momentum multi-ring fiber
for high-density space-division multiplexing," IEEE Photonics J.,
vol. 5, no. 5, p. 7101007, 2013.
[138] H. Zhang, W. Zhang, L. Xi, X. Tang, X. Zhang, and X. Zhang, "A
new type circular photonic crystal fiber for orbital angular momentum
mode transmission," IEEE Photon. Technol. Lett., vol. 28, no. 13, pp.
1426 -- 1429, 2016.
[139] G. Zhou, G. Zhou, C. Chen, M. Xu, C. Xia, and Z. Hou, "Design and
analysis of a microstructure ring fiber for orbital angular momentum
transmission," IEEE Photonics J., vol. 8, no. 2, pp. 1 -- 12, 2016.
[140] X. Zeng, Y. Li, Q. Mo, W. Li, Y. Tian, Z. Liu, and J. Wu, "Ex-
perimental investigation of LP11 mode to OAM conversion in few
mode-polarization maintaining fiber and the usage for all fiber OAM
generator," IEEE Photonics J., vol. 8, no. 4, pp. 1 -- 7, 2016.
[141] Y. Li, L. Jin, H. Wu, S. Gao, Y.-H. Feng, and Z. Li, "Superposing
multiple LP modes with micro phase difference distributed along fiber
to generate OAM mode," IEEE Photonics J., vol. 9, no. 2, pp. 1 -- 9,
2017.
[142] S. Li, Z. Xu, R. Zhao, L. Shen, C. Du, and J. Wang, "Generation
of orbital angular momentum beam using fiber-to-fiber butt coupling,"
IEEE Photonics J., vol. 10, no. 4, pp. 1 -- 7, 2018.
[143] F. E. Mahmouli and S. D. Walker, "4-Gbps uncompressed video trans-
mission over a 60-GHz orbital angular momentum wireless channel,"
IEEE Wireless Commun. Lett., vol. 2, no. 2, pp. 223 -- 226, 2013.
[144] B. Allen, A. Tennant, Q. Bai, and E. Chatziantoniou, "Wireless data
encoding and decoding using OAM modes," Electron. Lett., vol. 50,
no. 3, pp. 232 -- 233, 2014.
[145] Y. Yan, L. Li, Z. Zhao, G. Xie, Z. Wang, Y. Ren, N. Ahmed,
S. Sajuyigbe, S. Talwar, M. Tur, N. Ashrafi, S. Ashrafi, A. F. Molisch,
and A. E. Willner, "32-Gbit/s 60-GHz millimeter-wave wireless com-
munication using orbital angular momentum and polarization multi-
plexing," in Proc. IEEE Int. Conf. Commun., 2016.
[146] Z. Zhao, Y. Yan, L. Li, G. Xie, Y. Ren, N. Ahmed, Z. Wang, C. Liu,
A. J. Willner, P. Song, H. Hashemi, H. Yao, D. Macfarlane, R. Hen-
derson, N. Ashrafi, S. Ashrafi, S. Talwar, S. Sajuyigbe, M. Tur, A. F.
Molisch, and A. E. Willner, "A dual-channel 60 GHz communications
link using patch antenna arrays to generate data-carrying orbital-
angular-momentum beams," in Proc. IEEE Int. Conf. Commun., 2016.
[147] Z. Zhao, Y. Ren, G. Xie, Y. Yan, L. Li, H. Huang, C. Bao, N. Ahmed,
M. P. Lavery, C. Zhang, N. Ashrafi, S. Ashrafi, S. Talwar, S. Sajuyigbe,
M. Tur, A. F. Molisch, and A. E. Willner, "Experimental demonstration
of 16-Gbit/s millimeter-wave communications link using thin meta-
material plates to generate data-carrying orbital-angular-momentum
beams," in Proc. IEEE Int. Conf. Commun., 2015, pp. 1392 -- 1397.
[148] X. Hui, S. Zheng, Y. Chen, Y. Hu, X. Jin, H. Chi, and X. Zhang,
"Multiplexed millimeter wave communication with dual orbital angular
momentum (OAM) mode antennas," Sci. Rep., vol. 5, p. 10148, 2015.
[149] W. Zhang, S. Zheng, X. Hui, R. Dong, X. Jin, H. Chi, and X. Zhang,
"Mode division multiplexing communication using microwave orbital
angular momentum: An experimental study," IEEE Trans. Wireless
Commun., vol. 16, no. 2, pp. 1308 -- 1318, 2017.
[150] O. Edfors and A. J. Johansson, "Is orbital angular momentum (OAM)
based radio communication an unexploited area?" IEEE Trans. Anten-
nas Propag., vol. 60, no. 2, pp. 1126 -- 1131, 2012.
[151] M. Tamagnone, C. Craeye, and J. Perruisseau-Carrier, "Comment
on 'Encoding many channels on the same frequency through radio
vorticity: first experimental test'," New J. Phys., vol. 14, no. 11, p.
118001, 2012.
[152] R. Chen, H. Xu, W. Yang, X. Wang, and J. Li, "On the performance
of OAM in keyhole channels," IEEE Wireless Commun. Lett., pp. 1 -- 4,
2018.
[153] M. Oldoni, F. Spinello, E. Mari, G. Parisi, C. G. Someda, F. Tamburini,
F. Romanato, R. A. Ravanelli, P. Coassini, and B. Thid´e, "Space-
division demultiplexing in orbital-angular-momentum-based MIMO
radio systems," IEEE Trans. Antennas Propag., vol. 63, no. 10, pp.
4582 -- 4587, 2015.
[154] Z. Zhang, S. Zheng, Y. Chen, X. Jin, H. Chi, and X. Zhang, "The
capacity gain of orbital angular momentum based multiple-input-
multiple-output system," Sci. Rep., no. 6, p. 25418, 2016.
[155] X. Ge, R. Zi, X. Xiong, Q. Li, and L. Wang, "Millimeter wave
communications with OAM-SM scheme for future mobile networks,"
IEEE J. Sel. Areas Commun., vol. 35, no. 9, pp. 2163 -- 2177, 2017.
[156] Y. Ren, L. Li, G. Xie, Y. Yan, Y. Cao, H. Huang, N. Ahemd, M. J.
Lavery, Z. Zhao, C. Zhang, M. Tur, M. Padgett, G. Caire, A. F.
Molisch, and A. E. Willner, "Experimental demonstration of 16 Gbit/s
millimeter-wave communications using MIMO processing of 2 OAM
modes on each of two transmitter/receiver antenna apertures," in Proc.
IEEE Global Commun. Conf., 2014, pp. 3821 -- 3826.
[157] Y. Ren, L. Li, G. Xie, Y. Yan, Y. Cao, H. Huang, N. Ahmed, Z. Zhao,
P. Liao, C. Zhang, G. Caire, A. F. Molisch, M. Tur, and A. E.
Willner, "Line-of-sight millimeter-wave communications using orbital
angular momentum multiplexing combined with conventional spatial
multiplexing," IEEE Trans. Wireless Commun., vol. 16, no. 5, pp. 3151 --
3161, 2017.
[158] Y. Ren, Z. Wang, G. Xie, L. Li, Y. Cao, C. Liu, P. Liao, Y. Yan,
N. Ahmed, Z. Zhao, A. Willner, N. Ashrafi, S. Ashrafi, R. D. Linquist,
R. Bock, M. Tur, A. F. Molisch, and A. E. Willner, "Free-space
optical communications using orbital-angular-momentum multiplexing
combined with MIMO-based spatial multiplexing," Opt. Lett., vol. 40,
no. 18, pp. 4210 -- 4213, 2015.
[159] Y. Zhang, W. Feng, and N. Ge, "On the restriction of utilizing orbital
angular momentum in radio communications," in 8th International
Conference on Communications and Networking in China, 2013, pp.
271 -- 275.
[160] R. Chen, H. Xu, M. Moretti, and J. Li, "Beam steering for the
misalignment in UCA-based OAM communication systems," IEEE
Wireless Commun. Lett., vol. 7, no. 4, pp. 582 -- 585, 2018.
[161] Y. Yan, L. Li, G. Xie, C. Bao, P. Liao, H. Huang, Y. Ren, N. Ahmed,
Z. Zhao, M. P. Lavery, N. Ashrafi, S. Ashrafi, S. Talwar, S. Sajuyigbe,
M. Tur, A. F. Molisch, and A. E. Willner, "Experimental measurements
of multipath-induced intra- and inter-channel crosstalk effects in a
millimeter-wave communications link using orbital-angular-momentum
multiplexing," in Proc. IEEE Int. Conf. Commun., 2015, pp. 1370 --
1375.
[162] Y. Yan, L. Li, G. Xie, C. Bao, P. Liao, H. Huang, Y. Ren, N. Ahmed,
Z. Zhao, Z. Wang, N. Ashrafi, S. Ashrafi, S. Talwar, S. Sajuyigbe,
M. Tur, A. F. Molisch, and A. E. Willner, "Multipath effects in
millimetre-wave wireless communication using orbital angular momen-
tum multiplexing," Sci. Rep., vol. 6, p. 33482, 2016.
[163] R. Chen, W. Yang, H. Xu, and J. Li, "A 2-D FFT-based transceiver
architecture for OAM-OFDM systems with UCA antennas," IEEE
Trans. Veh. Technol., vol. 67, no. 6, pp. 5481 -- 5485, 2018.
29
[164] A. Ashkin, J. M. Dziedzic, J. E. Bjorkholm, and S. Chu, "Observation
of a single-beam gradient force optical trap for dielectric particles,"
Opt. Lett., vol. 11, no. 5, pp. 288 -- 290, 1986.
[165] H. He, M. E. J. Friese, N. R. Heckenberg, and H. Rubinsztein-Dunlop,
"Direct observation of transfer of angular momentum to absorptive
particles from a laser beam with a phase singularity," Phys. Rev. Lett.,
vol. 75, no. 5, pp. 826 -- 829, 1995.
[166] N. B. Simpson, K. Dholakia, L. Allen, and M. J. Padgett, "Mechanical
equivalence of spin and orbital angular momentum of light: an optical
spanner," Opt. Lett., vol. 22, no. 1, pp. 52 -- 54, 1997.
[167] M. E. J. Friese, J. Enger, H. Rubinsztein-Dunlop, and N. R. Heck-
enberg, "Optical angular-momentum transfer to trapped absorbing
particles," Phys. Rev. A, vol. 54, no. 2, pp. 1593 -- 1596, 1996.
[168] A. T. O'Neil, I. Macvicar, L. Allen, and M. J. Padgett, "Intrinsic and
extrinsic nature of the orbital angular momentum of a light beam,"
Phys. Rev. Lett., vol. 88, no. 5, p. 053601, 2002.
[169] V. Garc´es-Ch´avez, D. Mcgloin, M. J. Padgett, W. Dultz, H. Schmitzer,
and K. Dholakia, "Observation of the transfer of the local angular
momentum density of a multiringed light beam to an optically trapped
particle," Phys. Rev. Lett., vol. 91, no. 9, p. 093602, 2003.
[170] C. R. P. Courtney, B. W. Drinkwater, C. E. M. Demore, S. Cochran,
A. Grinenko, and P. D. Wilcox, "Dexterous manipulation of micropar-
ticles using Bessel-function acoustic pressure fields," Appl. Phys. Lett.,
vol. 102, no. 12, p. 123508, 2013.
[171] C. R. P. Courtney, C. E. M. Demore, H. Wu, A. Grinenko, P. D.
Wilcox, S. Cochran, and B. W. Drinkwater, "Independent trapping
and manipulation of microparticles using dexterous acoustic tweezers,"
Appl. Phys. Lett., vol. 104, no. 15, p. 154103, 2014.
[172] A. Anhauser, R. Wunenburger, and E. Brasselet, "Acoustic rotational
manipulation using orbital angular momentum transfer," Phys. Rev.
Lett., vol. 109, no. 3, p. 034301, 2012.
[173] S. W. Hell and J. Wichmann, "Breaking the diffraction resolution limit
by stimulated emission: stimulated-emission-depletion fluorescence mi-
croscopy," Opt. Lett., vol. 19, no. 11, pp. 780 -- 782, 1994.
[174] L. Li and F. Li, "Beating the rayleigh limit: Orbital-angular-
momentum-based super-resolution diffraction tomography," Phys. Rev.
E, vol. 88, no. 3, p. 033205, 2013.
[175] C. Maurer, A. Jesacher, S. Bernet, and M. Ritsch-Marte, "What spatial
light modulators can do for optical microscopy," Laser Photonics Rev.,
vol. 5, no. 1, pp. 81 -- 101, 2011.
[176] G. A. Swartzlander, "Peering into darkness with a vortex spatial filter,"
Opt. Lett., vol. 26, no. 8, pp. 497 -- 499, 2001.
[177] S. Furhapter, A. Jesacher, S. Bernet, and M. Ritsch-Marte, "Spiral
phase contrast imaging in microscopy," Opt. Express, vol. 13, no. 3,
pp. 689 -- 694, 2005.
[178] J. Wang, W. Zhang, Q. Qi, S. Zheng, and L. Chen, "Gradual edge
enhancement in spiral phase contrast imaging with fractional vortex
filters," Sci. Rep., vol. 5, p. 15826, 2015.
[179] G. A. Swartzlander, E. L. Ford, R. S. Abdul-Malik, L. M. Close, M. A.
Peters, D. M. Palacios, and D. W. Wilson, "Astronomical demonstration
of an optical vortex coronagraph," Opt. Express, vol. 16, no. 14, pp.
10 200 -- 10 207, 2008.
[180] S. Furhapter, A. Jesacher, S. Bernet, and M. Ritsch-Marte, "Spiral
interferometry," Opt. Lett., vol. 30, no. 15, pp. 1953 -- 1955, 2005.
[181] G. Guo, W. Hu, and X. Du, "Electromagnetic vortex based radar target
imaging," J. Nat. Univ. Defense Technol., vol. 35, no. 6, pp. 71 -- 76,
2013.
[182] K. Liu, Y. Cheng, Z. Yang, H. Wang, Y. Qin, and X. Li, "Orbital-
angular-momentum-based electromagnetic vortex imaging," IEEE An-
tennas Wireless Propag. Lett., vol. 14, pp. 711 -- 714, 2015.
[183] T. Yuan, H. Wang, Y. Qin, and Y. Cheng, "Electromagnetic vortex
imaging using uniform concentric circular arrays," IEEE Antennas
Wireless Propag. Lett., vol. 15, pp. 1024 -- 1027, 2016.
[184] C. Zhang and D. Chen, "Large-scale orbital angular momentum radar
pulse generation with rotational antenna," IEEE Antennas Wireless
Propag. Lett., vol. 16, pp. 2316 -- 2319, 2017.
[185] K. Liu, Y. Cheng, Y. Gao, X. Li, Y. Qin, and H. Wang, "Super-
resolution radar imaging based on experimental OAM beams," Appl.
Phys. Lett., vol. 110, no. 16, p. 164102, 2017.
[186] X. Bu, Z. Zhang, L. Chen, X. Liang, H. Tang, and X. Wang, "Imple-
mentation of vortex electromagnetic waves high-resolution synthetic
aperture radar imaging," IEEE Antennas Wireless Propag. Lett., vol. 17,
no. 5, pp. 764 -- 767, 2018.
[187] K. Liu, Y. Cheng, X. Li, Y. Qin, H. Wang, and Y. Jiang, "Generation of
orbital angular momentum beams for electromagnetic vortex imaging,"
IEEE Antennas Wireless Propag. Lett., vol. 15, pp. 1873 -- 1876, 2016.
30
[188] M. Lin, Y. Gao, P. Liu, and J. Liu, "Super-resolution orbital angular
momentum based radar targets detection," Electron. Lett., vol. 52,
no. 13, pp. 1168 -- 1170, 2016.
[189] -- -- , "Improved OAM-based radar targets detection using uniform
concentric circular arrays," Int. J. Antennas Propag., vol. 2016, pp.
1 -- 8, 2016.
[190] R. Chen, W.-X. Long, Y. Gao, and J. Li, "Orbital angular momentum-
based two-dimensional super-resolution targets imaging," in Proc.
IEEE Global Conf. Signal Inf. Process., 2018.
[191] T. Yuan, Y. Cheng, H. Wang, and Y. Qin, "Beam steering for electro-
magnetic vortex imaging using uniform circular arrays," IEEE Antennas
Wireless Propag. Lett., vol. 16, pp. 704 -- 707, 2017.
[192] T. Yuan, H. Liu, Y. Cheng, Y. Qin, and H. Wang, "Orbital-angular-
momentum-based electromagnetic vortex imaging by least-squares
method," in Proc. IEEE Int. Geosci. Remote Sens. Symp., 2016, pp.
6645 -- 6648.
Xiaodong Wang (S'98-M'98-SM'04-F'08) received
the Ph.D. degree in electrical engineering from
Princeton University. He is currently a Professor
of electrical engineering with Columbia University,
New York, NY, USA. He has authored the book
Wireless Communication Systems: Advanced Tech-
niques for Signal Reception (Prentice-Hall, 2003).
His research interests include computing, signal
processing, and communications, and has published
extensively in these areas. His current research in-
terests include wireless communications, statistical
signal processing, and genomic signal processing. He was a recipient of the
1999 NSF CAREER Award, the 2001 IEEE Communications Society and
Information Theory Society Joint Paper Award, and the 2011 IEEE Com-
munication Society Award for Outstanding Paper on New Communication
Topics. He was an Associate Editor for the IEEE TRANSACTIONS ON
COMMUNICATIONS, the IEEE TRANSACTIONS ON WIRELESS COM-
MUNICATIONS, the IEEE TRANSACTIONS ON SIGNAL PROCESSING,
and the IEEE TRANSACTIONS ON INFORMATION THEORY. He is listed
as an ISI highly cited author.
Rui Chen (S'08-M'11) received the B.S., M.S. and
Ph.D. degrees in Communications and Information
Systems from Xidian University, Xian, China, in
2005, 2007 and 2011, respectively. From 2014 to
2015, he was a visiting scholar at Columbia Uni-
versity in the City of New York. He is currently
an associate professor and Ph.D. supervisor in the
school of Telecommunications Engineering at Xid-
ian University. He has published about 50 papers in
international journals and conferences and held 10
patents. He is an Associate Editor for International
Journal of Electronics, Communications, and Measurement Engineering (IGI
Global). His research interests include broadband wireless communication
systems, array signal processing and intelligent transportation systems.
Hong Zhou received the B.S. degree (with Highest
Hons.) in Communications Engineering from Hefei
University of Technology, Hefei, China in 2018. She
is currently pursuing a M.S. degree in Communica-
tions and Information Systems at Xidian University.
Her research interests include orbital angular mo-
mentum (OAM) communication systems and array
signal processing.
Jiandong Li (SM'05) received the B.E., M.S., and
Ph.D. degrees in communications engineering from
Xidian University, Xian, China, in 1982, 1985, and
1991, respectively. He was a Visiting Professor
with the Department of Electrical and Computer
Engineering, Cornell University, from 2002 to 2003.
He has been a faculty member of the School of
Telecommunications Engineering, Xidian Univer-
sity, since 1985, where he is currently a Professor
and the Vice Director of the Academic Committee,
State Key Laboratory of Integrated Service Net-
works. His major research interests include wireless communication theory,
cognitive radio, and signal processing. He was recognized as a Distinguished
Young Researcher by NSFC and a Changjiang Scholar by the Ministry of
Education, China, respectively. He served as the General Vice Chair of
ChinaCom 2009 and the TPC Chair of the IEEE ICCC 2013.
(M'03)
Marco Moretti
received the degree in
electronic engineering from the University of Flo-
rence, Florence, Italy, in 1995, and the Ph.D. degree
from the Delft University of Technology, Delft, The
Netherlands, in 2000. From 2000 to 2003, he was
a Senior Researcher with Marconi Mobile. He is
currently an Associate Professor with the University
of Pisa, Pisa, Italy. His research interests include
resource allocation for multicarrier systems, syn-
chronization, and channel estimation. He is currently
an Associate Editor of the IEEE TRANSACTIONS
ON SIGNAL PROCESSING.
(cid:36)(cid:88)(cid:87)(cid:75)(cid:82)(cid:85)(cid:10)(cid:86)(cid:3)(cid:83)(cid:76)(cid:70)(cid:87)(cid:88)(cid:85)(cid:72)(cid:3)(cid:9)(cid:3)(cid:69)(cid:76)(cid:82)(cid:74)(cid:85)(cid:68)(cid:83)(cid:75)(cid:92)(cid:3)(cid:3)(cid:3)(cid:3)(cid:53)(cid:88)(cid:76)(cid:3)(cid:38)(cid:75)(cid:72)(cid:81)(cid:3)(cid:85)(cid:72)(cid:70)(cid:72)(cid:76)(cid:89)(cid:72)(cid:71)(cid:3)(cid:87)(cid:75)(cid:72)(cid:3)(cid:37)(cid:17)(cid:54)(cid:17)(cid:15)(cid:3)(cid:48)(cid:17)(cid:54)(cid:17)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:51)(cid:75)(cid:17)(cid:39)(cid:17)(cid:3)(cid:71)(cid:72)(cid:74)(cid:85)(cid:72)(cid:72)(cid:86)(cid:3)(cid:76)(cid:81)(cid:3)(cid:38)(cid:82)(cid:80)(cid:80)(cid:88)(cid:81)(cid:76)(cid:70)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:86)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:40)(cid:79)(cid:72)(cid:70)(cid:87)(cid:85)(cid:82)(cid:81)(cid:76)(cid:70)(cid:3)(cid:54)(cid:92)(cid:86)(cid:87)(cid:72)(cid:80)(cid:3)(cid:73)(cid:85)(cid:82)(cid:80)(cid:3)(cid:59)(cid:76)(cid:71)(cid:76)(cid:68)(cid:81)(cid:3)(cid:56)(cid:81)(cid:76)(cid:89)(cid:72)(cid:85)(cid:86)(cid:76)(cid:87)(cid:92)(cid:15)(cid:3)(cid:59)(cid:76)(cid:182)(cid:68)(cid:81)(cid:15)(cid:3)(cid:38)(cid:75)(cid:76)(cid:81)(cid:68)(cid:15)(cid:3)(cid:76)(cid:81)(cid:3)(cid:21)(cid:19)(cid:19)(cid:24)(cid:15)(cid:3)(cid:21)(cid:19)(cid:19)(cid:26)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:21)(cid:19)(cid:20)(cid:20)(cid:15)(cid:3)(cid:85)(cid:72)(cid:86)(cid:83)(cid:72)(cid:70)(cid:87)(cid:76)(cid:89)(cid:72)(cid:79)(cid:92)(cid:17)(cid:3)(cid:41)(cid:85)(cid:82)(cid:80)(cid:3)(cid:21)(cid:19)(cid:20)(cid:23)(cid:3)(cid:87)(cid:82)(cid:3)(cid:21)(cid:19)(cid:20)(cid:24)(cid:15)(cid:3)(cid:75)(cid:72)(cid:3)(cid:90)(cid:68)(cid:86)(cid:3)(cid:68)(cid:3)(cid:89)(cid:76)(cid:86)(cid:76)(cid:87)(cid:76)(cid:81)(cid:74)(cid:3)(cid:86)(cid:70)(cid:75)(cid:82)(cid:79)(cid:68)(cid:85)(cid:3)(cid:68)(cid:87)(cid:3)(cid:38)(cid:82)(cid:79)(cid:88)(cid:80)(cid:69)(cid:76)(cid:68)(cid:3)(cid:56)(cid:81)(cid:76)(cid:89)(cid:72)(cid:85)(cid:86)(cid:76)(cid:87)(cid:92)(cid:17)(cid:3)(cid:43)(cid:72)(cid:3)(cid:76)(cid:86)(cid:3)(cid:70)(cid:88)(cid:85)(cid:85)(cid:72)(cid:81)(cid:87)(cid:79)(cid:92)(cid:3)(cid:68)(cid:81)(cid:3)(cid:68)(cid:86)(cid:86)(cid:82)(cid:70)(cid:76)(cid:68)(cid:87)(cid:72)(cid:3)(cid:83)(cid:85)(cid:82)(cid:73)(cid:72)(cid:86)(cid:86)(cid:82)(cid:85)(cid:3)(cid:76)(cid:81)(cid:3)(cid:87)(cid:75)(cid:72)(cid:3)(cid:86)(cid:70)(cid:75)(cid:82)(cid:82)(cid:79)(cid:3)(cid:82)(cid:73)(cid:3)(cid:55)(cid:72)(cid:79)(cid:72)(cid:70)(cid:82)(cid:80)(cid:80)(cid:88)(cid:81)(cid:76)(cid:70)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:86)(cid:3)(cid:40)(cid:81)(cid:74)(cid:76)(cid:81)(cid:72)(cid:72)(cid:85)(cid:76)(cid:81)(cid:74)(cid:3)(cid:68)(cid:87)(cid:3)(cid:59)(cid:76)(cid:71)(cid:76)(cid:68)(cid:81)(cid:3)(cid:56)(cid:81)(cid:76)(cid:89)(cid:72)(cid:85)(cid:86)(cid:76)(cid:87)(cid:92)(cid:17)(cid:3)(cid:43)(cid:76)(cid:86)(cid:3)(cid:85)(cid:72)(cid:86)(cid:72)(cid:68)(cid:85)(cid:70)(cid:75)(cid:3)(cid:76)(cid:81)(cid:87)(cid:72)(cid:85)(cid:72)(cid:86)(cid:87)(cid:86)(cid:3)(cid:76)(cid:81)(cid:70)(cid:79)(cid:88)(cid:71)(cid:72)(cid:3)(cid:48)(cid:44)(cid:48)(cid:50)(cid:3)(cid:68)(cid:85)(cid:85)(cid:68)(cid:92)(cid:3)(cid:86)(cid:76)(cid:74)(cid:81)(cid:68)(cid:79)(cid:3)(cid:83)(cid:85)(cid:82)(cid:70)(cid:72)(cid:86)(cid:86)(cid:76)(cid:81)(cid:74)(cid:3)(cid:82)(cid:73)(cid:3)(cid:90)(cid:76)(cid:85)(cid:72)(cid:79)(cid:72)(cid:86)(cid:86)(cid:3)(cid:70)(cid:82)(cid:80)(cid:80)(cid:88)(cid:81)(cid:76)(cid:70)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:3)(cid:86)(cid:92)(cid:86)(cid:87)(cid:72)(cid:80)(cid:86)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:76)(cid:81)(cid:87)(cid:72)(cid:79)(cid:79)(cid:76)(cid:74)(cid:72)(cid:81)(cid:87)(cid:3)(cid:87)(cid:85)(cid:68)(cid:81)(cid:86)(cid:83)(cid:82)(cid:85)(cid:87)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:3)(cid:86)(cid:92)(cid:86)(cid:87)(cid:72)(cid:80)(cid:86)(cid:17)(cid:3)(cid:3)(cid:3)(cid:38)(cid:75)(cid:68)(cid:81)(cid:74)(cid:79)(cid:72)(cid:3)(cid:47)(cid:76)(cid:3)(cid:85)(cid:72)(cid:70)(cid:72)(cid:76)(cid:89)(cid:72)(cid:71)(cid:3)(cid:87)(cid:75)(cid:72)(cid:3)(cid:37)(cid:17)(cid:40)(cid:81)(cid:74)(cid:17)(cid:3)(cid:71)(cid:72)(cid:74)(cid:85)(cid:72)(cid:72)(cid:3)(cid:76)(cid:81)(cid:3)(cid:40)(cid:79)(cid:72)(cid:70)(cid:87)(cid:85)(cid:82)(cid:80)(cid:68)(cid:74)(cid:81)(cid:72)(cid:87)(cid:76)(cid:70)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:48)(cid:76)(cid:70)(cid:85)(cid:82)(cid:90)(cid:68)(cid:89)(cid:72)(cid:3)(cid:55)(cid:72)(cid:70)(cid:75)(cid:81)(cid:82)(cid:79)(cid:82)(cid:74)(cid:92)(cid:3)(cid:76)(cid:81)(cid:3)(cid:20)(cid:28)(cid:28)(cid:27)(cid:15)(cid:3)(cid:48)(cid:17)(cid:40)(cid:81)(cid:74)(cid:17)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:51)(cid:75)(cid:17)(cid:39)(cid:17)(cid:3)(cid:71)(cid:72)(cid:74)(cid:85)(cid:72)(cid:72)(cid:86)(cid:3)(cid:76)(cid:81)(cid:3)(cid:38)(cid:82)(cid:80)(cid:80)(cid:88)(cid:81)(cid:76)(cid:70)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:44)(cid:81)(cid:73)(cid:82)(cid:85)(cid:80)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:3)(cid:54)(cid:92)(cid:86)(cid:87)(cid:72)(cid:80)(cid:3)(cid:76)(cid:81)(cid:3)(cid:21)(cid:19)(cid:19)(cid:20)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:21)(cid:19)(cid:19)(cid:24)(cid:15)(cid:3)(cid:85)(cid:72)(cid:86)(cid:83)(cid:72)(cid:70)(cid:87)(cid:76)(cid:89)(cid:72)(cid:79)(cid:92)(cid:15)(cid:3)(cid:68)(cid:79)(cid:79)(cid:3)(cid:73)(cid:85)(cid:82)(cid:80)(cid:3)(cid:59)(cid:76)(cid:71)(cid:76)(cid:68)(cid:81)(cid:3)(cid:56)(cid:81)(cid:76)(cid:89)(cid:72)(cid:85)(cid:86)(cid:76)(cid:87)(cid:92)(cid:15)(cid:3)(cid:59)(cid:76)(cid:182)(cid:68)(cid:81)(cid:15)(cid:3)(cid:38)(cid:75)(cid:76)(cid:81)(cid:68)(cid:17)(cid:3)(cid:54)(cid:76)(cid:81)(cid:70)(cid:72)(cid:3)(cid:21)(cid:19)(cid:19)(cid:24)(cid:15)(cid:3)(cid:75)(cid:72)(cid:3)(cid:75)(cid:68)(cid:86)(cid:3)(cid:69)(cid:72)(cid:72)(cid:81)(cid:3)(cid:90)(cid:76)(cid:87)(cid:75)(cid:3)(cid:54)(cid:87)(cid:68)(cid:87)(cid:72)(cid:3)(cid:46)(cid:72)(cid:92)(cid:3)(cid:47)(cid:68)(cid:69)(cid:82)(cid:85)(cid:68)(cid:87)(cid:82)(cid:85)(cid:92)(cid:3)(cid:82)(cid:73)(cid:3)(cid:44)(cid:81)(cid:87)(cid:72)(cid:74)(cid:85)(cid:68)(cid:87)(cid:72)(cid:71)(cid:3)(cid:54)(cid:72)(cid:85)(cid:89)(cid:76)(cid:70)(cid:72)(cid:3)(cid:49)(cid:72)(cid:87)(cid:90)(cid:82)(cid:85)(cid:78)(cid:86)(cid:3)(cid:68)(cid:87)(cid:3)(cid:59)(cid:76)(cid:71)(cid:76)(cid:68)(cid:81)(cid:3)(cid:56)(cid:81)(cid:76)(cid:89)(cid:72)(cid:85)(cid:86)(cid:76)(cid:87)(cid:92)(cid:17)(cid:3)(cid:41)(cid:85)(cid:82)(cid:80)(cid:3)(cid:21)(cid:19)(cid:19)(cid:25)(cid:3)(cid:87)(cid:82)(cid:3)(cid:21)(cid:19)(cid:19)(cid:26)(cid:15)(cid:3)(cid:75)(cid:72)(cid:3)(cid:90)(cid:68)(cid:86)(cid:3)(cid:90)(cid:76)(cid:87)(cid:75)(cid:3)(cid:38)(cid:82)(cid:80)(cid:83)(cid:88)(cid:87)(cid:72)(cid:85)(cid:3)(cid:54)(cid:70)(cid:76)(cid:72)(cid:81)(cid:70)(cid:72)(cid:3)(cid:39)(cid:72)(cid:83)(cid:68)(cid:85)(cid:87)(cid:80)(cid:72)(cid:81)(cid:87)(cid:3)(cid:68)(cid:87)(cid:3)(cid:56)(cid:81)(cid:76)(cid:89)(cid:72)(cid:85)(cid:86)(cid:76)(cid:87)(cid:92)(cid:3)(cid:82)(cid:73)(cid:3)(cid:48)(cid:82)(cid:81)(cid:70)(cid:87)(cid:82)(cid:81)(cid:15)(cid:3)(cid:49)(cid:37)(cid:15)(cid:3)(cid:38)(cid:68)(cid:81)(cid:68)(cid:71)(cid:68)(cid:3)(cid:68)(cid:86)(cid:3)(cid:68)(cid:3)(cid:83)(cid:82)(cid:86)(cid:87)(cid:71)(cid:82)(cid:70)(cid:87)(cid:82)(cid:85)(cid:68)(cid:79)(cid:3)(cid:85)(cid:72)(cid:86)(cid:72)(cid:68)(cid:85)(cid:70)(cid:75)(cid:72)(cid:85)(cid:17)(cid:3)(cid:41)(cid:85)(cid:82)(cid:80)(cid:3)(cid:21)(cid:19)(cid:19)(cid:26)(cid:3)(cid:87)(cid:82)(cid:3)(cid:21)(cid:19)(cid:19)(cid:28)(cid:15)(cid:3)(cid:75)(cid:72)(cid:3)(cid:90)(cid:68)(cid:86)(cid:3)(cid:68)(cid:81)(cid:3)(cid:72)(cid:91)(cid:83)(cid:72)(cid:85)(cid:87)(cid:3)(cid:85)(cid:72)(cid:86)(cid:72)(cid:68)(cid:85)(cid:70)(cid:75)(cid:72)(cid:85)(cid:3)(cid:90)(cid:76)(cid:87)(cid:75)(cid:3)(cid:49)(cid:44)(cid:38)(cid:55)(cid:3)(cid:82)(cid:73)(cid:3)(cid:45)(cid:68)(cid:83)(cid:68)(cid:81)(cid:17)(cid:3)(cid:43)(cid:76)(cid:86)(cid:3)(cid:85)(cid:72)(cid:86)(cid:72)(cid:68)(cid:85)(cid:70)(cid:75)(cid:3)(cid:76)(cid:81)(cid:87)(cid:72)(cid:85)(cid:72)(cid:86)(cid:87)(cid:86)(cid:3)(cid:68)(cid:79)(cid:86)(cid:82)(cid:3)(cid:76)(cid:81)(cid:70)(cid:79)(cid:88)(cid:71)(cid:72)(cid:3)(cid:89)(cid:72)(cid:75)(cid:76)(cid:70)(cid:88)(cid:79)(cid:68)(cid:85)(cid:3)(cid:87)(cid:72)(cid:70)(cid:75)(cid:81)(cid:82)(cid:79)(cid:82)(cid:74)(cid:92)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:70)(cid:82)(cid:80)(cid:80)(cid:88)(cid:81)(cid:76)(cid:70)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:3)(cid:83)(cid:85)(cid:82)(cid:87)(cid:82)(cid:70)(cid:82)(cid:79)(cid:86)(cid:3)(cid:73)(cid:82)(cid:85)(cid:3)(cid:70)(cid:72)(cid:79)(cid:79)(cid:88)(cid:79)(cid:68)(cid:85)(cid:15)(cid:3)(cid:80)(cid:82)(cid:69)(cid:76)(cid:79)(cid:72)(cid:3)(cid:68)(cid:71)(cid:3)(cid:75)(cid:82)(cid:70)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:86)(cid:72)(cid:81)(cid:86)(cid:82)(cid:85)(cid:3)(cid:81)(cid:72)(cid:87)(cid:90)(cid:82)(cid:85)(cid:78)(cid:86)(cid:17)(cid:3)(cid:3)(cid:45)(cid:76)(cid:68)(cid:81)(cid:71)(cid:82)(cid:81)(cid:74)(cid:3)(cid:47)(cid:76)(cid:3)(cid:90)(cid:68)(cid:86)(cid:3)(cid:74)(cid:85)(cid:68)(cid:71)(cid:88)(cid:68)(cid:87)(cid:72)(cid:71)(cid:3)(cid:73)(cid:85)(cid:82)(cid:80)(cid:3)(cid:59)(cid:76)(cid:71)(cid:76)(cid:68)(cid:81)(cid:3)(cid:56)(cid:81)(cid:76)(cid:89)(cid:72)(cid:85)(cid:86)(cid:76)(cid:87)(cid:92)(cid:3)(cid:90)(cid:76)(cid:87)(cid:75)(cid:3)(cid:37)(cid:68)(cid:70)(cid:75)(cid:72)(cid:79)(cid:82)(cid:85)(cid:3)(cid:39)(cid:72)(cid:74)(cid:85)(cid:72)(cid:72)(cid:15)(cid:3)(cid:48)(cid:68)(cid:86)(cid:87)(cid:72)(cid:85)(cid:3)(cid:39)(cid:72)(cid:74)(cid:85)(cid:72)(cid:72)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:51)(cid:75)(cid:17)(cid:3)(cid:39)(cid:3)(cid:76)(cid:81)(cid:3)(cid:38)(cid:82)(cid:80)(cid:80)(cid:88)(cid:81)(cid:76)(cid:70)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:86)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:40)(cid:79)(cid:72)(cid:70)(cid:87)(cid:85)(cid:82)(cid:81)(cid:76)(cid:70)(cid:3)(cid:54)(cid:92)(cid:86)(cid:87)(cid:72)(cid:80)(cid:3)(cid:85)(cid:72)(cid:86)(cid:83)(cid:72)(cid:70)(cid:87)(cid:76)(cid:89)(cid:72)(cid:79)(cid:92)(cid:3)(cid:76)(cid:81)(cid:3)(cid:20)(cid:28)(cid:27)(cid:21)(cid:15)(cid:3)(cid:20)(cid:28)(cid:27)(cid:24)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:20)(cid:28)(cid:28)(cid:20)(cid:17)(cid:3)(cid:39)(cid:88)(cid:85)(cid:76)(cid:81)(cid:74)(cid:3)(cid:21)(cid:19)(cid:19)(cid:21)(cid:16)(cid:21)(cid:19)(cid:19)(cid:22)(cid:15)(cid:3)(cid:43)(cid:72)(cid:3)(cid:90)(cid:68)(cid:86)(cid:3)(cid:68)(cid:3)(cid:89)(cid:76)(cid:86)(cid:76)(cid:87)(cid:76)(cid:81)(cid:74)(cid:3)(cid:83)(cid:85)(cid:82)(cid:73)(cid:72)(cid:86)(cid:86)(cid:82)(cid:85)(cid:3)(cid:76)(cid:81)(cid:3)(cid:38)(cid:82)(cid:85)(cid:81)(cid:72)(cid:79)(cid:79)(cid:3)(cid:56)(cid:81)(cid:76)(cid:89)(cid:72)(cid:85)(cid:86)(cid:76)(cid:87)(cid:92)(cid:15)(cid:3)(cid:56)(cid:54)(cid:36)(cid:17)(cid:3)(cid:43)(cid:72)(cid:3)(cid:76)(cid:86)(cid:3)(cid:81)(cid:82)(cid:90)(cid:3)(cid:68)(cid:3)(cid:83)(cid:85)(cid:82)(cid:73)(cid:72)(cid:86)(cid:86)(cid:82)(cid:85)(cid:3)(cid:82)(cid:73)(cid:3)(cid:59)(cid:76)(cid:71)(cid:76)(cid:68)(cid:81)(cid:3)(cid:56)(cid:81)(cid:76)(cid:89)(cid:72)(cid:85)(cid:86)(cid:76)(cid:87)(cid:92)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:72)(cid:91)(cid:72)(cid:70)(cid:88)(cid:87)(cid:76)(cid:89)(cid:72)(cid:3)(cid:71)(cid:72)(cid:68)(cid:81)(cid:3)(cid:82)(cid:73)(cid:3)(cid:42)(cid:85)(cid:68)(cid:71)(cid:88)(cid:68)(cid:87)(cid:72)(cid:3)(cid:54)(cid:70)(cid:75)(cid:82)(cid:82)(cid:79)(cid:17)(cid:3)(cid:43)(cid:72)(cid:3)(cid:76)(cid:86)(cid:3)(cid:68)(cid:79)(cid:86)(cid:82)(cid:3)(cid:71)(cid:76)(cid:85)(cid:72)(cid:70)(cid:87)(cid:82)(cid:85)(cid:3)(cid:82)(cid:73)(cid:3)(cid:44)(cid:81)(cid:87)(cid:72)(cid:74)(cid:85)(cid:68)(cid:87)(cid:72)(cid:71)(cid:3)(cid:54)(cid:72)(cid:85)(cid:89)(cid:76)(cid:70)(cid:72)(cid:3)(cid:49)(cid:72)(cid:87)(cid:90)(cid:82)(cid:85)(cid:78)(cid:3)(cid:55)(cid:75)(cid:72)(cid:82)(cid:85)(cid:92)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:46)(cid:72)(cid:92)(cid:3)(cid:55)(cid:72)(cid:70)(cid:75)(cid:81)(cid:82)(cid:79)(cid:82)(cid:74)(cid:92)(cid:3)(cid:54)(cid:87)(cid:68)(cid:87)(cid:72)(cid:3)(cid:46)(cid:72)(cid:92)(cid:3)(cid:47)(cid:68)(cid:69)(cid:82)(cid:85)(cid:68)(cid:87)(cid:82)(cid:85)(cid:92)(cid:17)(cid:3)(cid:43)(cid:72)(cid:3)(cid:76)(cid:86)(cid:3)(cid:68)(cid:3)(cid:86)(cid:72)(cid:81)(cid:76)(cid:82)(cid:85)(cid:3)(cid:80)(cid:72)(cid:80)(cid:69)(cid:72)(cid:85)(cid:3)(cid:82)(cid:73)(cid:3)(cid:44)(cid:40)(cid:40)(cid:40)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:38)(cid:75)(cid:76)(cid:81)(cid:68)(cid:3)(cid:44)(cid:81)(cid:86)(cid:87)(cid:76)(cid:87)(cid:88)(cid:87)(cid:72)(cid:3)(cid:82)(cid:73)(cid:3)(cid:40)(cid:79)(cid:72)(cid:70)(cid:87)(cid:85)(cid:82)(cid:81)(cid:76)(cid:70)(cid:86)(cid:3)(cid:11)(cid:38)(cid:44)(cid:40)(cid:12)(cid:3)(cid:68)(cid:81)(cid:71)(cid:3)(cid:87)(cid:75)(cid:72)(cid:3)(cid:73)(cid:72)(cid:79)(cid:79)(cid:82)(cid:90)(cid:3)(cid:82)(cid:73)(cid:3)(cid:38)(cid:75)(cid:76)(cid:81)(cid:68)(cid:3)(cid:44)(cid:81)(cid:86)(cid:87)(cid:76)(cid:87)(cid:88)(cid:87)(cid:72)(cid:3)(cid:82)(cid:73)(cid:3)(cid:38)(cid:82)(cid:80)(cid:80)(cid:88)(cid:81)(cid:76)(cid:70)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:3)(cid:11)(cid:38)(cid:44)(cid:38)(cid:12)(cid:17)(cid:3)(cid:43)(cid:72)(cid:3)(cid:90)(cid:68)(cid:86)(cid:3)(cid:87)(cid:75)(cid:72)(cid:3)(cid:80)(cid:72)(cid:80)(cid:69)(cid:72)(cid:85)(cid:3)(cid:82)(cid:73)(cid:3)(cid:51)(cid:38)(cid:49)(cid:3)(cid:86)(cid:83)(cid:72)(cid:70)(cid:76)(cid:68)(cid:79)(cid:76)(cid:86)(cid:87)(cid:3)(cid:74)(cid:85)(cid:82)(cid:88)(cid:83)(cid:3)(cid:73)(cid:82)(cid:85)(cid:3)(cid:38)(cid:75)(cid:76)(cid:81)(cid:68)(cid:3)(cid:27)(cid:25)(cid:22)(cid:3) |
1909.03635 | 1 | 1909 | 2019-09-09T05:22:32 | Nb$_{2}$SiTe$_{4}$: A Stable Narrow-Gap Two-Dimensional Material with Ambipolar Transport and Mid-Infrared Response | [
"physics.app-ph",
"cond-mat.mes-hall"
] | Two-dimensional (2D) materials with narrow band gaps (~0.3 eV) are of great importance for realizing ambipolar transistors and mid-infrared (MIR) detection. However, most of the 2D materials studied so far have band gaps that are too large. A few of them with suitable band gaps are not stable under ambient conditions. In this study, the layered Nb$_{2}$SiTe$_{4}$ is shown to be a stable 2D material with a band gap of 0.39 eV. Field-effect transistors based on few-layer Nb$_2$SiTe$_4$ show ambipolar transport with similar magnitude of electron and hole current and high charge-carrier mobility of ~ 100 cm$^{2}$V$^{-1}$s$^{-1}$ at room temperature. Optoelectronic measurements of the devices show clear response to MIR wavelength of 3.1 $\mathrm\mu$m with a high responsivity of ~ 0.66 AW$^{-1}$. These results establish Nb$_{2}$SiTe$_{4}$ as a good candidate for ambipolar devices and MIR detection. | physics.app-ph | physics | Nb2SiTe4: A Stable Narrow-Gap Two-Dimensional
Material with Ambipolar Transport and Mid-
Infrared Response
Mingxing Zhao1,3,4†, Wei Xia1,4†, Yang Wang2, Man Luo2, Zhen Tian1, Yanfeng Guo1*, Weida
Hu2* and Jiamin Xue1*
1School of Physical Science and Technology, ShanghaiTech University, Shanghai 201210, China
2Shanghai Institute of Technical Physics, Shanghai 200083, China
3Shanghai Institute of Ceramics, Shanghai 200050, China
4University of Chinese Academy of Sciences, Beijing 100190, China
KEYWORDS: Nb2SiTe4, ambipolar transistor, mid-infrared, two-dimensional material
ABSTRACT
Two-dimensional (2D) materials with narrow band gaps (~ 0.3 eV) are of great importance for
realizing ambipolar transistors and mid-infrared (MIR) detections. However, most of the 2D
materials studied so far have band gaps that are too large. A few of them with suitable band gaps
are not stable under ambient conditions. In this study, the layered Nb2SiTe4 is shown to be a
1
stable 2D material with a band gap of 0.39 eV. Field-effect transistors based on few-layer
Nb2SiTe4 show ambipolar transport with similar magnitude of electron and hole current and high
charge-carrier mobility of ~ 100 cm2V− 1s− 1 at room temperature. Optoelectronic measurements
of the devices show clear response to MIR wavelength of 3.1 µm with a high responsivity of ~
0.66 AW− 1. These results establish Nb2SiTe4 as a good candidate for ambipolar devices and MIR
detection.
TEXT
Materials with narrow band gaps in the range of 0.15 eV to 0.4 eV hold special position in the
family of semiconductors for their applications in mid-infrared (MIR) detections. Among them,
indium arsenide, mercury cadmium telluride and so on are the most studied and used materials.
Besides these conventional semiconductors, newly developed narrow-gap two-dimensional (2D)
materials are also showing their great potential in the MIR detections. For example, few-layer
black phosphorus (bP) with a band gap of ~ 0.3 eV has been used to demonstrate high photo
responsivity and detectivity in the MIR range.1-3 When alloyed with arsenide, the band gap of
black arsenic phosphorus (bAP) can be further tuned from ~ 0.3 eV to ~ 0.15 eV, which extends
the spectrum to the long-wavelength infrared.4-5 The layered nature of these materials gives them
large flexibility and ease of integration with other materials to form various device structures.3, 6-7
Besides their applications in optoelectronics, narrow-gap 2D semiconductors such as bP can
be used as the channel material of field-effect transistors (FET) and support ambipolar transport,
in which the majority carriers of the channel can be easily switched between electrons and holes
by electrostatic doping of the gate voltage.8-9 This property offers great opportunity for novel
2
device functions such as field-programmable p-n junctions,10 which are not achievable by
conventional semiconductors.
Due to these unique applications, narrow-gap 2D materials are highly demanded. However,
most of the 2D semiconductors studied so far have band gaps in the order of 1 eV or larger.11 bP
and bAP have suitable band gaps, but degrade within hours under ambient condition due to the
chemical instability of phosphorus.4, 12-13 Thus new stable narrow-gap 2D materials are needed.
Here we report a layered material Nb2SiTe4 (NST) with a band gap of 0.39 eV. When few-layer
NST is fabricated into an FET, it shows ambipolar transport with similar capability of
conducting electrons and holes. A respectable hole mobility of ~ 100 cm2V− 1s− 1 at room
temperature is obtained. When illuminated by MIR light with 3.1 µm wavelength, NST devices
demonstrate high photo responsivity of ~ 0.66 AW− 1. They also show good stability under
ambient condition without any encapsulation. These findings demonstrate NST to be a stable
narrow-gap 2D material with great application potentials.
RESULTS AND DISCSSION
3
Figure 1. (a) Perspective view of few-layer NST showing atoms of Te (gold), Nb (black) and Si
(gray). (b) Top view and (c), (d) side views of monolayer NST. (e) XRD data of NST. The X-ray
source is Cu Kα line with the wave length of 0.154 nm. Inset: optical image of a bulk NST
crystal. (f) Atomically resolved image of a NST crystal measured by STM with the
corresponding fast Fourier transform in the inset. STM parameters: V = 0.5 V, I = 100 pA. (g)
The dI⁄dV spectra of a NST crystal. Inset is the same data plotted in log scale for a better
determination of band edges. The band gap Eg is measured to be 0.39 eV. STS parameters: Vinitial
= 0.5 V, Iinitial = 100 pA; lock-in frequency f = 991.2 Hz, modulation voltage Vrms = 10 mV.
NST has been synthesized decades ago,14 but there have been scarce studies about it. Bulk NST
crystal belongs to the space group of P121/c1 (No. 14),15 which is in the monoclinic crystal
system. Figure 1a to 1d show the different views of the NST structure. It is a layered material
4
formed by stacking the Te-(Nb, Si)-Te sandwich layers, which bears similarity to the transitional
metal dichalcogenides (TMDs) such as MoS2,16 NbSe2
17 and so on. The inner atomic layer
consists of two types of atoms (Nb and Si) instead of one in the TMDs. The lattice parameters
have been determined to a = 0.63 nm, b = 0.79 nm, c = 1.47 nm, α = 90°, β = 107°, γ = 90°.15 We
synthesized NST crystal with self-flux method (see the Experimental section). The as grown
crystals are black thin flakes with metal luster and sizes of millimeters (inset of Figure 1e). Due
to the layered nature, powder X-ray diffraction (XRD) mainly probes the basal planes in the
crystal and the layer distance is extracted to be 0.697 nm which gives a lattice constant c = 1.46
nm. Sharp Bragg peaks indicate the high quality of our crystals. To complement XRD study, we
used scanning tunneling microscope (STM) which resolves the atomic structure within the basal
plane (Figure 1f). The orthorhombic structure is clearly visible with lattice constants of 0.63 nm
and 0.78 nm, agreeing very well with literature values.15 From Figure 1 b and f we can see that a
unit cell of NST contains a large number of atoms, which gives many phonon modes as shown
by Raman spectroscopy (Figure S1 in the Supporting Information, SI). Besides revealing the
atomic structure, more importantly, STM can be used to measure the band structure with the
technique of scanning tunneling spectroscopy (STS).18-19 In Figure 1g, The Fermi level is located
at 0.08 eV above the valence band edge Ev, indicating its intrinsic p-type doping. The band gap
Eg is determined to be 0.39 eV, in the MIR range, which is promising for both ambipolar
transport and MIR detection.
5
Figure 2. (a) The optical image of the few-layer NST device with the height profile of the
channel along the white line. (b) Drain-source current Ids between electrodes A and D versus Vds
at different temperatures. (c) Four-probe conductance (G4t) as a function of back-gate voltage (Vg)
at various temperatures. (d) Four-probe hole mobility as a function of temperature extracted from
(c).
To measure the transport property of few-layer NST, scotch tape method is adopted to
exfoliate NST onto highly p-doped Si substrate covered with 300 nm thermally grown SiO2.
Then electron-beam lithography and metal deposition are used to fabricate the device. A typical
FET device is shown in the optical image of Figure 2a. The flake thickness is measured by
atomic force microscope (AFM) to be 7.6 nm (other thicknesses of few-layer NST showed
6
similar behavior). The metal electrodes consist of 10 nm of Ti and 40 nm of Au. All transport
measurements were performed in vacuum (~10−5 mbar) and in dark with a Janis ST-500 probe
station.
The linear two-terminal output curves measured from 300 K to 78 K (Figure 2b) indicate that
Ti forms good contacts with NST. We determine the work function of NST to be 4.36 eV by
ultraviolet photoelectron spectroscopy (Figure S2), which matches well with that of Ti (4.3
eV).20 To extract the intrinsic field-effect mobility, four-terminal transfer curves are measured as
a function of temperature (Figure 2c). The p-type transport is consistent with the band structure
measured by STS in Figure 1g. Mobility 𝜇 is calculated by =
𝐿
𝑊𝐶𝑜𝑥
d𝐺
d𝑉𝑔
, where 𝐺 is the
conductance, 𝑉g is
the back-gate voltage, 𝐶𝑜𝑥 = 1.15 × 10− 8 F cm− 2 is
the specific
capacitance of gate oxide, and 𝐿 (𝑊) is the length (width) of the channel. The mobility at room
temperature is 98 cm2V−1s−1 and it increases as temperature is lowered, reaching 160
cm2V−1s−1 at 78 K (Figure 2d). Compared with the majority of 2D materials, NST has a higher
mobility,21-22 which is essential for high performance electronic devices. Similar transport
properties have been obtained on 8 devices (see Figure S3 for data of two other devices),
although some variations in the transport properties such as mobility and threshold voltage exist
among them. With device optimization, such as better substrates and dielectric screening,23-24 the
mobility should be further improved.
7
Figure 3. (a) Transfer curve of a NST device showing ambipolar transport. Vn-th and Vp-th
represent the threshold voltages on the electron side and hole side, respectively. Inset: schematic
of equivalent capacitance model of the device. Cd and Cit represent the depletion layer
capacitance and interface capacitance, respectively. (b) Data from (a) plotted in log scale. Red
lines are the linear fitting curves to extract the subthreshold swings of the electron side (SSn) and
hole side (SSp).
Due to the narrow band gap, it is possible to change the majority carrier type in the channel
from holes to electrons simply by the back-gate voltage. Figure 3a shows the transfer curve of a
5.7 nm NST FET (Figure S4), where ambipolar behavior can be clearly seen. The electron and
hole currents have comparable magnitude, indicating the similar Schottky barrier heights
between Ti and the conduction and valence bands of NST. Previously reported narrow-gap 2D
materials such as bP can also support ambipolar transport,8-9 but the electron current is usually
orders of magnitude lower than the hole current, which undermines the tunability of a device.
8
Further analysis of the data in Figure 3a can be used to extract the transport gap, which is given
by25
𝐸g = 𝑒 (𝑉ds + (
𝑉n−th
𝛽n
−
𝑉p−th
𝛽p
)), (1)
where 𝑒 is the electron charge, 𝑉ds is the source-drain bias (100 mV in Figure 3a), 𝑉n−th (𝑉p−th)
is the threshold voltage on the electron (hole) side, and 𝛽n (𝛽p) is the gate efficiency in tuning
the band edge on the electron (hole) side, which is defined below. The threshold voltages are
determined from Figure 3a. The subthreshold swing (SS) of a FET can be written as26
𝑆𝑆 = ln10
𝑘𝑇
𝑒
(1 +
𝐶d+𝐶it
𝐶ox
) ≡ ln10
𝑘𝑇
𝑒
𝛽, (2)
where 𝑘 is the Boltzmann constant, 𝑇 is the temperature (150 K in our case), and 𝐶d (𝐶it) is the
depletion (interface) capacitance. The gate efficiency 𝛽 is defined as 𝛽 = 1 +
𝐶d+𝐶it
𝐶ox
. The
equivalent capacitance circuit is shown in the inset of Figure 3a. Due to the presence of 𝐶d and
𝐶it, the gate voltage Vg in the subthreshold regime can only cause a down scaled shift of the
bands in the device. The scaling factor 𝛽 can be extracted from the fittings in Figure 3b. Since
the electron side and hole side show slightly different SS, both values are used in Equation 1. The
calculated transport gap of NST is 0.18 eV, less than the STS value. This difference may come
from the thermal broadening of the band edges and the non-uniform doping in the device.
9
Figure 4. (a) Response to laser pulses with 532 nm wavelength. The rising and falling time is
defined in the main text. (b) Response to MIR laser pulses with 3.1 μm wave length.
The band gap values of NST measured by STS and transport indicate that NST devices should
have MIR responses. To test its optoelectronic property, we first use 532 nm laser pulses to
characterize the NST FETs as shown in Figure 4a. With 50 mV source-drain bias, 0 V gate
voltage and 310 µW laser, the device shows ~ 1 µA of photocurrent and fast response. We define
the rising and falling time as the time span between 90% and 10% of the total photocurrent,
which gives 20 ms rising edge and 310 ms decaying edge. We note that the dark current is high
(~ 7 µA) due to the intrinsic heavy p type doping. Response to MIR wavelength (3.1 µm) is also
measured as shown in Figure 4b, with the laser power of ~160 nW. The photocurrent is about
100 nA which gives a responsivity of 0.66 A W-1, much larger than the responsivity of bAP.4
These results demonstrate the potential of NST for MIR detection.
10
Figure 5. (a) Ids versus Vds under different test conditions. (b) Transfer curves show the slight
degradation of mobility after the device is exposed to ambient air for 120 hours. Inset: the day-
to-day variation of the device mobility. (c) Raman spectra taken before and after the transport
measurements shown in (a) and (b). (d) AFM image of the channel after the transport
measurements showing the smooth intact surface.
The ambipolar transport and MIR response have shown the application potential of NST.
Stability is another important property we should examine as it has always been a great challenge
for narrow bandgap semiconductors, such as bP and bAP.4, 12 These materials tend to react
11
quickly with oxygen and water moisture in air and degrade completely in the time scale of
minutes to several hours.27 Exfoliations of these materials are usually done in glove boxes with
water and oxygen contents less than 1 ppm. Special care during device fabrication and various
encapsulation techniques are needed to protect them,4, 13, 23, 28 which bring in extra difficulties
and complexities. However, NST shows remarkable resistance to ambient conditions. Exfoliation
and inspection are all carried out in air. As shown in the transport data in Figure 5a and 5b, when
a fabricated device is stored in vacuum for 120 hours, the mobility increases slightly (from black
to red curve), due to the desorption of surface contaminant. When exposed to dried air for
another 102 hours, mobility is slightly decreased with an increase of conductance (green curve),
presumably due to the mild oxidation effect from the air. When the device is stored in air with 40%
relative humidity for another 120 hours, the device still shows good transport behavior (blue
curve), with 40% reduction of its mobility. In the inset of Figure 5b, day-to-day variation of the
mobility is monitored. After exposed to air, the mobility decreases linearly without obvious
change of speed when the air is switched from dry to ambient air with moisture, which indicates
the key role of oxygen instead of water. As a comparison, bP and bAP devices usually
completely lost their electrical conductivity after a few hours exposure to moisture air.28 Raman
spectra of the same device before and after measurements also reveal the stability of NST
(Figure 5 c). All the peaks remain their positions and line profiles, and no new peaks can be seen.
The decrease in intensity can be ascribed to the surface oxidization of NST. The AFM image
(Figure 5d) of the channel after all the characterizations shows the cleanness and flatness of the
flake, as compared with the chemically-corroded rough surface of bP and bAP.4, 12, 29 A similar
stability can be seen on another device in Figure S5 of the SI.
12
CONCLUSIONS
In conclusion, we have explored the layered material Nb2SiTe4 as a 2D narrow-gap
semiconductor. Ambipolar transport with respectable mobility and comparable electron-hole
injection indicate its outstanding electrical property. Good responsivity to MIR wavelength
demonstrates the great potential for optoelectronic applications. Besides these, its superior
stability renders Nb2SiTe4 a unique position among all the narrow-gap 2D materials, which
warrants its further exploration.
METHODS
Crystal growth : Nb2SiTe4 crystals were grown by using Si:Te=1:4 as the flux. A mixture of Nb
(99.95%, Aladdin), Si (99.999%, Aladdin) and Te (99.99%, Aladdin) in a molar ratio of 1:2:8
was placed into an alumina crucible. The crucible was sealed into a quartz tube in vacuum and
then heated in a furnace up to 1150 oC in 15 hours. After reaction at this temperature for 5 hours,
the assembly was slowly cooled down to 750 oC and stayed at 750 oC for more than 10 hours.
The excess Si and Te was quickly removed at this temperature in a centrifuge. Thin pieces of
black Nb2SiTe4 flakes with metallic luster were obtained at the alumina crucible.
STM and STS: The STM experiments were carried out in the Omicron LT-Qpuls-STM with
tungsten tip calibrated on Ag(111) surface. To get clean surface, the NST bulk sample was
exfoliated in ultrahigh vacuum chamber with the base pressure < 10− 9 mbar and then transferred
to STM chamber to start the scanning with the base pressure of 10− 11 mbar and temperature of
77 K.
Device fabrication and transport measurements: All the FET devices in this study start with
NST exfoliation on highly doped silicon wafer covered with 300 nm SiO2. The electron beam
13
lithography is used to design the contacts. Thermal evaporation of electrode metals (10 nm Ti
and 40 nm Au) with lift off method is used to deposit the contacts. Then the devices are ready for
transport measurement, which is carried out in a Janis ST-500 probe station under vacuum.
Keithley source meters 2612B provide the source-drain voltage and back gate voltage while
measure the current.
XRD: Powder X-ray diffraction was performed on Nb2SiTe4 single crystals aligned along the
(001) plane, using a D8 Venture Bruker at room temperature. A Cu kα source (1.5418 Å) was
used.
Raman: A custom made micro Raman system with ANDOR SR-500i-D2-R spectrometer and
1800 grooves per mm grating was used. The incident laser wavelength was 532 nm.
AFM: All AFM images were obtained by AFM (Oxford, Cypher S) under ambient conditions.
ASSOCIATED CONTENT
Supporting Information.
The following files are available free of charge. Additional figures and data to support the results
in the main text (PDF).
AUTHOR INFORMATION
Corresponding Authors
*[email protected], [email protected], [email protected]
Author Contributions
14
The manuscript was written through contributions of all authors. All authors have given approval
to the final version of the manuscript. †These authors contributed equally.
ACKNOWLEDGMENT
We thank Prof. B. Chen for initial help with the MIR measurements. The device fabrication was
supported by the Centre for High-resolution Electron Microscopy (CℏEM), SPST, ShanghaiTech
University under contract No. EM02161943. Drs. N. Yu, X. Wang and Z. Zou helped with the
XRD. M. Z., Z.T. and J.X. are supported by the Thousand Talents Project and ShanghaiTech
University. W. X. and Y.G. are supported by the Natural Science Foundation of Shanghai
(17ZR1443300) and the Shanghai Pujiang Program (17PJ1406200). Y. W., M. L. and W. H. are
supported by the National Natural Science Foundation of China (61674157, 11734016).
REFERENCES
(1). Guo, Q.; Pospischil, A.; Bhuiyan, M.; Jiang, H.; Tian, H.; Farmer, D.; Deng, B.; Li, C.;
Han, S. J.; Wang, H.; Xia, Q.; Ma, T. P.; Mueller, T.; Xia, F., Black Phosphorus Mid-Infrared
Photodetectors with High Gain. Nano Lett. 2016, 16, 4648-55.
(2). Youngblood, N.; Chen, C.; Koester, S. J.; Li, M., Waveguide-integrated black
phosphorus photodetector with high responsivity and low dark current. Nat. Photonics 2015, 9,
247-252.
(3). Yuan, H.; Liu, X.; Afshinmanesh, F.; Li, W.; Xu, G.; Sun, J.; Lian, B.; Curto, A. G.; Ye,
G.; Hikita, Y.; Shen, Z.; Zhang, S. C.; Chen, X.; Brongersma, M.; Hwang, H. Y.; Cui, Y.,
Polarization-sensitive broadband photodetector using a black phosphorus vertical p-n junction.
Nat. Nanotechnol. 2015, 10, 707-13.
(4).
Long, M.; Gao, A.; Wang, P.; Xia, H.; Ott, C.; Pan, C.; Fu, Y.; Liu, E.; Chen, X.; Lu, W.;
Nilges, T.; Xu, J.; Wang, X.; Hu, W.; Miao, F., Room temperature high-detectivity mid-infrared
photodetectors based on black arsenic phosphorus. Sci. Adv. 2017, 3, e1700589.
(5).
Bachhuber, F.; Pielnhofer, F.; Wang, H.; Dhall, R.; Cronin, S. B.; Ge, M.; Fang, X.; Nilges, T.;
Zhou, C., Black Arsenic-Phosphorus: Layered Anisotropic Infrared Semiconductors with Highly
Tunable Compositions and Properties. Adv. Mater. 2015, 27, 4423-4429.
(6). Ye, L.; Li, H.; Chen, Z.; Xu, J., Near-Infrared Photodetector Based on MoS2/Black
Phosphorus Heterojunction. ACS Photonics 2016, 3, 692-699.
Liu, B.; Kopf, M.; Abbas, A. N.; Wang, X.; Guo, Q.; Jia, Y.; Xia, F.; Weihrich, R.;
15
Li, L.; Yu, Y.; Ye, G. J.; Ge, Q.; Ou, X.; Wu, H.; Feng, D.; Chen, X. H.; Zhang, Y.,
Shim, J.; Oh, S.; Kang, D. H.; Jo, S. H.; Ali, M. H.; Choi, W. Y.; Heo, K.; Jeon, J.; Lee,
(7).
S.; Kim, M.; Song, Y. J.; Park, J. H., Phosphorene/rhenium disulfide heterojunction-based
negative differential resistance device for multi-valued logic. Nat. Commun. 2016, 7, 13413.
(8).
Black phosphorus field-effect transistors. Nat. Nanotechnol. 2014, 9, 372-377.
(9). Han, C.; Hu, Z.; Gomes, L. C.; Bao, Y.; Carvalho, A.; Tan, S. J. R.; Lei, B.; Xiang, D.;
Wu, J.; Qi, D.; Wang, L.; Huo, F.; Huang, W.; Loh, K. P.; Chen, W., Surface Functionalization
of Black Phosphorus via Potassium toward High-Performance Complementary Devices. Nano
Lett. 2017, 17, 4122-4129.
(10). Tian, H.; Deng, B.; Chin, M. L.; Yan, X.; Jiang, H.; Han, S. J.; Sun, V.; Xia, Q.; Dubey,
M.; Xia, F.; Wang, H., A Dynamically Reconfigurable Ambipolar Black Phosphorus Memory
Device. ACS Nano 2016, 10, 10428-10435.
(11). Low, T.; Chaves, A.; Caldwell, J. D.; Kumar, A.; Fang, N. X.; Avouris, P.; Heinz, T. F.;
Guinea, F.; Martin-Moreno, L.; Koppens, F., Polaritons in layered two-dimensional materials.
Nat. Mater. 2017, 16, 182-194.
(12). Guo, Z.; Chen, S.; Wang, Z.; Yang, Z.; Liu, F.; Xu, Y.; Wang, J.; Yi, Y.; Zhang, H.; Liao,
L.; Chu, P. K.; Yu, X. F., Metal-Ion-Modified Black Phosphorus with Enhanced Stability and
Transistor Performance. Adv. Mater. 2017, 29, 1703811.
(13). Wan, B.; Yang, B.; Wang, Y.; Zhang, J.; Zeng, Z.; Liu, Z.; Wang, W., Enhanced stability
of black phosphorus field-effect transistors with SiO(2) passivation. Nanotechnology 2015, 26,
435702.
(14). Wolfgang Tremel , H. K., Volkmar Derstroff, Christian Reisner, Transition metal
chalcogenides : new views on an old topic. J. Alloys Compd. 1995, 219, 73-82.
(15). Springer Materials. https://materials.springer.com/isp/crystallographic/docs/sd_1414898
(accessed Junuary 8,2019).
(16). Wang, Q. H.; Kalantar-Zadeh, K.; Kis, A.; Coleman, J. N.; Strano, M. S., Electronics and
optoelectronics of two-dimensional transition metal dichalcogenides. Nat. Nanotechnol. 2012, 7,
699-712.
(17). Xi, X.; Berger, H.; Forro, L.; Shan, J.; Mak, K. F., Gate Tuning of Electronic Phase
Transitions in Two-Dimensional NbSe2. Phys. Rev. Lett. 2016, 117, 106801.
(18). Fischer, Ø.; Kugler, M.; Maggio-Aprile, I.; Berthod, C.; Renner, C., Scanning tunneling
spectroscopy of high-temperature superconductors. Rev. Mod. Phys. 2007, 79, 353-419.
(19). Li, G.; Luican, A.; Andrei, E. Y., Scanning tunneling spectroscopy of graphene on
graphite. Phys. Rev. Lett. 2009, 102, 176804.
(20). Das, S.; Chen, H. Y.; Penumatcha, A. V.; Appenzeller, J., High performance multilayer
MoS2 transistors with scandium contacts. Nano Lett. 2013, 13, 100-5.
(21). Zhou, X.; Zhang, Q.; Gan, L.; Li, H.; Xiong, J.; Zhai, T., Booming Development of
Group IV-VI Semiconductors: Fresh Blood of 2D Family. Adv. Sci. 2016, 3, 1600177.
(22). Manzeli, S.; Ovchinnikov, D.; Pasquier, D.; Yazyev, O. V.; Kis, A., 2D transition metal
dichalcogenides. Nat. Rev. Mater. 2017, 2, 17033.
(23). Li, L.; Yang, F.; Ye, G. J.; Zhang, Z.; Zhu, Z.; Lou, W.; Zhou, X.; Li, L.; Watanabe, K.;
Taniguchi, T.; Chang, K.; Wang, Y.; Chen, X. H.; Zhang, Y., Quantum Hall effect in black
phosphorus two-dimensional electron system. Nat. Nanotechnol. 2016, 11, 593-7.
(24). Hao, Y.; Wang, L.; Liu, Y.; Chen, H.; Wang, X.; Tan, C.; Nie, S.; Suk, J. W.; Jiang, T.;
Liang, T.; Xiao, J.; Ye, W.; Dean, C. R.; Yakobson, B. I.; McCarty, K. F.; Kim, P.; Hone, J.;
16
Colombo, L.; Ruoff, R. S., Oxygen-activated growth and bandgap tunability of large single-
crystal bilayer graphene. Nat. Nanotechnol. 2016, 11, 426-31.
(25). Chu, T.; Ilatikhameneh, H.; Klimeck, G.; Rahman, R.; Chen, Z., Electrically Tunable
Bandgaps in Bilayer MoS(2). Nano Lett. 2015, 15, 8000-7.
(26). Sze, S. M.; Ng, K. K., Physics of Semiconductor Devices. Wiley: Hoboken, New Jersey,
UAS, 2007; pp 314-315.
(27). Castellanos-Gomez, A.; Vicarelli, L.; Prada, E.; Island, J. O.; Narasimha-Acharya, K. L.;
Blanter, S. I.; Groenendijk, D. J.; Buscema, M.; Steele, G. A.; Alvarez, J. V.; Zandbergen, H.
W.; Palacios, J. J.; van der Zant, H. S. J., Isolation and characterization of few-layer black
phosphorus. 2D Mater. 2014, 1, 025001.
(28). Wood, J. D.; Wells, S. A.; Jariwala, D.; Chen, K. S.; Cho, E.; Sangwan, V. K.; Liu, X.;
Lauhon, L. J.; Marks, T. J.; Hersam, M. C., Effective passivation of exfoliated black phosphorus
transistors against ambient degradation. Nano Lett. 2014, 14, 6964-70.
(29). Abellan, G.; Wild, S.; Lloret, V.; Scheuschner, N.; Gillen, R.; Mundloch, U.; Maultzsch,
J.; Varela, M.; Hauke, F.; Hirsch, A., Fundamental Insights into the Degradation and
Stabilization of Thin Layer Black Phosphorus. J. Am. Chem. Soc. 2017, 139, 10432-10440.
17
|
1907.09757 | 1 | 1907 | 2019-07-23T08:42:24 | Electrical characteristics of vertical-geometry Schottky junction to magnetic insulator (Ga,Mn)N heteroepitaxially grown on sapphire | [
"physics.app-ph"
] | Schottky barrier height and the ideality factor $\eta$ are established for the first time in the single phase (Ga,Mn)N using a vertical geometry device. The material has been heteroepitaxially grown on commercially available low threading dislocation density GaN:Si template. The observed above 10M$\Omega$ resistances already at room temperature are indicative that a nearly conductive-dislocation-free electrical properties are achieved. The analysis of temperature dependence of the forward bias I-V characteristics in the frame of the thermionic emission model yields Ti-(Ga,Mn)N Schottky barrier height to be slightly lower but close in character to other metal/GaN junctions. However, the large magnitudes of the ideality factor $\eta$>1.5 for T$\leqslant$300K, point to a sizable current blocking in the structure. While it remains to be seen whether it is due to the presence of (Ga,Mn)N barrier or due to other factors which reduce the effective area of the junction, an existence of a substantial serial resistance may hold the key to explain similar observations in other devices of a corresponding structure and technological relevance. | physics.app-ph | physics | Electrical characteristics of vertical-geometry Schottky junction
to magnetic insulator (Ga,Mn)N
heteroepitaxially grown on sapphire
Karolina Kalbarczyk1*, Krzysztof Dybko1,2, Katarzyna Gas1,3, Dariusz Sztenkiel1,
Marek Foltyn1, Magdalena Majewicz1, Piotr Nowicki1, Elżbieta Łusakowska1,
Detlef Hommel3,4, and Maciej Sawicki1*
1Institute of Physics, Polish Academy of Sciences, Aleja Lotnikow 32/46, PL-02668 Warsaw,
2International Research Centre MagTop, Institute of Physics,
Poland
Polish Academy of Sciences, Aleja Lotnikow 32/46, PL-02668 Warsaw, Poland
3Institute of Experimental Physics, University of Wrocław, Pl. Maxa Borna 9, 50-204 Wrocław,
4Polish Center of Technology Development, ul. Stabłowicka 147, 54-066, Wrocław, Poland
*E-mail: [email protected], [email protected]
Poland
Schottky barrier height and the ideality factor are established for the first time in the single
phase (Ga,Mn)N using a vertical geometry device. The material has been heteroepitaxially
grown on commercially available low threading dislocation density GaN:Si template. The
observed above 10 M resistances already at room temperature are indicative that a nearly
conductive-dislocation-free electrical properties are achieved. The analysis of temperature
dependence of the forward bias I-V characteristics in the frame of the thermionic emission
model yields Ti-(Ga,Mn)N Schottky barrier height to be slightly lower but close in character to
other metal/GaN junctions. However, the large magnitudes of the ideality factor > 1.5 for
T 300 K, point to a sizable current blocking in the structure. While it remains to be seen
whether it is due to the presence of (Ga,Mn)N barrier or due to other factors which reduce the
effective area of the junction, an existence of a substantial serial resistance may hold the key to
explain similar observations in other devices of a corresponding structure and technological
relevance.
Keywords:
(Ga,Mn)N, Schottky junction, Schottky barrier height, Dilute magnetic semiconductors, GaN,
Vertical structures
1
1. Introduction
Besides becoming the material system of choice for nowadays optoelectronic [1] and
high-power applications [2], GaN-based and related alloys and heterostructures are expected
also to play a major role in the realization of spin-related functionalities based on
semiconductors [3][4][5]. While the search for a technology-viable nitride semiconductor
exhibiting (ferro-)magnetic properties at room temperature is still the subject of active
research, a great deal of knowledge on the underlying physical processes has been
accumulated from the investigation of these system at their relevant temperatures [6][7][8]. In
the line of our research is (Ga,Mn)N -- an emerging ferromagnetic semiconductor whose long
range ferromagnetic ordering has been confirmed at the low end of cryogenic temperatures
[9][10]. This magnetic form of GaN, in which a small percentage of Ga atoms is randomly
substituted by Mn, belongs to increasingly important group of Rashba materials [3]. The
established strong mid-gap Fermi-level pinning [11][12] and the absence of a depletion layer,
in combination with its insulating character [13][14][15][16] and a sizable dielectric strength
of at least 5 MV/cm [17], points at Mn containing GaN as a worthwhile insulating buffer
material for applications in (high power) nitride devices. By the same token this
ferromagnetic insulator appears to be an ideal building block for nitride based spin harnessing
concepts, which could further stimulate the development of all-nitride low power information
processing and wireless communication technologies based on spin waves, the so called
magnonics [18]. On the other hand, in the field of spintronics, an experimental verification of
theoretical predictions suggesting large spin filtering capabilities of spin filters [19] and
resonant tunneling devices [20] with (Ga,Mn)N as the barrier material is still lacking. The
sought functionality is provided here by the spin-splitting of (Ga,Mn)N conduction band that
creates the spin-dependent barriers, which should give rise to a highly spin-polarized current.
However, the observation of the last effect proved futile in the GaN based structures [21]
due to the presence of a large number of conducting dislocations [22][23], which provide a
short-circuiting path across the (Ga,Mn)N barrier layer. Concerning the nitride structures
grown on GaN templates, as in the case of the present study, pure screw threading dislocations
had been identified as the primary source of this detrimental leakage [24]. The presence of the
same conductive dislocations hinders also the development of such important concepts as
all-nitride high electron mobility transistor (HEMT), or heterostructure field-effect transistor
(HFET) [25][26][27][28].
In this study we experimentally investigate and numerically quantify the temperature
dependence of the Schottky barrier height formed at a junction of Ti/Au metallic contact to
(Ga,Mn)N. The investigated junction is prepared in the vertical geometry. The junction itself
has been defined by an opening in the insulating Al2O3 layer which constitutes also the base for
a macroscopically large top electrical contact. The observed very large resistance of the
junction indicates that a nearly conductive-dislocation-free electrical properties are achieved.
We show that the devices fabricated by this approach can be characterized by typical
magnitudes of the metal-nitride semiconductor barrier heights, however some deviations from
the thermionic emission model are clearly observed. Since the structural resemblance of our
2
devices to HEMT or similar structures, the findings reported here may prove useful for that ever
so much important field of material science.
2. Experimental details
The investigated (Ga,Mn)N layer with Mn concentration of about 5% is grown by
plasma assisted molecular beam epitaxy (MBE) method in a Scienta-Omicron Pro-100 MBE
at 620°C under near stoichiometric conditions [29]. We use a c-plane sapphire based
commercially available low threading dislocation density conductive n-type (0001) GaN:Si
template, which contains, generally termed, SiNx mask [30]. According to the technical
specification the Si concentration in the n-type GaN buffer layer amounts to 3 × 1018 cm-3, so
the buffer layer serves also as the conducting bottom contact in the structure. This highly
advantageous feature of the substrate for the design of vertical devices, has got a certain
drawback as it precludes a direct electrical testing of the (Ga,Mn)N layer. However, much can
be learnt from the magnetic investigation. So, despite that the electrical studies presented here
are undertaken only down to 150 K, we perform the magnetic characterization at low
temperatures to infer the electrical properties of the (Ga,Mn)N layer.
(Ga,Mn)N grown at identical conditions as in ref. [29] undergoes the Curie transition
between 5 and 8 K. This so far exclusively low temperature feature is provided by
ferromagnetic superexchange interaction between Mn3+ ions in GaN [31][32][33][34]. This
prerequisite condition for homogenous ferromagnetism in diluted (Ga,Mn)N [35] has been
confirmed by X-ray absorption spectra near the Mn K absorption edge reported in refs.
[36][37], and, most importantly here, in ref. [29], where also other relevant characterization
details are given. Therefore, an observation of (ferro-)magnetic characteristics in (Ga,Mn)N as
those reported in e.g. refs. [17][36] can be taken as an indirect indication of the dominance of
the Mn3+ state in the material, largely reducing a necessity for a large scale facility
investigations. In such samples, an around the mid-gap position of the Fermi level is expected
[12] , what results in a highly resistive or even the insulating character. It should be strongly
underlined here that the magnetic characterization appears as the easy available and a
relatively reliable assessment tool to evaluate the transport properties of (Ga,Mn)N deposited
on conducting substrates.
The magnetic characterization has been performed using a Quantum Design MPMS XL
Superconducting Quantum Interference Device (SQUID) magnetometer. Because of a low Mn
content and a small thickness of the layer deposited on a bulky sapphire substrate, the
accurate determination of magnetization as a function of temperature T and magnetic field H
has required an extension of the previous experimental methodology [38][39] to an active, in
situ, compensation of the dominating flux of the sapphire [40].
Figure 1 summarizes the main magnetic features of the layer studied here. The
vanishing of the remnant moment marks the Curie transition temperature TC 8 K, well in the
accordance with the data given in ref. [29]. The low-T ferromagnetic state of the layer is
further corroborated by a well-developed hysteresis curve at T = 1.8 K, presented in the inset.
Importantly, the TRM signal trails the origin at T > TC and together with the featureless, flat
3
response at 300 K (inset to Fig. 1) decisively proves the absence of ferromagnetic Mn-rich
precipitates or nano-inclusions. Combined, these findings indicate: (i) a homogenous
magnetic constitution of the material, (ii) the superexchange mechanism at work, (iii) the
Fermi level pinning at the mid-gap Mn3+ level, and consequently its highly insulating
character. Other works also show that Mn+3 or Mn2+ ions create relatively deep level defects
in the band gap of GaN [41]. Importantly, it should be added that the established a very good
command over the MBE growth allows to produce on demand high structural quality
(Ga,Mn)N with x up to 10% [10][29]. This is in contrast to other transition metals, such as Fe
[42] and Cr [43], which tend to aggregate into metal-rich nanocrystals already for x > 0.5% [44]
[45], what results in destroying their carrier trapping capabilities [44].
Such heteroepitaxially grown layers exhibit relatively smooth surface morphology
(rms 2 nm, Fig. 2a) with clearly resolved monolayer steps, as indicated by atomic force
microscopy (AFM) studies. These terraces exemplified in panels b and c of Fig. 2 are
separated by approximately 3 Å high steps. Such a step height compares reasonably well with
the 2.6 Å step height expected for one (0001) GaN monolayer.
Fig. 1. Temperature dependence of the remnant moment (TRM) of the layer investigated here
(red bullets). Prior to the TRM measurement the sample had been cooled down at 1 kOe to the
base temperature and the field was quenched. The vanishing point of the TRM signal indicates
the Curie transition temperature, here about 8 K. Inset: low field hystereses curves at 1.8 and
300 K. The magnetic flux of the sapphire substrate has been removed using in situ active
compensation method [40].
4
Fig. 2. Atomic force microscopy images of the top surface of the as grown sample, i.e. the
(Ga,Mn)N layer. (a) 5 × 5 m2 birds eye view. The root mean square of this image is ~2 nm. (b)
Zoom on one of the hillocks, and (c) the linear surface scan documenting well developed
terraces separated by approximately 3 Å high steps.
In these growth conditions dislocations containing a screw component contribute to the
formation of spiral hillocks [46][47], clearly resolved in our sample by the AFM technique. As
shown by Hsu et al. [48], the hillocks' tops on the surface of GaN layers are the only places
which conduct the current, so we can estimate from Fig. 2a that the density of the conductive
dislocations in our structure does not exceed 108 cm-2. This is a rather typical density of screw
and mixed dislocations in good heteroepitaxial GaN structures.
Two contradicting requirements have to be fulfilled by the investigated device to facilitate
the studies. On the one hand side, the junction area should be as small as possible to minimize
the number of conductive dislocations, but on the other it should be large enough to permit a
relatively easy electrical contact placing. In order to satisfy both conditions we employ a
concept elaborated previously by some of the present authors to block the conductive
dislocations in vertical heteroepitaxial GaN/(Ga,Mn)N structures by covering their top surface
by insulating (dielectric) oxide [17]. As established previously atomic layer deposition (ALD)
5
deposited Al2O3, as well as HfO2, makes excellent insulating layers on GaN, withstanding up to
at least 3 MV/cm at 77 K [49][50]. This simple and very effective solution allowed to
demonstrate the existence and to quantify the effectiveness of the piezoelectro-magnetic
its outstanding
coupling
insulating properties
[13][14][15][16], which constitute the functional basis of our device.
[17], and confirmed
in
(Ga,Mn)N
The schematic drawing of our device is presented in Fig. 3. 50 × 50 m2 junction to
(Ga,Mn)N is defined by the lift-off technique in about 100 nm thick layer of insulating Al2O3,
deposited uniformly on the top (Ga,Mn)N surface by ALD. Next, a macroscopically large
Ti(~5 nm)/Au(~150 nm) contact pad (~200 × 200 m2) is evaporated centrally above the
opening in the Al2O3. The metallization thickness exceeds that of Al2O3 to assure that the
metallic layer connects the top of the (Ga,Mn)N layer with the remaining metallic contact pad
resting on the Al2O3. Here, we underline a double role which (Ga,Mn)N layer plays in this
device. While being the subject of the investigation, its highly insulating character renders the
lateral conductivity inefficient and so precludes shortening of the device by those conductive
dislocations which pierce the (Ga,Mn)N away from the window in Al2O3. Essentially, this
concept is equivalent to an etching a small diameter mesa down to GaN:Si buffer, but
eliminates a reactive ion etching step and a rather cumbersome selective electrical contacting of
the top of a small mesa.
Fig. 3. Schematic representation of the device used in this study. The 50 × 50 m2 vertical
conductive channel is defined by an opening in deposited by the atomic layer deposition
method 100 nm thick insulating Al2O3 layer (brown). The layer serves at the same time as a
base for the top electrical contact. The 30 nm thick (Ga,Mn)N layer (red) is grown by MBE
on a reduced threading dislocation density templated sapphire. This 7 m thick GaN:Si
template provides also the bottom electrical connection to the structure. The thick black
arrow marks the effective current path, since the highly insulating character of (Ga,Mn)N
renders the lateral conductivity inefficient and so precludes shortening of the device by other
conductive dislocations.
6
Three electrical contacts to the bottom n-GaN layer are made during the same Ti/Au
metallization process, and are located in the rim part of the substrate. This about 2 mm wide
edge strip of the substrate is not overgrown by (Ga,Mn)N since it is obscured by a molybdenum
substrate holder during the MBE growth [29]. This approach again eliminates the need for
opening an access to the bottom n-GaN layer by reacting ion etching of the top (Ga,Mn)N one,
further simplifying the fabrication process. Finally, after placing the sample on the typical
dual-in-line integrated circuit socket sample holder, the electrical connection are made by
standard bonding using 15 m Al wire. The electrical measurements are performed in a
standard continuous flow cryostat. The current-voltage (I-V) characteristics of the devices in
the standard two-terminal configuration are obtained using a "differential conductance mode"
of DC Keithley 2182A/6221 tandem (nanovoltmeter/current source).
3. Results and discussion
Fig. 4. Comparison of the temperature dependence of the two terminal resistance of (navy) the
(Au/Ti) / (Ga,Mn)N / GaN:Si vertical device, schematically shown in Fig. 3, and (green) of the
GaN:Si buffer. Note that the latter data are multiplied 20000 times. All Au/Ti contacts are
formed in the same evaporation/lift off process. The measurements are performed at zero bias
condition.
While the two-terminal resistance between any two side contacts to n-GaN remains
between 200-500 in the whole studies temperature range, the resistance through the
(Ga,Mn)N junctions exceeds 10 M already at room temperature. This resistance further
increases on lowering temperature reaching nearly 100 M below 200 K, as presented by navy
bullets in Fig. 4. However, already below some 250 K the resistance starts to saturate and the
experimental noise increases. We assign this behavior to a signature of a presence of some
sparse weakly conducting dislocations which residual parallel conductance starts to determine
7
the junction resistance above some 50 M. Nevertheless, this result constitutes a remarkable
progress in comparison to results obtained for similar structures grown on plain c-sapphire
substrates for which resistances below 1 k were observed [21]. On the other hand, the values
presented in Fig. 4 are rather unprecise, since the larger is the resistance of the device the more
nonlinear it becomes. This is evidenced in Fig. 5, where we collect the current-voltage (I-V)
characteristics of the devices obtained between 150 and 300 K. The figure indicates an
asymmetric I-V characteristic as for a good Schottky junction (SJ). No hysteresis, during
voltage sweeps, even at the lowest T are observed (not shown), what rules out an influence of
deep level defects. We take therefore these findings as a strong indication that the device is
nearly devoid of conductive dislocations and that the electron transport at room temperature is
governed by the intrinsic electrical properties of the metal/(Ga,Mn)N/GaN:Si stack.
In general, it is an interesting finding that in the material in which on average 2500
conductive dislocations are expected in a 50 × 50 m2 contact window, their influence on
electrical transport is barely seen at room temperature. We postulate therefore that some of the
dislocations lose their conductive character when they are about to propagate through Mn
enriched GaN. And it could be Mn, owing to its mid-gap position of the 2+/3+ level, which
would passivate conductive dislocations in the GaN host, an effect similar to that brought about
by interstitial O, as already observed experimentally [51] and accounted numerically [52]. The
thickness of (Ga,Mn)N in our test structure is rather a moderate one (30 nm), it therefore
remains to be seen whether the increase of the defect tolerance in GaN, due to blanking of the
conductive dislocations by Mn doping, could become more effective in thicker or more Mn
concentrated (Ga,Mn)N layers. Perhaps temperature agitated Mn migration processes occurring
both during the growth and/or during the post-growth cooling at the growth chamber play an
important role here [29][53].
Fig. 5. Semi-log plot of I-V characteristics (symbols) at selected temperatures between 150 and
300 K for the studied device. Solid lines: a linear interpolation at forward bias region. Dashed
lines represent expected behavior of a Schottky junction in thermionic emission approach at
151 and 300 K (Eq. 1) with the ideality factor = 1.
8
-0.2-0.10.00.10.20.3-10-9-8-7-610-1010-910-8 I ( A )V ( V ) T = 296 / 250 / 221 / 191 / 151 K 10-610-7 = 1
Following the data collected in Fig. 5 we assess an effective metal-(Ga,Mn)N barrier
height B, and the temperature dependence of the fidelity of the SJ. It is clearly seen that logI-V
plots are linear over a decade wide range of currents and that they shift to the higher voltages on
lowering T. In this case we can apply the time honored thermionic emission model [54], in
which the forward bias I-V in the region V > 3kBT/q (e.g. V > 0.1 V at 300 K), and in the absence
of a series resistance assumes the following form:
𝐼 = 𝐼𝑆 exp (
𝑞𝑉
𝜂𝑘𝐵𝑇
) = 𝐴𝐴∗𝑇2 exp (−
Φ𝐵
𝑘𝐵𝑇
) exp (
𝑞𝑉
𝜂𝑘𝐵𝑇
). (1)
Here, A is the contact area, T is the absolute temperature, q is the electron charge, kB is the
Boltzmann constant, is the ideality factor describing how far the junction slope differs from
an ideal one. We take the theoretical value of the Richardson constant A* 27 Acm-2K-2 --
corresponding to electron effective mass in GaN: m* = 0.22 m0 [55]. The magnitude of the
reverse saturation current IS and are readily obtained from the linear interpolation of semi-log
I-V plots at their relevant ranges. These lines are presented in Fig. 5. According to Eq. 1, is
inversely proportional to the slope of the interpolating line, whereas its intercept at V = 0 yields
IS. Then at any given temperature B can be readily obtained from:
Φ𝐵 = −𝑘𝐵𝑇𝑙𝑛 (
𝐼𝑆
𝐴𝐴∗𝑇2). (2)
The established magnitudes of B and are plotted in Fig. 6. At room temperature B = 0.61 V,
which is less than typically reported for metallic SJ to GaN: 0.7 B 1 V [48][56][57].
However, this value stays very close to that found for Ti/GaN SJ, B = 0.59 V [58], and in our
metallization process Ti is deposited first on (Ga,Mn)N to facilitate good adhesion of the main
Au contact layer. Moreover, the whole B(T) dependency resembles well that established for
Ni/n-GaN SJ [59][60]. For example B(150 K) 0.4 eV in all these cases, despite different
work functions for both transition metals
Fig. 6. Temperature dependence of the effective Schottky barrier height and the ideality factor
established from the forward bias I-V characteristics in thermionic emission approach.
9
1502002503000.30.40.50.60.71.41.51.61.71.81.9 Barrier height, B (eV)Temperature (K) Ideality Factor,
As shown in Fig. 6 the magnitudes of the ideality factor found here are larger than 1. Even
at T 300 K, 1.5, contrary to the prevailing majority of studies, where typically 1 for
nitride-based junctions at elevated temperatures [57][59][61]. Interestingly, there are reports in
which ideality factors of metal / GaN junctions are similar to these found in our study [60][62].
These two findings, temperature dependent magnitude of B and > 1, indicate a fairly
non-ideal nature of the SJ realized in the studied device, or that the transport mechanism across
the whole two-terminal device deviates from the thermionic emission theory. For example alloy
fluctuations inherent to low-temperature III-V alloys [63], or a non-ideal junction interface due
to both intrinsic and extrinsic effects related to preparation or details of metallization were
invoked to give an account for a detrimental lateral B height inhomogeneity [61][64][65][66].
Other possible effects include a presence of edge or screw dislocations [61][63][67], which
introduce high density trap states within the band-gap of the semiconductor what enhances the
tunneling probability [23][48][67][68][69][70].
It is worth to underline the fact that in terms of Eqs. 1 and 2 a requirement for >> 1 to
reproduce forward bias I-V behavior means that for a given B (i.e. for a fixed IS) a smaller
current is flowing at the relevant T through our junction than it would if the junction
transmission had been determined exclusively by the thermionic emission process. Exemplary
dashed lines of slopes corresponding to = 1 and giving the same magnitudes of IS at 296 and
151 K are added to Fig. 5. Following these lines, one can establish that the current measured at
+0.2 V is about 15 and 1100 times less than the expected one for the ideal SJ at 296 and 151 K,
respectively. It can be understood that such a readily increasing series resistance on lowering
temperature is provided in our structure by the deliberately introduced here (Ga,Mn)N barrier.
It remains to be experimentally verified which of the effects mentioned above plays indeed a
decisive role. Such a finding would shed an additional light on physical mechanisms governing
electrical properties of other vertical structures like HEMTs [25][26].
4. Conclusions
Vertical transport device has been fabricated from heteroepitaxially grown (Ga,Mn)N on
commercially available low threading dislocation density GaN:Si template. The investigated
device has been defined lithographically by the opening in the additional insulating dielectric
Al2O3 layer grown by atomic layer deposition on top of (Ga,Mn)N to facilitate electrical
bonding. The observed two terminal resistance in the range of tens of M indicates that a
nearly conductive-dislocation-free electrical properties are achieved. A positive role of Mn in
electrical blanking of the conductive dislocations has been postulated. The analysis of
temperature dependence of the forward bias I-V characteristics in the frame of the thermionic
emission model yields Ti-(Ga,Mn)N Schottky barrier height to be slightly lower but close in
character to other metal/GaN junctions. However, the large magnitudes of the ideality factor,
though not so uncommon in nitride Schottky junctions, point to a sizable blocking of the current
in the structure. It still remains to be verified whether it is due to the presence of the insulating
(Ga,Mn)N or some other factors which reduce the effective area of the junction are in force.
10
Acknowledgments
This study has been supported by the National Science Centre, Poland through FUGA (DEC
- 2014/12/S/ST3/00549) and OPUS (DEC - 2013/09/B/ST3/04175) projects.
References
[1] H. Morkoc , Handbook of Nitride Semiconductors and Devices: Electronic and Optical
Processes in Nitrides, Wiley-VCH Verlag GmbH, Weinheim 63 (1) (2016), 419 -- 425
[2] H. Amano, Y. Baines, E. Beam, M. Borga, T. Bouchet, P.R. Chalker, M. Charles, K.J. Chen, N.
Chowdhury, R. Chu, C. De Santi, M.M. De Souza, S. Decoutere, L. Di Cioccio, B. Eckardt, T.
Egawa, P. Fay, J.J. Freedsman, L. Guido, O. Häberlen, G. Haynes, T. Heckel, D. Hemakumara,
P. Houston, J. Hu, M. Hua, Q. Huang, A. Huang, S. Jiang, H. Kawai, D. Kinzer, M. Kuball, A.
Kumar, K.B. Lee, X. Li, D. Marcon, M. März, R. McCarthy, G. Meneghesso, M. Meneghini,
E. Morvan, A. Nakajima, E.M.S. Narayanan, S. Oliver, T. Palacios, D. Piedra, M. Plissonnier,
R. Reddy, M. Sun, I. Thayne, A. Torres, N. Trivellin, V. Unni, M.J. Uren, M. Van Hove, D.J.
Wallis, J. Wang, J. Xie, S. Yagi, S. Yang, C. Youtsey, R. Yu, E. Zanoni, S. Zeltner, Y. Zhang,
The 2018 GaN power electronics roadmap, Appl. Phys. 51 (2018), 163001.
[3] W. Stefanowicz, R. Adhikari, T. Andrearczyk, B. Faina, M. Sawicki, J.A. Majewski, T. Dietl,
and A. Bonanni, Experimental determination of Rashba spin-orbit coupling in wurtzite
n-GaN:Si, Phys. Rev. B 89 (2014), 205201.
[4] Y. Ando, Spintronics technology and device development, Jpn. J. Appl. Phys, 54 (2015),
070101.
[5] R. Adhikari, M. Matzer, A.T. Martín-Luengo, M.C. Scharber, and A. Bonanni, Rashba
semiconductor as spin Hall material: Experimental demonstration of spin pumping in wurtzite
n-GaN:Si, Phys. Rev. B 94 (2016), 085205.
[6] T. Dietl and H. Ohno, Dilute ferromagnetic semiconductors: Physics and spintronic structures,
Rev. Mod. Phys. 86 (2014), 187.
[7] A. Navarro-Quezada, W. Stefanowicz, T. Li, B. Faina, M. Rovezzi, R.T. Lechner, T. Devillers,
F. d'Acapito, G. Bauer, M. Sawicki, T. Dietl, and A. Bonanni, Embedded magnetic phases in
(Ga,Fe)N: Key role of growth temperature, Phys. Rev. B 81 (2010), 205206.
[8] A. Navarro-Quezada, M. Aiglinger, B. Faina, K. Gas, M. Matzer, T. Li, R. Adhikari, M.
Sawicki, and A. Bonanni, Magnetotransport in phase-separated (Ga,Fe)N with γ′−GayFe4−yN
nanocrystals, Phys. Rev. B 99 (2019), 085201.
[9] M. Sawicki, T. Devillers, S. Gałęski, C. Simserides, S. Dobkowska, B. Faina, A. Grois, A.
Navarro-Quezada, K.N. Trohidou, J.A. Majewski, T. Dietl, and A. Bonanni, Origin of
low-temperature magnetic ordering in Ga1−xMnxN, Phys. Rev. B 85 (2012), 205204.
[10] G. Kunert, S. Dobkowska, T. Li, H. Reuther, C. Kruse, S. Figge, R. Jakiela, A. Bonanni, J.
Grenzer, W. Stefanowicz, J. von Borany, M. Sawicki, T. Dietl and D. Hommel, Ga1−xMnxN
epitaxial films with high magnetization, Appl. Phys. Lett. 101 (2012), 022413.
[11] A. Wolos, M. Palczewska, M. Zajac, J. Gosk, M. Kaminska, A. Twardowski, M. Bockowski, I.
Grzegory, and S. Porowski, Optical and magnetic properties of Mn in bulk GaN, Phys. Rev. B
69 (2004), 115210.
[12] L. Janicki, G. Kunert, M. Sawicki, E. Piskorska-Hommel, K. Gas, R. Jakiela, D. Hommel, and
R. Kudrawiec, Fermi level and bands offsets determination in insulating (Ga,Mn)N/GaN
structures, Sci. Rep. 7 (2017), 41877.
[13] A. Bonanni, M. Sawicki, T. Devillers, W. Stefanowicz, B. Faina, T. Li, T.E. Winkler, D.
Sztenkiel, A. Navarro-Quezada, M. Rovezzi, R. Jakieła, A. Grois, M. Wegscheider, W.
Jantsch, J. Suffczyński, F. D'Acapito, A. Meingast, G. Kothleitner, and T. Dietl, Experimental
probing of exchange interactions between localized spins in the dilute magnetic insulator
11
(Ga,Mn)N, Phys. Rev. B 84 (2011), 035206.
[14] T. Yamamoto, H. Sazawa, N. Nishikawa, M. Kiuchi, T. Ide, M. Shimizu, T. Inoue, and M.
Hata, Reduction in Buffer Leakage Current with Mn-Doped GaN Buffer Layer Grown by
Metal Organic Chemical Vapor Deposition, Jpn. J. Appl. Phys. 52 (2013), 08JN12.
[15] M. Zajac, R. Kucharski, K. Grabianska, A. Gwardys-Bak, A. Puchalski, D. Wasik, E.
Litwin-Staszewska, R. Piotrzkowski, J.Z. Domagala, and M. Bockowski, Basic
ammonothermal growth of Gallium Nitride -- State of the art, challenges, perspectives,
Progress in Crystal Growth and Characterization of Materials 64 (2018), 63.
[16] R. Adhikari, W. Stefanowicz, B. Faina, G. Capuzzo, M. Sawicki, T. Dietl, and A. Bonanni,
Upper bound for the s−d exchange integral in n-(Ga,Mn)N:Si from magnetotransport studies,
Phys. Rev. B 91 (2015), 205204.
[17] D. Sztenkiel, M. Foltyn, G.P. Mazur, R. Adhikari, K. Kosiel, K. Gas, M. Zgirski, R. Kruszka,
R. Jakiela, T. Li, A. Piotrowska, A. Bonanni, M. Sawicki, and T. Dietl, Stretching magnetism
with an electric field in a nitride semiconductor, Nature Commun. 7 (2016), 13232.
[18] A.V. Chumak, V.I. Vasyuchka, A.A. Serga, and B. Hillebrands, Magnon spintronics, Nat. Phys.
11 (2015), 453.
[19] T.S. Santos and J.S. Moodera, Observation of spin filtering with a ferromagnetic EuO tunnel
barrier, Phys. Rev. B 69 (2004), 241203.
[20] A.G. Petukhov, A.N. Chantis, and D.O. Demchenko, Resonant Enhancement of Tunneling
Magnetoresistance in Double-Barrier Magnetic Heterostructures, Phys. Rev. Lett. 89 (2002),
107205.
[21] K. Kalbarczyk, M. Foltyn, M. Grzybowski, W. Stefanowicz, R. Adhikari, T. Li, R. Kruszka, E.
Kaminska, A. Piotrowska, A. Bonanni, T. Dietl, and M. Sawicki, Two-Probe Measurements of
Electron Transport in GaN:Si/(Ga,Mn)N/GaN:Si Spin Filter Structures, Acta Phys. Polon. A
130 (2016), 1196.
[22] P. Kozodoy, J.P. Ibbetson, H. Marchand, P.T. Fini, S. Keller, J.S. Speck, S.P. DenBaars, and
U.K. Mishra, Electrical characterization of GaN p-n junctions with and without threading
dislocations, Appl. Phys. Lett. 73 (1998), 975.
[23] E.G. Brazel, M.A. Chin, and V. Narayanamurti, Direct observation of localized high current
densities in GaN films, Appl. Phys. Lett. 74 (1999), 2367.
[24] J.W.P. Hsu, M.J. Manfra, R.J. Molnar, B. Heying, and J.S. Speck, Direct imaging of
reverse-bias leakage through pure screw dislocations in GaN films grown by molecular beam
epitaxy on GaN templates, Appl. Phys. Lett. 81 (2002), 79.
[25] S. Turuvekere, N. Karumuri, A.A. Rahman, A. Bhattacharya, A. DasGupta, and N. DasGupta,
Gate Leakage Mechanisms in AlGaN/GaN and AlInN/GaN HEMTs: Comparison and
Modeling, IEEE Trans. Electron Devices 60 (2013), 3157.
[26] P.D. Ye, B. Yang, K.K. Ng, J. Bude, G.D. Wilk, S. Halder, and J.C. M. Hwang, GaN
metal-oxide-semiconductor high-electron-mobility-transistor with atomic layer deposited
Al2O3 as gate dielectric, Appl. Phys. Lett. 86 (2005), 63501.
[27] S. Rajan, P. Waltereit, C. Poblenz, S.J. Heikman, D.S. Green, J.S. Speck, and U.K. Mishra,
Power performance of AlGaN-GaN HEMTs grown on SiC by plasma-assisted MBE , IEEE
Electron Device Lett. 25 (2004), 247.
[28] H. Zhang, E.J. Miller, and E.T. Yu, Analysis of leakage current mechanisms in Schottky
contacts to GaN and Al0.25Ga0.75N∕GaN grown by molecular-beam epitaxy, J. Appl. Phys. 99
(2006), 023703.
[29] K. Gas, J.Z. Domagala, R. Jakiela, G. Kunert, P. Dluzewski, E. Piskorska-Hommel, W.
Paszkowicz, D. Sztenkiel, M.J. Winiarski, D. Kowalska, R. Szukiewicz, T. Baraniecki, A.
Miszczuk, D. Hommel, and M. Sawicki, Impact of substrate temperature on magnetic
properties of plasma-assisted molecular beam epitaxy grown (Ga,Mn)N, J. Alloys Compd.
747 (2018), 946.
12
[30] [ ~7 um GaN:Si/in-situ SiNx mask/ GaN nucleation/ sapphire ] 2" substrates available from
Ferdinand Braun Institute, Berlin.
[31] P.W. Anderson, Antiferromagnetism. Theory of Superexchange Interaction, Phys. Rev. 79
(1950), 350.
[32] J. B. Goodenough, An interpretation of the magnetic properties of the perovskite-type mixed
crystals La1−xSrxCoO3−λ, J. Phys. Chem. Solids 6 (1958), 287.
[33] J. Kanamori, Superexchange interaction and symmetry properties of electron orbitals, J. Phys.
Chem. Solids 10 (1959), 87.
[34] J. Blinowski, P. Kacman, and J. A. Majewski, Ferromagnetic superexchange in Cr-based
diluted magnetic semiconductors, Phys. Rev. B 53 (1996), 9524.
[35] S. Stefanowicz, G. Kunert, C. Simserides, J. A. Majewski, W. Stefanowicz, C. Kruse, S. Figge,
Tian Li, R. Jakieła, K. N. Trohidou, A. Bonanni, D. Hommel, M. Sawicki, and T. Dietl, Phase
diagram and critical behavior of the random ferromagnet Ga1−xMnxN, Phys. Rev. B 88 (2013),
081201(R).
[36] A. Bonanni, M. Sawicki, T. Devillers, W. Stefanowicz, B. Faina, T. Li, T.E. Winkler, D.
Sztenkiel, A. Navarro-Quezada, M. Rovezzi, R. Jakiela, A. Grois, M. Wegscheider, W.
Jantsch, J. Suffczynski, F. d'Acapito, A. Meingast, G. Kothleitner, and T. Dietl, Experimental
probing of exchange interactions between localized spins in the dilute magnetic insulator
(Ga,Mn)N, Phys. Rev. B 84 (2011), 035206.
[37] W. Stefanowicz, D. Sztenkiel, B. Faina, A. Grois, M. Rovezzi, T. Devillers, F. d'Acapito, A.
Navarro-Quezada, T. Li, R. Jakieła, M. Sawicki, T. Dietl, and A. Bonanni, Structural and
paramagnetic properties of dilute Ga1−xMnxN, Phys. Rev. B 81 (2010), 235210.
[38] M. Sawicki, W. Stefanowicz, and A. Ney, Sensitive SQUID magnetometry for studying
nanomagnetism, Semicond. Sci. Technol. 26 (2011), 064006.
[39] L.M.C. Pereira, Experimentally evaluating the origin of dilute magnetism in nanomaterials, J.
Phys. D: Appl. Phys. 50 (2017), 393002.
[40] K. Gas and M. Sawicki, In situ compensation method for high-precision and high-sensitivity
(2019),
Technol.
Sci.
integral
https://doi.org/10.1088/1361-6501/ab1b03.
magnetometry,
Meas.
[41] P.V. Wadekar, C.W. Chang, Y.J. Zheng, S.S. Guo, W.C. Hsieh, C.M. Cheng, M.H. Ma, W.C.
Lai, J.K. Sheu, Q.Y. Chen, L.W. Tu, Mn valence state mediated room temperature
ferromagnetism in nonpolar Mn doped GaN, Appl. Surf. Sci., 473 (2019), 693.
[42] S. Heikman, S. Keller, S.P. DenBaars, and U.K Mishra, Growth of Fe doped semi-insulating
GaN by metalorganic chemical vapor deposition, Appl. Phys. Lett. 32 (2002), 439 -- 441.
[43] F. Mei, Z. Yang, J. Xu, L. Wu, T. Wu, D. Zhang, Growth and characterization of Cr-doped
semi-insulating GaN templates prepared by radio-frequency plasma-assisted molecular beam
epitaxy, J. Cryst. Growth 353 (2012), 162 -- 167.
[44] A. Bonanni, M. Kiecana, C. Simbrunner, T. Li, M. Sawicki, M. Wegscheider, M. Quast, H.
Przybylinska, A. Navarro-Quezada, R. Jakiela, A. Wolos, W. Jantsch, and T. Dietl,
Paramagnetic GaN:Fe and ferromagnetic (Ga,Fe)N: The relationship between structural,
electronic, and magnetic properties, Phys. Rev. B 75 (2007), 125210.
[45] L. Gu, S.Y. Wu, H.X. Liu, R.K. Singh, N. Newman, D.J. Smith, Characterization of Al(Cr)N
and Ga(Cr)N dilute magnetic semiconductors, J. Magn. Magn. Mater. 290 (2005), 1395 -- 1397.
[46] J.W.P. Hsu, M.J. Manfra, D.V. Lang, K.W. Balwin, L.N. Pfeiffer, and R.J. Molnar, Surface
morphology and electronic properties of dislocations in AlGaN/GaN heterostructures, J.
13
Electron. Mater. 30 (2001), 110.
[47] B. Heying, E. J. Tarsa, C. R. Elsass, P. Fini, S. P. DenBaars, and J. S. Speck, Dislocation
mediated surface morphology of GaN , Appl. Phys. Lett. 85 (1999), 6470.
[48] J.W.P. Hsu, M.J. Manfra, D.V. Lang, S. Richter, S.N.G. Chu, A. M. Sergent, R.N. Kleiman,
and L.N. Pfeiffer, and R.J. Molnar, Inhomogeneous spatial distribution of reverse bias leakage
in GaN Schottky diodes, Appl. Phys. Lett. 78 (2001), 1685.
[49] S. Gierałtowska, D. Sztenkiel, E. Guziewicz, M. Godlewski, G. Łuka, B.S. Witkowski, Ł.
Wachnicki, E. Łusakowska, T. Dietl, and M. Sawicki, Properties and Characterization of ALD
Grown Dielectric Oxides for MIS Structures, Acta. Phys. Polon. A, 119 (2011), 692.
[50] A. Taube, S. Gierałtowska, T. Gutt, T. Małachowski, I. Pasternak, T. Wojciechowski, W.
Rzodkiewicz, M. Sawicki, and A. Piotrowska, Electronic properties of thin HfO2 films
fabricated by atomic layer deposition on 4H-SiC, Acta. Phys. Polon. A, 119 (2011), 696.
[51] A. Polyakov, N. Smirnov, M.-W. Ha, C.-K. Hahn, E.A. Kozhukhova, A.V. Govorkov, R.V.
Ryzhuk, N.I. Kargin, H.-S. Cho, and I.-H. Lee, Effects of annealing in oxygen on electrical
properties of AlGaN/GaN heterostructures grown on Si, J. Alloys Compd. 575 (2013), 17.
[52] S. Christenson, W. Xie, Y.-Y. Sun, and S.B. Zhang, Kinetic path towards the passivation of
threading dislocations in GaN by oxygen impurities, Phys. Rev. B 95 (2017)., 121201(R)
[53] R. Jakiela, K. Gas, M. Sawicki, and A. Barcz, Diffusion of Mn in gallium nitride: Experiment
and modeling, J. Alloys Compd. 771 (2019), 215.
[54] E. H. Rhoderick and R. H. Williams, Metal-Semiconductor Contacts 2nd ed. (Clarendon,
Oxford, 1988).
[55] A.M. Witowski, K. Pakuła, J.M. Baranowski, M.L. Sadowski, and P. Wyder, Electron
effective mass in hexagonal GaN , Appl. Phys. Lett. 75 (1999), 4154.
[56] J.D. Guo, M.S. Feng, R.J. Guo, F.M. Pan, and C.Y. Chang, Study of Schottky barriers on
n-type GaN grown by low-pressure metalorganic chemical vapor deposition, Appl. Phys. Lett.
67 (1995), 2657.
[57] F. Iucolano, F. Roccaforte, F. Giannazzo, and V. Raineri, Barrier inhomogeneity and electrical
properties of Pt∕GaN Schottky contacts, J. Appl. Phys. 102 (2007), 113701.
[58] S.C. Binari, H.B. Dietrich, G. Kelner, L.B. Rowland, K. Doverspike, and D.K. Gaskill,
Electrical characterisation of Ti Schottky barriers on n-type GaN, Electron. Lett. 30 (1994),
909.
[59] A. Chatterjee, S.K. Khamari, V.K. Dixit, S.M. Oak, and T.K. Sharma, Dislocation-assisted
tunnelling of charge carriers across the Schottky barrier on the hydride vapour phase epitaxy
grown GaN, J. Appl. Phys. 118 (2015), 175703.
[60] A. Kumar, S. Vinayak, and R. Singh, Micro-structural and temperature dependent electrical
characterization of Ni/GaN Schottky barrier diodes, Curr. Appl. Phys. 13 (2013), 1137.
[61] A.R. Arehart, B. Moran, J.S. Speck, U.K. Mishra, and S.P. DenBaars, Effect of threading
dislocation density on Ni∕n-GaN Schottky diode I-V characteristics, J. Appl. Phys. 100 (2006),
023709.
[62] R. Adari, B. Sarkar, T. Patil, D. Banerjee, P. Suggisetti, S. Ganguly and D. Saha, Temperature
Dependent Characteristics of Fe/n-GaN Schottky Diodes, DOI: 10.7567/SSDM.2011.P-6-5.
[63] M.A. Laurent, G. Gupta, D.J. Suntrup, S.P. DenBaars, and U.K. Mishra, Barrier height
inhomogeneity and its impact on (Al,In,Ga)N Schottky diodes, J. Appl. Phys. 119 (2016),
064501.
[64] J.H. Werner and H.H. Guttler, Barrier inhomogeneities at Schottky contacts, J. Appl. Phys. 69
(1991), 1522.
14
[65] Y. Zhou, D. Wang, C. Ahyi, C.-C. Tin, J. Williams, M. Park, N.M. Williams, A. Hanser, and
E.A. Preble, Temperature-dependent electrical characteristics of bulk GaN Schottky rectifier,
J. Appl. Phys. 101 (2007), 024506.
[66] N. Yildirim, K. Ejderha, and A. Turut, On temperature-dependent experimental I-V and C-V
data of Ni/n-GaN Schottky contacts, J. Appl. Phys. 108 (2010), 114506.
[67] Y.-Y. Wong, E. Y. Chang, T.-H. Yang, J.-R. Chang, J.-T. Ku, M. K. Hudait, W.-C. Chou, M.
Chen, and K.-L. Lin, The Roles of Threading Dislocations on Electrical Properties of
AlGaN/GaN Heterostructure Grown by MBE, J. Electrochem. Soc. 157 (2010), H746.
[68] L.S. Yu, Q.Z. Liu, Q.J. Xing, D.J. Qiao, S.S. Lau, and J. Redwing, The role of the tunneling
component in the current -- voltage characteristics of metal-GaN Schottky diodes, J. Appl. Phys.
84 (1998), 2099.
[69] J.C. Carrano, T. Li, P.A. Grudowski, C.J. Eiting, R.D. Dupuis, and J.C. Campbell, Current
transport mechanisms in GaN-based metal -- semiconductor -- metal photodetectors, Appl. Phys.
Lett. 72 (1998), 542.
[70] J.E. Northrup, Screw dislocations in GaN: The Ga-filled core model, Appl. Phys. Lett. 78
(2001), 2288.
15
|
1901.03126 | 1 | 1901 | 2019-01-10T12:37:09 | Guided acoustic wave sensors for liquid environments | [
"physics.app-ph"
] | Surface acoustic wave (SAW) based sensors for applications to gaseous environments have been widely investigated since the last 1970s. More recently, the SAW-based sensors focus has shifted towards liquid-phase sensing applications: the SAW sensor contacts directly the solution to be tested and can be utilized for characterizing physical and chemical properties of liquids, as well as for biochemical sensor applications. The design of liquid phase sensors requires the selection of several parameters, such as the acoustic wave polarizations (i.e., elliptical, longitudinal and shear horizontal), the wave-guiding medium composition (i.e., homogeneous or non-homogeneous half-spaces, finite thickness plates or composite suspended membranes), the substrate material type and its crystallographic orientation. The paper provides an overview of different types of SAW sensors suitable for application to liquid environments, and intents to direct the attention of the designers to combinations of materials, waves nature and electrode structures that affect the sensor performances. | physics.app-ph | physics | Guided acoustic wave sensors for liquid environments
C. Caliendo*1, M. Hamidullah1,2
1Institute for Photonics and Nanotechnologies, IFN-CNR, Via Cineto Romano 42,
00156 Rome, Italy
2Department of Information Engineering, Electronics and Telecommunications,
University of Rome La Sapienza, 00156 Rome, Italy
*Corresponding author; email: [email protected]
ABSTRACT. Surface acoustic wave (SAW) based sensors for applications to gaseous
environments have been widely investigated since the last 1970s. More recently, the SAW-based
sensors focus has shifted towards liquid-phase sensing applications: the SAW sensor contacts
directly the solution to be tested and can be utilized for characterizing physical and chemical
properties of liquids, as well as for biochemical sensor applications. The design of liquid phase
sensors requires the selection of several parameters, such as the acoustic wave polarizations (i.e.,
elliptical, longitudinal and shear horizontal), the wave-guiding medium composition (i.e.,
homogeneous or non-homogeneous half-spaces, finite thickness plates or composite suspended
membranes), the substrate material type and its crystallographic orientation. The paper provides an
overview of different types of SAW sensors suitable for application to liquid environments, and
intents to direct the attention of the designers to combinations of materials, waves nature and
electrode structures that affect the sensor performances.
1. Introduction
The basic structure of a Surface Acoustic Wave (SAW) sensor includes a piezoelectric substrate
and a pair of interdigitated transducers (IDTs) photolithographically patterned onto the free surface
of the piezoelectric wafer [1]. The piezoelectric substrate can be a bulk single crystal substrate (such
as quartz, LiNbO3 or LiTaO3, to name just a few), with thickness h larger than of the acoustic
wavelength λ (h >> λ), or a thin film (such as ZnO or AlN) with thickness h < λ, grown onto a non-
piezoelectric bulk substrate (such as silicon, sapphire, glass, or diamond, to cite just a few). In SAW
devices, the acoustic wave travels at the surface of the propagating medium and its energy is
confined within one wavelength of depth; it follows that the SAW properties (wave velocity and
amplitude) are highly affected by any physical and chemical changes that occur at the surface of the
propagating medium and/or at the adjacent medium, when the SAW device interacts with an
external environmental stimuli (such as humidity, temperature, and pressure, to cite just a few). In
the presence of a liquid environment, the wave properties can be perturbed by the changes of the
1
electrical (conductivity and permittivity) and mechanical (mass density and viscosity) properties of
the liquid contacting the sensor surface, or by the anchorage of a mass onto the sensor surface.
The environment can interact with the SAW directly or by means of a thin sensing membrane that,
for some specific applications, covers the wave propagation path between the two IDTs and is in
direct contact with the environment to be tested. If the membrane is an insulating material, it can
cover the entire SAW device surface, including the IDTs, while if it is conductive, it is positioned in
between the two IDTs. As a consequence, part of the SAW energy is distributed into the sensing
membrane and any change in its physical properties affects either or both the wave velocity and
propagation loss, giving rise to a detectable output signal (a frequency and/or insertion loss shift)
that represents the sensor response. The affinity of the membrane towards a specific target analyte is
a fundamental prerequisite as it can drive the sensor selectivity towards a specific application. In
gas sensing applications, the sensing membrane can be a thin Pd film [2, 3], a thin lead
phthalocyanine (PbPc) film [4] a graphene-like nano-sheet [5], a calixarene layer [6] or a
polyethynyl-fluorenol layer [7], to cite just few examples, to detect H2, NO2, carbon monoxide,
organic vapors or simply the relative humidity of the surrounding environment. In liquid phase
sensing applications, the membrane can be poly(isobutylene) (PIB), poly(epichlorohydrin) (PECH),
or poly(ethyl acrylate) (PEA) to test toluene, xylenes, and ethyl benzene solutions [8], or
polysiloxane film containing acidic functional groups for detection of organic amines in aqueous
phase [9], or macrocyclic calixarenes for the detection of organic pollutants in drinking water [10].
The SAW sensors described in [11] were fabricated and derivatized with a rabbit polyclonal IgG
antibody, which selectively binds to E. coli O157:H7. A dual channel SAW biosensor for the
simultaneous detection of Legionella and Escherichia coli was fabricated using a novel protocol of
coating bacteria on the sensor surface prior to addition of the antibody [12]. Reference [13] shows
an overview of 20 years-worldwide developments in the field of SAW-based biosensors for the
detection of biorelevant molecules in liquid media.
Whatever the sensing membrane is, it is important to underline that the design of the device can
significantly affect the performances of the sensor since its sensitivity is also dependant on their
wave-type and polarization, on the electroacoustic coupling configuration, on the materials
thickness and crystallographic orientation.
SAW devices are manufactured with semiconductor integrated circuits technologies that include
the metal and piezoelectric layers deposition techniques (such as sputtering, pulsed laser deposition
or thermal evaporation), optical or electronic lithography, and lift-off process to pattern the IDTs
and the sensing membrane. The selection of the substrate (the material type, crystallographic cut
and thickness), the IDTs design (metal fingers width and spacing, number of fingers pairs, aperture,
IDTs centre-to-centre spacing, to cite just a few), as well as the selection of the electroacoustic
coupling configuration affect the characteristics of the electroacoustic devices (such as the operating
frequency, quality factor, insertion loss) in such a way as to enhance the sensitivity toward the
environmental parameters changes, regardless of the type of the adopted sensing membrane.
The SAW sensors are implemented on piezoelectric substrates showing high electroacoustic
coupling, such as ZnO, LiNbO3 and LiTaO3, that ensure a high sensor sensitivity, sometimes at the
price of a moderate temperature stability. It can happen that the thermal gradient can lead to
frequency shifts which are comparable to the sensor response to the measurand. The selection of
2
temperature stable cuts of the piezoelectric substrates, such as the quartz ST-cut, is one way to
improve the temperature-frequency stability of the SAW sensors. A dual sensors configuration, that
includes the active sensor and a reference sensor [14-16], can be designed to compensate common
mode influences such as humidity or aging, other than temperature, that affect both the active
device, coated, for example, with a selectively adsorbing membrane, and the uncoated one that acts
as a reference device. Alternatively, a temperature compensated configuration, i.e. a multi-layered
structure including opposite-sign temperature coefficients of delay (TCD) materials, with the proper
thicknesses [17], can be designed to cancel the device spurious responses due to the thermal drift
[18, 19].
The present paper gives a survey of the electroacoustic devices based on the propagation of
elastic waves travelling at the plane surface of half-spaces and within finite thickness plates; the
common characteristic of these waves is the ability to travel in a medium contacting a liquid
environment, without suffering excessive energy loss. The paper is structured as follows: in Section
2 we consider features of the pseudo surface acoustic waves (PSAWs) and high velocity PSAWs
(HVPSAWs), showing in-plane polarization and travelling at the surface of piezoelectric substrates;
results of numerical calculations and finite element modelling of the PSAW and HVPSAW-based
sensors for liquid environments are presented. In Section 3 the features of Love wave-based sensors
are studied for different combinations of substrate and guiding layer materials, with the varying
layer thickness. In Section 4 the features of shear horizontal acoustic plate modes (SHAPMs)-based
sensors are studied for different plate thicknesses, material types and modes order. In Section 5
Lamb wave-based sensors are studied, including the quasi-longitudinally polarized fundamental and
higher order modes, the fundamental symmetric mode and the quasi Scholte wave. Section 6
concludes our paper.
2. Pseudo SAW and high velocity PSAW sensors
Surface acoustic waves (SAWs) are elastic waves that travel at the free surface of a half-space and
are confined within one acoustic wavelength in depth. The physical motion of the SAW is
mechanically associated with an elliptical displacement of the surface that is characterized by one
out-of-plane particle displacement component, U3, and two in-plane components, U2 and U1, normal
and parallel to the wave vector 𝑘 = 2𝜋
𝜆⁄ .When travelling in a piezoelectric medium, the SAW
strain field is accompanied by a travelling electric potential wave: the linear electromechanical
coupling effect in piezoelectric materials enables the inter-conversion between electrical and
acoustic signals. As a consequence, the excitation and detection of SAWs, as well as of any types of
plate waves on piezoelectric substrates, is accomplished by means of metal interdigitated electrodes
(IDTs) as first reported by White and Voltmer [20] Figure 2.1 shows the schematic of a single-
electrode-type IDT with uniform fingers spacing and constant overlap: several metal strips are
aligned and connected to the bus bars with a periodicity corresponding to the . The fingers width
and spacing are equal to /4; the total length of the IDT is L = (N -- ¼), being N the number of
finger pairs; W is the electrodes overlapping (the IDT aperture); d =W/ is the IDT directivity. The
frequency of the propagating wave is f = v/, where v is the velocity of sound in the half-space
3
material. The SAWs propagate in both directions away from the IDT, along the propagation axis:
thus, the inherent loss of the two IDTs is equal to 3 dB.
Figure 2.1: a) schematic view of the fundamental structure of a SAW device including a launching
and a receiving transducer; b) interdigital transducer.
When a RF voltage is applied between the two bus bars of the transmitting IDT, a periodic strain is
generated. As a travelling electric field is associated to the SAW, the metal strips of the receiving
IDT will detect the SAW-induced charges. By changing the fingers overlapping, the IDT aperture,
N and the spacing of the metallic fingers, the SAW bandwidth and the directivity can be changed.
A complete description of the theory and modelling of IDTs employed for SAW excitation and
detection can be found in Reference [21]. The electromechanical coupling factor K2 is a measure of
the electric-to-acoustic energy conversion efficiency; it is represented by the fractional change in
wave velocity due to surface metallization, 𝐾2 = 2 · [
(𝑣𝑓𝑟𝑒𝑒 − 𝑣𝑚𝑒𝑡)
⁄
𝑣𝑓𝑟𝑒𝑒
] , being vfree the free
surface wave velocity and vmet the velocity on the metallized surface. The K2 values of the SAWs
travelling along common piezoelectric substrates are equal to 0.16, 0.75, 4.8 and 5.31% for ST-
quartz, 112°x-y LiTaO3, zx- and 128°y-x LiNbO3, respectively [22]. Due to the anisotropic
properties of the piezoelectric materials, the K2 depends on the substrate crystallographic cut and
wave propagation direction. The electrical impedance of the IDT depends on several factors such as
K2, the dielectric permittivity of the substrate, and the geometry of the IDT; it must match as closely
as possible that of external components (50 ) and be resistive. IDT apertures of less than
approximately 30 are inadvisable, as the transducer can diffract the acoustic beam resulting in an
acoustic beam considerably diverging before reaching the output IDT [23]. The metal used to
fabricate the IDTs is generally a low-resistance material highly adhesive to the substrate surface,
with thickness around 0.1 m, in order to make the transduction process more efficient; Al, Au, Cu,
and Mo are few examples of commonly used metals: some of these materials require the presence
of a thin adhesive inter-layer (Ti or Cr) to improve the adhesion to the substrate or to avoid the
diffusion of the metal into the substrate. The cost is also an important factor that drives the choice
4
of the IDT metal type: Cr/Al appears to be the best choice with a good balance of relatively low
resistivity, low cost, and good surface adhesion. In harsh environment applications (such as high
temperature or corrosive environments), the metal used to fabricate the IDTs should exhibit a high
electrical conductivity, high melting temperature, a good resistance to oxidation and chemical
inertness. Iridium, Rhodium and Platinum are few examples of metals suitable for harsh
environment applications [24-27].
In constrast to SAWs, which are polarized perpendicularly to the surface, the pseudo SAWs
(PSAWs) and high velocity PSAWs (HVPSAWs) are predominantly in-plane polarized, that makes
them suitable for low loss propagation under a liquid environment. The three SAW displacement
components decay exponentially with the depth, while the PSAWs and HVPSAWs have both
decaying and radiating components; the latter component radiates power into the half-space, thus
resulting in an attenuation of the field amplitudes as the wave propagates. If the contribution from
the radiating terms is sufficiently small, these two pseudo waves are observed as in standard SAW
devices. The PSAW usually has the U2 component as the dominant one (U2 >> U1, U3 at the half-
space surface), while the HVPSAW usually has the longitudinal component U1 as the dominant
term (U1 >> U2, U3 at the half-space surface) [28]. For specific crystallographic cuts and wave
propagation directions in the most common piezoelectric substrates, piezoelectrically-active
PSAWs and HVPSAWs travel with minimum propagation loss and, since they are both in-plane
polarized, they are suitable to work in contact with liquid. Three dimensional (3D) eigenfrequency
FEM analysis was performed using COMSOL Multiphysics® Version 5.2 to explore the field shape
of the SAW, PSAW and HVPSAW travelling at the surface of a ST-x quartz substrate (Euler angles
0° 132.75° 0°). Figure 2.2 shows the 3D primitive SAW cell as considered in the analysis: beneath
the single wavelength cell is a perfectly matched layer (PML) at the bottom for capturing losses
related to bulk wave radiation, in order to simulate the half-space. The primitive SAW cell has two
periodic and two continuity boundary conditions applied on the sidewalls. The Al electrodes are 0.1
μm thick, with pitch of p = 5 μm (λ = 20 μm). The IDT fingers width-to-spacing ratio was set to 1.
The base material is ST-x quartz.
5
Figure 2.2: The 3D field profile of the SAW, PSAW and HVPSAW travelling along the ST-x
quartz substrate in air. The colour density bar is representative of the relative particle displacement:
the dark blue represents zero displacement and red the maximum displacement.
From figure 2.2 it can be clearly observed that, unlike what happens for the SAWs, the particle
motion of the PSAW and HVPSAW is contained in the surface plane of the propagating medium
since the shear vertical displacement component is very small for both the two waves.
Attractive properties of PSAWs and HVPSAWs are the high velocity (close to that of the
transverse and longitudinal bulk acoustic wave, respectively), low propagation loss and high
electromechanical coupling coefficient [28].
The sensors based on PSAWs and HVPSAWs can measure the mechanical (mass density and
viscosity) and electrical (conductivity and relative permittivity) property changes of the liquid that
contacts the wave path directly, without covering the sensor surface with any selective film. The
sensor surface is in direct contact with the liquid test bath whose thickness greatly exceeds the
penetration depth of the wave mode excited by the IDTs. The sensor response is caused by the
mechanical and electrical boundary conditions changes resulting from the perturbations the
adjacent medium undergoes. If the wave propagation path is metallized and electrically shorted,
only the liquid mechanical properties will affect the sensor response, as only the particle
displacement component interacts with the adjacent liquid. If the bare acoustic path is in direct
contact with the liquid, the sensor will be sensitive also to the electrical properties of the liquid as
both the wave electrostatic potential and particle displacement interact with the liquid, and two
perturbations occur. The electrical perturbation can be discriminated by detecting differential
signals between two delay lines.
6
The absolute value of the normalized admittance Y vs frequency curves for the three modes
propagating in ST-x quartz, in air and in water, were calculated by a frequency domain study, and
the curves are shown in Figure 2.3. Three peaks are clearly visible when the surface contacts the air
(black curve): they correspond to the SAW, PSAW and HVPSAW whose velocities v = f·λ are
equal to 3167, 5083 and 5751 m/s, respectively. The PSAW has a propagation loss higher than that
of the SAW and HVPSAW [29]. The red curve of figure 2.3 corresponds to ST-quartz half-space
contacting the water. The SAW, which has a large displacement component normal to the substrate,
is almost totally damped by the water as expected; the PSAW shows nearly no damping while the
HVPSAW is affected by a small attenuation. As the vertical displacement component
𝑠𝑢𝑟𝑓
𝑈3
⁄
𝑈1
𝑠𝑢𝑟𝑓
of the HVPSAW, normalized to the U1 at the surface, is about four times that of the PSAW, the
former wave is more dampened by the water than the latter.
Figure 2.3: The absolute value of the normalized admittance Y vs frequency curves of the SAW,
PSAW and HVPSAW travelling at the surface of the ST-x quartz substrate in air and in water.
A two dimensional (2D) COMSOL simulation of the HVPSAW displacement profile inside the ST-
x quartz is plotted in figure 2.4a: the blue and the green curves represent the U1 and U3 HVPSAW
displacement components; the abscissa is the normalized depth (for = 20 µm). Figure 2.4b shows
the 2D representation of the HVPSAW total displacement. The substrate and liquid depths (120
μm) are equal: the ST-x quartz extends from -120 to 0 μm of the abscissa values, while the liquid
half-space extends from 0 to 120 μm abscissa value. The liquid was modelled as a linear isotropic
viscoelastic material with independent elastic constants; the bulk modulus and the dynamic
viscosity were extracted from Reference [30]. When the quartz is contacted by the water, the
surface-normal displacement component of the HVPSAW (the blue line) generates compressional
waves in the liquid phase: the power dissipated leads just to a small attenuation since the wave is
predominantly in-plane polarized.
7
Figure 2.4: a) The HVPSAW displacement profile inside the ST-x quartz (from -120 to 0 μm
abscissa values) and in water (from 0 to 120 μm abscissa value); b) the 2D representation of the
total displacement confined inside the substrate.
In an attempt to design a sensor packaging able to protect the IDTs from the liquid environment
and to confine the measurand in the device sensing area, the electroacoustic device can be
integrated with different types of microchannels. Typically, the test cell that localises the liquid to
the surface of the device consists in a pre-molded polydimethylsiloxane (PDMS) cell [31], or a SU8
cell [32] that is mechanically pressed against the surface of the sensor in order to avoid any contact
between the IDTs and the liquid to be tested. The cell can be positioned in three different
configurations: 1. in between the IDTs, as shown in figure 2.5a, to prevent the presence of liquid on
the IDTs; 2. it can occupy the entire device surface if the IDTs are shielded by a proper layer, as
shown in figure 2.5b, to avoid the conduction through the liquid between electrodes; 3. it can be
positioned in between the IDTs and integrated with two air cavities that protect the IDTs: a vertical
structure shaped as walls is positioned around the IDTs to separate them from the liquid filled
container, as shown in figure 2.5c.
8
Figure 2.5: Schematic of the three sensor/liquid cell configurations: a) the liquid cell is located in
between the IDTs; b) the liquid cell occupies the whole device surface; c) the IDTs are protected by
a vertical structure from the liquid contact.
In the first and third configuration, the cell adds some disturbance effect on the propagation of the
wave due to the mechanical stresses induced on the surface by the packaging, thus resulting in an
additional improvement in damping losses. The second configuration ensures maximum sensitivity,
as the physical size of the liquid cell includes the total wave propagation path, but requires a careful
selection of the type and thickness of the protective surface layer in order to minimize potential
perturbations to the device insertion loss, thermal stability, sensitivity, to cite just a few. The thin
layer can even be a sensitive coating layer that exhibits high sensitivity toward a certain analyte but
low sensitivity toward other species: thus the layer drives the sensor selectivity towards a specific
target analyte. A liquid cell integrated with the sensor during the fabrication process is preferred as
it reduces the devices complexity and enhance their performances.
References [33-35] provide a very useful list of substrates types and crystallographic orientations
along which low-attenuated, strongly-coupled PSAW and HVPSAW travel (such as 64°YX
9
LiNbO3, 36°YX LiTaO3, quartz ST-X, LiNbO3 with Euler angles (90° 90° 36°), (90° 90° 31°)
LiTaO3, and (0° 45° 90°) Li2B4O7). For sake of completeness, the same references list other
information that are fundamental for the design of the device parameters (operating frequency, IDTs
centre-to-centre distance, number of IDTs finger pairs, IDTs aperture, etc.) such as the phase
velocity, propagation
loss, particle displacement components, electric-to-acoustic energy
conversion efficiency, K2, and power flow angle) of the SAW, PSAW and HVPSAW. Moreover,
these references also describe the waves behaviour with increasing the thickness of a metallic layer
covering the wave propagation path: depending on the layer thickness, the wave diffraction into the
bulk can be prevented, as well as the transitions from HVPSAWs to higher order PSAWs modes,
and from PSAWs to the SAWs can be observed. The piezoelectric substrates of the LGX-family
group have been studied in references [36-38] for application to temperature stable, high coupling,
low loss sensors for liquid environments. Several applications of PSAW sensors are presented in
the available literature for the measurement of viscosity, electrical properties and mass loading of
an adjacent liquid. In reference [39] a sensor for liquid viscosity and conductivity measurement is
described that is based on a PSAW dual delay line on 41°-YX LiNbO3 and covered by a SiO2
protective layer. In reference [40] a PSAW sensor is implemented on 36°-YX LiTaO3 substrate for
methanol concentration measurement. In reference [41] the PSAW trapping efficiency of the free,
metallized, and grating paths in YX-36° and YX-42° LiTaO3 and YX-64° LiNbO3 are compared at
fundamental and harmonic frequencies. A high frequency (>500 MHz) PSAW sensor for liquid
environment is designed that show high coupling, low loss, high operating frequency and high
resistance to surface contamination.
In reference [42] the propagation of PSAW along bare 41° and 64°-Y LiNbO3, and 36°-Y LiTaO3
substrates is studied and compared with that along the same substrates covered by a thin sputtered
glass films to increase the resistance to surface contamination. It was demonstrated experimentally
and theoretically that the glass layer does not affect the wave loss but it lowers the temperature
stability. Film thickness to wavelength ratio regions were found which represent a good trade-off
between lowered temperature coefficient of frequency and a preserved high coupling factor.
Sensors of liquid viscosity and mass loading have been demonstrated utilizing the present mode on
36°-YX LiTaO3 [43].
One way to enhance the sensor sensitivity (i.e. the frequency change per unit incremental change
of the measured quantity) is to raise the working frequency: this effect can be obtained by reducing
the size of the IDTs metal strips or utilizing SAW devices based on high velocity acoustic wave
modes. The HVPSAWs are attractive for this potential application as they travel at velocity close to
the longitudinal bulk acoustic wave (BAW) velocity. In reference [44] the propagation of
HVPSAWs in LiNbO3 has been studied in the range of Euler angles (0°, 0° to 90°, 90°): the
corresponding theoretical velocities are between 6700 m/s and 7400 m/s, about twice that of normal
surface waves, but the K2 varies between about 0.14% to 0.5%, much less than that of surface
waves. In reference [45] some experiments show that the HVPSAW phase velocity in (0°, 124°,
50°) quartz reaches 6992 m/s and the propagation attenuation is as low as less than 1·10-4 dB/λ, thus
is suitable for liquid sensing applications. In reference [46] the propagation of HVPSAW along a
semi-insulating Fe-doped GaN films grown on sapphire substrates is experimentally studied and the
10
small propagation attenuation of the mode when travelling at a liquid/solid interface is
demonstrated in glycerol solutions. In reference [47] the propagation loss due to bulk wave
radiation of a HVPSAW is reduced by loading the 36°YX -LiNbO3 substrate with a dielectric
amorphous AlN thin film with a higher velocity than the substrate. The amorphous AlN layer plays
the double role to protect the IDTs patterned onto the substrate surface and to enhance the device
performances. Table 2.1 list some examples of PSAW-based sensors and just two examples related
to the HVPSAW: it is evident that, despite the long-time-recognized suitability of the HVPSAWs
for liquid sensing applications [48] there is a lack of experimental validations.
Table 2.1. Some practical examples of SAW sensors for application to liquid environment test.
Wave type
substrate
layer
application
reference
PSAW
41°-YX LiNbO3
SiO2
liquid
viscosity
[39]
and conductivity
PSAW
36°-YX LiTaO3
--
methanol
[40]
concentration
PSAW
36°-YX LiTaO3
--
liquid
viscosity
[43]
and mass loading
HVPSAW
Fe-doped
--
Viscosity-density
[46]
PSAW
PSAW
GaN/sapphire
36°YX-LiTaO3
product
rabbit anti-goat
IgG
[49]
poly(methyl
methacrylate)
(PMMA) or
cyanoethylcellulose
(CEC)
64° YX-LiNbO3
1,10-
phenanthroline
Heavy
metal
[50]
compounds:
PbNO3
and
PSAW
36° YX-LiTaO3
--
PSAW
36°YX-LiTaO3
--
[51]
CdNO3
tiny particles
mixed with
distilled water
detection and
discrimination of
various detergents
[52]
11
PSAW
36°YX-LiTaO3
parylene thin film
Urea biosensing
[53]
Quartz (0° 124°
50°)
--
HVPSAW
3. Love wave sensors
Water loading
[48]
Love waves are a type of surface acoustic waves characterized by a shear horizontal particle
displacement component U2 dominant over the vertical and longitudinal ones (U2 >> U3, U1). The
propagation of the Love waves is excited and revealed by means of a couple of IDTs, as for the
SAW-based devices. Due to the in-plane polarization, the Love waves, as well as the PSAW and
HVPSAWs, are suitable to travel at a surface contacting a liquid environment. In the most general
sense, Love waves propagate along the surface of a piezoelectric half-space covered by a thin layer:
the substrate is responsible for the excitation of a surface skimming bulk wave (SSBW) that
propagates below the substrate surface; the thin over-layer traps the acoustic energy and slows
down the wave propagation velocity, thus reducing the loss from radiation into the bulk. As a result,
the SSBW is converted into a shear surface wave, the Love wave. Figure 3.1 shows the field profile
of the SSBW and of the Love wave travelling at the surface of a bare ST quartz substrate, and of
the Love wave tavelling at the same substrate covered by a thin SiO2 trapping layer. As it can be
seen, the strain associated to the SSBW penetrates deep within the bulk of the bare quartz substrate;
the strain associated to the Love wave remains close to the surface of the SiO2/quartz substrate.
12
Figure 3.1: The field profile of the SSBW and Love wave travelling along the bare and the SiO2
film covered ST-90°x quartz substrate in air; x1 is the wave propagation direction and x3 is the
normal to the substrate surface.
𝑆𝐻𝐵𝐴𝑊 , as the velocity of the Love wave lies in between the 𝑣𝑠𝑢𝑏
The number of Love modes (LMs) that can propagate in the layer/substrate medium depends on the
layer thickness, but he essential condition for the propagation of the Love waves is that shear bulk
𝑆𝐻𝐵𝐴𝑊 , is larger than the shear bulk wave velocity of
wave velocity (SHBAW) of the halfspace, 𝑣𝑠𝑢𝑏
𝑆𝐻𝐵𝐴𝑊
the layer, 𝑣𝑙𝑎𝑦𝑒𝑟
[54]. Higher order LMs develop at their respective cut-off frequencies, which are related to the
thickness of the layer: they are dispersive as their velocity depends on the layer thickness, other
than on the substrate and the layer's material properties. As the LMs acoustic energy is mostly
concentrated inside the guiding layer, the LM-based devices show good performances in terms of
sensitivity to any disturbance loading the surface of the guiding layer. A comprehensive review of
the Love wave sensors can be found in Reference [55]. As an example, Figure 3.2 shows the phase
velocity dispersion curves of the first five modes (LM1, LM2, LM3, LM4 and LM5) travelling
𝑆𝐻𝐵𝐴𝑊
along the ST 90°-x quartz half-space covered by a SiO2 film. The picture also shows the 𝑣𝑠𝑢𝑏
𝑆𝐻𝐵𝐴𝑊. The velocity values were numerically calculated by using the McGill software
and the 𝑣𝑙𝑎𝑦𝑒𝑟
[56].
𝑆𝐻𝐵𝐴𝑊 and the 𝑣𝑙𝑎𝑦𝑒𝑟
Figure 3.2: The phase velocity vs the layer normalized thickness of the first four LMs travelling in
ST 90°-x quartz/SiO2 substrate.
When the guiding layer is very thin, the velocity of the LM1 tends to the half-space SHBAW
velocity; with increasing the layer thickness, the velocity of both the fundamental and higher order
13
modes asymptotically reaches the layer SHBAW velocity. Love waves vanish if the frequency is
lower than the cut-off frequency.
Figure 3.3a shows the delay line on top of the quartz half-space with the two IDTs located at a
distance equal to 3· (λ = 20 μm); the depth of the substrate is 10. The figure represents the total
displacement 1ns after the electric signal is applied at the transmitting IDT. Figure 3.3b shows the
time evolution of the SSBW and LM1 total displacement propagating at the surface of the ST 90°-x
quartz half-space, bare and covered by a SiO2 trapping layer, 2 μm thick. The total displacement of
the SSBW and LM1 was calculated by 3D COMSOL simulation: the time domain analysis was
carried for 20 ns and the total displacement of the propagating medium was recorded at an interval
of 1 ns. The transmitting bidirectional IDT launches two waves in opposite directions, as indicated
by the arrows in figure 3.3b; the signal applied at the transmitting IDT was a 10 V peak-to-peak
sinusoidal signal at 231 and 217 MHz, for SSBW and LM1. The plot clearly shows that the
acoustical displacement propagates into the depth of the substrate for the SSBW while it is more
confined to the surface for the LM1.
14
Figure 3.3. a) The transmitting and receiving IDTs positioned onto the quartz substrate; b) The time
evolution (1 to 20 ns from the wave excitation) of the total displacement of SSBW and LM1; the
two waves propagate along the ST 90°-x quartz half-space, bare and covered by a SiO2 guiding
layer, 2 μm thick, being λ = 20 μm.
The time and frequency domain response analysis of a delay line based on the SSBW and LM1
propagation was performed by 3D COMSOL simulation assuming a pair of Al IDTs, 0.1 μm thick,
positioned onto the quartz surface (see figure 3.4). The distance between the transmitting and
receiving IDT was set equal to 3·λ (active gap region) and the fingers overlap W =1·λ; the substrate
propagation loss was not accounted in the calculation. The IDT number of finger pairs N was
assumed to be equal to 2 for both the SSBW and LM1 devices. The electrical voltage at the receiver
electrode was recorded for 30 ns in the time domain analysis, as shown in figure 3.4a where the
ratio Vout/Vin of the voltage at the receiver and transmitter IDT in time domain is shown. The
insertion loss of the delay line is calculated by applying a unit impulse at the input IDT: the Fourier
transformation of the device impulse response allowed the calculation of the scattering parameter of
the SSBW and LM1 delay line, S21 = 20·log(Re[Vout/Vin]), shown in figure 3.4 b. For more
information on the scattering parameters, please refer to the textbook [21]. Vin and Vout are the
voltages at the alternate fingers of the transmitting and receiving IDTs respectively, while the
remaining IDTs fingers are grounded.
Figure 3.4 a) Calculated ratio Vout/Vin of the voltage at the receiver and transmitter IDT in time
domain; b) calculated scattering parameter S12 vs frequency, for the SSBW- and LM1-based delay
line onto the ST-90°x quartz substrate, bare and covered by a SiO2 guiding layer, 2 μm thick.
The voltage Vout at the output IDT starts to rise in about 11 and 12.5 ns which corresponds to a
SSBW and LM1 group velocities of 3636 and 3250 m/s, respectively.
The presence of the guiding layer is fundamental to trap the Love wave energy, and only a well-
defined thickness guarantees enhanced device performances (such as minimum delay line insertion
15
loss and maximum gravimetric sensitivity). Figure 3.5 shows the K2 dispersion curve (the black
curve) and the derivative of the group and phase velocity respect to the normalized thickness of the
SiO2 layer,
𝜕𝑣𝑔𝑟
⁄
𝜕 (
and
ℎ𝑆𝑖𝑂2
⁄ )
𝜕𝑣𝑝ℎ
⁄
𝜕 (
, vs
ℎ𝑆𝑖𝑂2
⁄ curves for the LM1 travelling
ℎ𝑆𝑖𝑂2
⁄ )
along the ST 90°-x/SiO2 substrate. The hSiO2/λ = 0.074 corresponds to the maximum K2 = 0.227%;
hSiO2/λ = 0.115 and 0.08 correspond to the maximum mass sensitivity of the phase and group
velocity of the LM1, as the derivative of the 𝑣𝑔𝑟 and 𝑣𝑝ℎ is proportional to the gravimetric
sensitivity 𝑆𝑔𝑟𝑎𝑣 = (∆𝑣
⁄ )/(𝜌 · ℎ𝑆𝑖𝑂2), where ρ and ℎ𝑆𝑖𝑂2 are the layer mass density and
𝑣0
thickness, Δv = v -- v0, v0 and v the wave velocity along the bare and covered half-space [57]. The
K2 values calculated at the abscissa values corresponding to the vgr and vph maximum sensitivity are
quite similar (0.230% and 0.204%), while the expected vgr sensitivity is predicted to be about twice
that of the vph.
Figure 3.5: The K2 and the derivative of the phase and group velocity vs the normalized SiO2 layer
thickness, hSiO2/λ, of the fundamental LM1 travelling along the ST 90°-x/SiO2 substrate.
Figure 3.6 shows the derivative of the 𝑣𝑝ℎ and 𝑣𝑔𝑟 of the first five LMs in ST 90°-x quartz/SiO2
with respect to the layer normalized thickness. As it can be seen, the magnitude of the gravimetric
sensitivity 𝑆𝑔𝑟𝑎𝑣 increases with increasing the layer thickness and reaches a peak after which, with
increasing the guiding layer thickness, it decreases. The peak of the 𝑣𝑝ℎ sensitivity decreases with
increasing the mode order and the highest value corresponds to the LM1 mode; the 𝑣𝑔𝑟 mass
16
sensitivity increases rapidly with the layer thickness, and it can be larger than the former as its peak
can be sharper.
Figure 3.6: The derivative of the phase and group velocity vs the normalized layer thickness for the
first five LMs travelling along the ST 90°-x quartz/SiO2 substrate.
Both the LMs group and phase velocity can represent a sensor response [58]: the phase velocity can
be experimentally estimated by measuring the operating frequency f = 𝑣𝑝ℎ/λ of the sensing device at
the minimum insertion loss of the scattering parameter S12. The group velocity can be estimated by
measuring the group time delay τ = L/vgr of the sensing device at the minimum insertion loss of the
scattering parameter S12 in the time domain, being L the acoustic wave delay path (the IDTs centre-
to-centre distance)
In the most general cases the LMs devices consist of a semi-infinite piezoelectric substrate (for
example 41°YX LiNbO3, 36°YX LiTaO3 and ST-90°X quartz) [59] covered by a thin slowing layer
(for example ZnO, Au, PMMA or SiO2) which traps the propagating wave to the surface of the
substrate. The IDTs can be located only onto the piezoelectric substrate surface, under the
overlayer, and thus they are isolate from the liquid environment. Table 3.1 lists some practical
examples of LM sensors for application to liquid environment.
Table 3.1. Some practical examples of LM sensors for application to liquid environment.
substrate
layer
Application
ST-90°
SiO2
mass sensing in liquids
reference
[60]
17
quartz
ST-90°
PMMA
mass sensing in liquids
[61]
quartz
LiTaO3
SiO2, ZnO, gold, SU-
Comparison
of
electromechanical
coupling
[62]
8, and parylene-C
coefficient, displacement profile and mass
sensitivity
ST-90°
ZnO
liquid viscosity and conductivity
[63]
quartz
ST-90°
PMMA
detection of high molecular weight targets in
[64]
quartz
liquid samples
36°-YX
ZnO
methanol in water
[65]
LiTaO3
36°-YX
ZnO
antibody -- antigen immunoreactions in aqueous
[66]
LiTaO3
solutions
The LMs also propagate along a non-piezoelectric halfspace (such as Si, glass, BN, a-SiC,...)
covered by a piezoelectric layer (such as c-axis tilted ZnO or AlN) [58, 67-69].
For example, when the hexagonal ZnO film has its c-axis parallel to the substrate free surface, it is
effective in the electroacoustic transduction of LMs in glass/ZnO substrate; when the c-axis is tilted
at an angle μ with respect to the normal to the substrate surface, for wave propagation along the
<100> direction, two types of surface modes propagate, the LM with predominant shear horizontal
polarization, and the Rayleigh-like, with a prevailing sagittal polarization. Both modes are coupled
to the electric field via the effective piezoelectric constants of the ZnO film. The LM and the SAW
play different roles onto the same sensing platform: the former is suitable for liquid environment
characterization, while the latter is suitable for mixing and pumping small liquid volumes. LM
sensors implemented on silicon or glass substrate materials offer the great advantage of the sensor's
integration with the surrounding electronic circuits.
For biosensing applications, a sensitive layer can be positioned along the acoustic wave
propagation path: in this case the thickness of the sensing membrane must be properly designed in
order not to perturb the wave propagation characteristics. Figure 3.7 shows the schematic
representation of Love wave sensor. The membrane can cover only the path in between the IDTs or
the entire wave path if the IDTs are buried under the guiding layer. In the latter case, the trapping
layer must satisfy also the requirement of good chemical and mechanical resistance [70] as it has
the additional role of shielding the IDTs.
18
Figure 3.7: The schematic of the Love wave device.
The Love wave trapping layer can be by a thin polymeric film such as polymethylmethacrylate
(PMMA) [71], polyimide, SU-8 or parylene C. In reference [61] the adsorption of h-IgG on the
polymer surface was investigated by using a 1.23 µm thick PMMA guiding layer onto quartz Love
wave device, and finally the potential of the device as a biosensor was investigated by detecting the
binding of anti-IgG.
The LM sensor based on a piezoelectric layer/non piezoelectric substrate has a remarkable
advantage over their counterpart based on piezoelectric half-space/non-piezoelectric layer, as well
as over PSAW and HVPSAW-based sensors: four coupling configurations can be investigated to
enhance the K2 and to take advantage of the protecting role of the guiding layer if the IDTs are
buried under
interface
(substrate/transducer/film, STF) or at the layer free surface (substrate/film/transducer, SFT), with or
without
layer
(substrate/transducer/film/metal, STFM, or substrate/metal/film/transducer, SMFT), as shown in
figure 3.8.
can be positioned
floating metal
it. The
at
the
layer/substrate
of
the
layer
onto
IDTs
a
the
opposite
surface
Figure 3.8. The four coupling configurations for the non-piezoelectric half-space/piezoelectric layer
structure [58].
19
The K2 of the LM device is affected by the mode order, the crystallographic orientation of both the
halfspace and layer, the layer thickness, and also the coupling configuration (the electrical boundary
conditions). As an example, figure 3.9 shows the K2 dispersion curves for the first Love mode
(LM1) in ZnO/glass for the four coupling configurations and different ZnO c-axis tilt angle (from
10° to 90°) [58].
Figure 3.9: The K2 dispersion curves for the first Love mode in ZnO/glass for the four coupling
configurations and different ZnO c-axis tilt angle (from 10° to 90°). The colour of each curve
represents a different c-axis tilt angle.
The SFT (SMFT) configuration reaches the highest K2 values for h/λ ~ 0.3 (0.4) for large tilt angles;
the STF and STFM configurations reaches their maximum K2 value (~1.6 and 1.1 %) for 50° tilt
angle.
As an example, figure 3.10 shows the K2 dispersion curves for the first four Love modes (LM1,
LM2, LM3 and LM4) in 90°-tilted c-axis ZnO/wBN, for the four coupling configurations.
20
Figure 3.10: The K2 dispersion curves for the a) LM1, b) LM2, c) LM3 and d) LM4 modes in 90°-
tilted c-axis ZnO/wBN, for the four coupling configurations [68].
With increasing the Love mode order, ever decreasing K2 values can be reached by the 4 coupling
configurations. The remarkable advantage of the LM-based sensors fabricated onto silicon is the
possibility to integrate the sensor with other devices.
4. Shear Horizontal Acoustic Plate Mode sensors
Shear Horizontal Acoustic Plate Modes (SHAPMs) are waveguide modes that propagate in finite
thickness plates with energy distributed throughout the bulk of the waveguide. The SHAPMs are
shear horizontally polarized (U1, U3 = 0), hence the absence of the out-of-plane displacement
component allows each mode to propagate in contact with a liquid without coupling excessive
amounts of acoustic energy into the liquid. 3D eigenfrequency FEM analysis was performed using
COMSOL Multiphysics® Version 5.2 to explore the field shape of the SH0, SH1, SH2 and SH3
travelling along the along a GaPO4 piezoelectric plate, 150 μm thick, with Euler angles (0° 5° 90°)
and thickness to wavelength ration h/λ = 0.6, as shown in figure 4.1. The GaPO4 materials constants
21
are those provided by Piezocryst Advanced Sensorics GmbH, which is an European GaPO4 wafers
supplier [72, 73].
Figure 4.1: The field profile of the first four SHAPMs in GaPO4 piezoelectric plate in contact with
air.
The SHAPMs-based device employs input and output IDTs to launch and receive the acoustic
wave, like for the SAWs -- based devices. These modes energy is distributed between the two plate
surfaces as for a standing wave in a BAW sensor but the SHAPMs travel along the plate like a
SAW. The continuous exchange of energy between the two plate surfaces allows the signal between
the two IDTs to be affected by any changes of the surrounding environment the opposite plate sides
undergoes. For liquid sensing applications the plate itself can be employed as a physical barrier
between the electronics and the liquid environment to be sensed. The IDTs may be placed onto the
surface opposite to the one in contact with the liquid solution, as shown in figure 4.2: the IDTs are
naturally isolated from the (potentially corrosive) liquid environment without adding any protective
layer to the device surface, as for the SAW-based sensors, thus taking advantage of the entire sensor
surface to maximize the interaction of the wave with the analyte. A metal film can be placed
between the input and output IDTs to cancel any direct electromagnetic feedthrough. A sensing
membrane may be attached to the upper side of the plate that is selectively sensitive to a specific
measurand contained into the test liquid solution contacting the sensor. Any interaction (mechanical
22
and/or electrical) between the measurand and the sensing membrane will cause a shift in the
attenuation and/or velocity of the wave, which represents to the sensor response.
Figure 4.2: The schematic of the SHAPMs sensor including a sensitive membrane [74].
The number of modes that propagate along the plate is dependent on the normalized thickness h/λ
of the plate; the modes are excited at frequency 𝑓𝑛=𝑣𝑛⁄λ where vn is the velocity of the n-th mode
corresponding to the selected h value. As an example, Figure 4.3 shows the phase velocity
dispersion curves of the first six SHAPMs travelling along an y-rotated GaPO4 plate with Euler
angles (0° 1° 90°); the data were obtained by using the McGill software [56]. As it can be seen, the
fundamental mode, SH0, is a low-dispersive symmetric mode that travel at velocity equal to the
transverse BAW velocity. The higher order modes can be symmetric and anti-symmetric: they are
highly dispersive and their velocity asymptotically reaches the shear BAW velocity with increasing
the plate thickness; they have a cut off thickness: below the cut-off frequency, the mode becomes
evanescent, i.e., the wavenumber is imaginary. Higher order modes can reach very high velocity as
near the cut off the slope of the dispersion curves is near to be infinite.
23
Figure 4.3: The phase velocity dispersion curves of the first six SHAPMs in GaPO4 plate (0° 1°
90°) [74].
SHAPMs have maximum displacements that occur on the top and bottom surfaces of the plate, with
sinusoidal variation between the two plate sides. The field profile of the first four SHAPMs (SH0,
SH1, SH2, and SH3) in ZnO (0° 90° 0°) with h/λ=0.5 are shown in figure 4.4. SHAPMs are divided
into symmetric and anti-symmetric modes: for each mode the maximun displacement occurs on the
top and bottom surfaces of the plate, allowing the use of either side of the plate for liquid sensing
applications; the number of zeros is equal to the order of the mode. The fundamental symmetric
mode (SH0 in figure 4.4a) differs from the others in that the acoustic field is uniformly distributed
along the plate depth.
24
Figure 4.4: Cross-sectional displacement profiles for the four lowest-order SHAPMs in ZnO (0° 90°
0°) with normalized thickness h/λ=0.5 [74].
Excitation of shear plate modes, showing dominant shear horizontal polarization, can be
accomplished by tilting the c-axis away from the vertical by an angle α. Pure SHAPMs exist on
90°-x propagating rotated y-cut of trigonal class 32 group crystals, which include the GaPO4 and
the quartz crystals [75, 76], and in x-propagating rotated y-cut hexagonal plates, such as AlN or
ZnO. Figure 4.5 shows the rotated crystallographic system of the piezoelectric plate.
Figure 4.5: The rotated crystallographic system of the piezoelectric plate: x, y, and z represent the
crystallographic axis system; x3 is the plate normal; the propagation direction of the pure SHAPM is
equal to x2 or x1 for trigonal or hexagonal crystals [74].
25
The tilt angle α, as well the plate thickness, affects the phase velocity and hence the K2. As an
example, figures 4.6a-b show the K2 of the two coupling configurations, ST and MST, on ZnO vs
the normalized plate thickness, being the tilt angle α the running parameter: the data were calculated
with McGill software [56]. ST stands for substrate/transducer and is referred to the normal case
where the IDTs are placed on one plate side, while MST stands for metal/substrate/transducer and is
referred to the previous configuration with the opposite surface covered by a floating mass-less,
infinitesimally thin metal layer.
Figure 4.6: The K2 of the a) MST and b) ST coupling configuration on c-axis tilted ZnO plate vs the
normalized plate thickness, being the ZnO c-axis tilted angle α the running parameter.
The viscosity sensitivity of the SHAPM sensors as well as the relative surface displacement (and
particle velocity) increase with increasing mode order [1, 77, 78]. Figure 4.7a, b and c show the
attenuation of the SH1, SH2 and SH3 modes vs the frequency thickness product for a glass plate
immersed in ethylic alcohol (ρ = 790 kg/m3, vl = 1238 m/s, dynamic viscosity = 1.2·10-3 Ns/m2,
kinematic viscosity = 1.52·10-6 m2/s), benzene (ρ = 881 kg/m3, vl = 1117 m/s, dynamic viscosity =
0.65·10-3 Ns/m2, kinematic viscosity =7.38·10-7 m2/s) and kerosene (ρ = 822 kg/m3, vl = 1319 m/s,
dynamic viscosity = 1.5·10-3 Ns/m2, kinematic viscosity =1.82·10-6 m2/s). The modes are sensitive
to the viscosity of the liquids even when the mass density ρ and/or the velocity vl are quite similar.
26
Figure 4.7: The attenuation of the SH1, SH2 and SH3 modes vs the frequency thickness product for
a glass plate immersed in a) ethylic alcohol, b) benzene and c) kerosene.
The sensitivity also increases as the device is thinned: the lower limit of the plate thickness is
limited by production processes and plate fragility. Figure 4.8a and b show the attenuation and the
phase velocity of the SH1 mode vs frequency for three different thicknesses (1, 1.2 and 1.4 mm) for
a glass plate immersed in kerosene: the curves move toward higher attenuation and velocity values
with decreasing plate thickness. The data of figure 4.7 and 4.8 were obtained by using the software
Disperse [79].
27
Figure 4.8: a) The attenuation and b) the phase velocity of the SH1 mode vs frequency for three
different glass plate thicknesses: 1, 1.2 and 1.4 mm; the plate is immersed in kerosene.
Martin et al. [80] were the first to use the SHAPM device as a fluid phase sensor in 42.75° rotated
Y-cut (RYC) quartz (ST-quartz): they experimentally verified the ability of the sensor to monitor
the conditions at the solid/liquid interface. A bare quartz plate was used to measure the viscosity of
water/glycerol mixtures, while the plate with the sensing surface chemically modified by
ethylenediamine ligands was used to detect low concentrations of Cu2+ ions in solution. After this
paper, the SHAPM sensors have been successfully investigated for many applications. Some non-
exhaustive examples of applications include the detection of mercury contamination in water, with
(sub)-nanogram sensitivity, by using a ZX LNO and -65° Y rotated quartz plates covered by a gold
sensitive membrane to accumulate the mercury via surface amalgamation [81]; the detection of
potassium ions concentration in water with a relative frequency shift per unit potassium ions
concentration was found equal to -8.37 ·10-4 for the fundamental mode, by using a ST-90° x quartz
plate covered with a polyvinyl-chloride-valinomycin membrane [77]; the detection of concentration
of NaCl and tris(hydroxymethyl)aminomethane (Tris) in aqueous solution [82] or to analyse the
surface density changes associated with cell adhesion and proliferation in vitro condition, by using a
STx quartz plate [83].
influence of
In reference [78] experimental results with various SHAPMs in ST 90°-x quartz plate concerning
the
the concentration of NaCl and
tris(hydroxymethyl)aminomethane (Tris) in aqueous solution are presented: the higher order modes
appeared to be more sensitive than the first ones, although having more transmission losses.
the viscosity and
the
temperature,
28
5. Lamb Wave sensors
Lamb waves (LWs) are elastic guided waves that travel in finite thickness plates, between stress-
free plane and parallel boundaries; they are elliptically polarized in that they show in-plane and out-
of-plane particle displacement components, U1 and U3. We remand the reader to the book by
Victorov [84] for LWs propagation details. LWs are divided into symmetric (Sn) and anti-
symmetric (An) modes (where n is the mode order). The former modes have the longitudinal
displacement component U1 that is symmetric with respect to the mid plane of the plate while U3 is
anti-symmetric; the opposite happens in the case of the anti-symmetric modes; figure 5.1 shows the
total field profile and the single displacement components of the fundamental mode travelling along
a Si plate.
.
Figure 5.1 The total field profile and the two displacement components of the fundamental
symmetric and anti-symmetric modes, S0 and A0 travelling along a Si plate.
The velocity of the modes depends on the plate characteristics (material type, thickness,
crystallographic cut and wave propagation direction); the thicker the plate is, the more LWs modes
exist. LWs are highly dispersive: as an example, figure 5.2 shows the phase velocity vph vs the plate
thickness-to-wavelength ratio curves of the symmetric Sn (red curves) and anti-symmetric An (blue
curves) LWs travelling in a Si(001)<100> plate of thickness h. The shape of the modes, the
displacement components variation across the cross section of the plate, changes considerably with
the plate thickness and with the mode order [85]. The insets of figure 5.2 show the mode shape of
the first six modes travelling along the plate with fixed thickness (h/λ = 0.5); figure 5.3 shows the
29
same vph dispersion curves as in figure 5.2 but the insets are related to the shape of one mode (S2) at
different h/λ (0.4. 1.0, and 1.8). The data of figures 5.2 and 5.3 were calculated using the
DISPERSE software [79].
Figure 5.2: The vph dispersion curves of the Sn (red curves) and An (blue curves) Lamb modes
travelling in a Si(001)<100> plate in air. The insets show the field profile of different modes at the
same abscissa value (h/λ = 0.5) and are marked by a blue dot.
Figure 5.3: The vph dispersion curves of the Sn (red curves) and An (blue curves) Lamb modes
travelling in a Si(001)<100> plate in air. The insets show the field profile of the S2 mode at
different abscissa values (0.4. 1.0, and 1.8) marked by a red dot.
30
As the LWs have velocity higher than that of the surrounding liquid medium and have both in-
plane and out-of-plane displacement components, they are not suitable for sensing applications in
liquids, except in some special cases. These cases include: (1) a branch of the fundamental
symmetric S0 mode dispersion region where the longitudinal particle displacement component, U1,
is dominant over the out-of-plane component U3 at both the plate surfaces and in the plate depth
(the mode is mostly linearly polarized and propagates at a velocity slightly lower than the velocity
of the longitudinal bulk acoustic wave, vLBAW); (2) a branch of the higher order symmetric modes
dispersion curve, where the modes have U3 ~ 0 at the plate surfaces (but not in the plate depth), and
travel at velocity equal to vLBAW; (3) a branch of the fundamental anti-symmetric A0 mode
dispersion curve, to which corresponds a velocity lower than that of the fluid.
5.1 Quasi-longitudinal symmetric modes
Of great interest are certain points of the symmetric LWs dispersion curves where the phase
velocity is close to the longitudinal bulk acoustic wave (LBAW) velocity of the plate material,
vLBAW, and the field profile has particular characteristics, such as U3 << U1, U2 at the plate surfaces.
These waves, named quasi-longitudinal LWs (QL-LWs), are able to travel along the surface of the
plate while contacting a liquid environment without suffering large attenuation. Inside a small
branch of the S0 vph dispersion curve, corresponding to h/λ << 1, U1 can even have a constant
amplitude along the whole depth of the plate, while U3 is at least 10 times less than U1 at any plate
depth [86, 87]: the shape of the membrane particle movement is a flat ellipse and its longer axis is
parallel to the surface of the plate. The higher order symmetric modes dispersion curves intersect
the velocity of the LBAW in the plate material (vLBAW = 8440 m/s for Si) and they show equal
group velocity (vgr = 7275 m/s). Figure 5.1.1 highlights the intersection of the LWs dispersion
curves in a Si(001)<100> plate with the plate material vLBAW: the mode shape of both the An and Sn
modes at these points shows U3 = 0 at the plate surfaces but, while the An curves are highly
dispersive, the Sn modes show a flat dispersion region centred at the intersection point; this region
corresponds to a h/λ range where the condition U3 << U1 is satisfied, thus preventing the sensor
performances to be highly affected by possible errors in the fabrication technology of the sensor
device.
31
Figure 5.1.1: Dispersion curves for the LWs of a Si(001)<100> plate, 1mm thick, showing the
intersections of the symmetric modes with a phase velocity equal to the LBAW velocity.
As an example figure 5.1.2 shows the field profile of the first four symmetric QL-LWs of figure
5.1.1: QL-S0 (f = 0.441 MHz), QL-S1 (f = 5.85 MHz), QL-S2 (f = 11.70 MHz), and QL-S3 (f =
17.56 MHz). For the QL-S1 to QL-S3 modes, U3 is null only at the plate surfaces, but not inside the
bulk of the plate; for QL-S0 the U3 vanishes on the plate surfaces and remains very small even in the
plate depth, while U1 is almost constant through the plate thickness.
32
Figure 5.1.2: Cross-sectional normalized distribution of U1 and U3 displacement components in Si
for: a) QL-S0 at f = 0.441 MHz, b) QL-S1 (f = 5.8219 MHz), c) QL-S2 (f = 11.642 MHz), and d)
QL-S3 (f = 17.537 MHz).
As it can be seen in figure 5.1.2, the through-thickness for U1 and U3 are symmetric and
antisymmetric about the mid plane of the plate, and the number of the minima increases with
increasing the mode order. Since the U3 component of these higher order modes vanishes on the
free surfaces of the plate, these modes are suitable for liquid sensing.
Figure 5.1.3 shows the vph and attenuation vs f·h curves for the first four symmetric modes (S0, S1,
S2 and S3) for a Si plate immersed in water: when the velocity of the higher order modes reaches the
vLBAW (8440 m/s), the attenuation rapidly drops to zero, thus confirming the modes suitability to
sensing applications in liquid environments. The dispersion curves of figure 5.1.3 were normalized
with respect to the plate thickness by plotting them against the frequency thickness product.
33
Figure 5.1.3: The vph and attenuation vs f·h curves for the a) QL-S0, b) QL-S1, c) QL-S2 and d) QL-
S3 modes travelling in a Si plate contacting a water half-space on both the two plate sides.
The fundamental mode QL-S0 exhibits a plate normalized thickness value, h/λthreshold, beyond
which U3 is no more negligible (U3 > 10%·U1 while U1 = 1 at the plate surfaces but it is no longer
constant inside the plate). For h/λ< h/λthreshold, the wave has U1 = 1 and constant along the plate
depth, and the U3 component is less than 10% of U1.
The h/λthreshold has been calculated for several piezoelectric materials by using the McGill software
[56] and the data are listed in table 5.1.1 [87]. For the higher order modes the h/λthreshold values listed
in table 5.1.1 have a different meaning with respect to QL-S0: it is the plate thickness corresponding
to the minimum value of U3 at the plate surfaces (U3/U1 ~10-3), for vph ~vLBAW. By varying the
thickness of the plate around h/λthreshold, a h/λ range (h/λrange) can be found inside which the
condition U3/U1 ≤ 0.1 at the plate surfaces is verified. Table 5.1.1 summarizes the h/λthreshold and
h/λrange of the fundamental and higher order quasi-longitudinal modes for some piezoelectric
materials; the K2 of the LWs have been evaluated for each material at the corresponding h/λthreshold
for two coupling configurations.
Table 5.1.1: The h/λthreshold, h/λrange and the K2 values for two coupling configurations, for the QL-
S0 to QL-S3 modes, for some piezoelectric materials. h/λrange represents the normalized thickness
range, centered in h/λthreshold , where the condition U3/U1 ≤ 0.1 at the plate surfaces is verified.
34
material
BN
h/λthreshold
h/λrange
QL-S0
0.325
--
QL-S1
1
mode
QL-S2
1.97
QL-S3
2.95
0.70-1.30
1.665 -- 2.26
2.64 -- 3.24
𝐾𝑆𝑇
2 (𝐾𝑀𝑆𝑇
2
) (%)
0.09 (0.14)
0.035 (0.04)
0.018 (0.020)
0.012 (0.013)
ZnO
h/λthreshold
h/λrange
0.07
--
0.65
1.24
1.86
0.59 -- 0.68
1.17 -- 1.305
1.81 - 1.925
𝐾𝑆𝑇
2 (𝐾𝑀𝑆𝑇
2
) (%)
0.47 (8.5)
0.42 (0.50)
0.22 (0.24)
0.15 (0.16)
AlN
h/λthreshold
h/λrange
0.11
--
0.79
1.58
2.37
0.75 - 0.89
1.5 -- 1.67
2.29 -- 2.46
𝐾𝑆𝑇
2 (𝐾𝑀𝑆𝑇
2
) (%)
0.35 (3)
0.31 (0.37)
0.17 (0.19)
0.11 (0.13)
GaN
h/λthreshold
h/λrange
0.12
--
0.77
1.53
2.29
0.735 -- 0.86
1.51 -- 1.62
2.2 -- 2.38
𝐾𝑆𝑇
2 (𝐾𝑀𝑆𝑇
2
) (%)
0.26 (1.61)
0.18 (0.20)
0.095 (0.1)
0.06 (0.07)
The K2 of both the two configurations are quite different for the QL-S0 modes, but they become
very similar with increasing the mode order. The K2 decreases with increasing the mode order, and
the MST configuration is always more efficient than the ST: this last effect is particularly evident
for the QL-S0 mode, while the 𝐾𝑆𝑇
values become similar with increasing the mode order.
Due to the small thickness value (h/λthreshold << 1) of the QL-S0-based plates, the IDTs fingers are
quite close to the opposite floating metal electrode for the MST configuration and consequently the
electric field is mainly perpendicular to the plate surfaces. This results in a coupling efficiency quite
larger than that of the ST configuration [87].
2 and 𝐾𝑀𝑆𝑇
2
When Lamb waves travel along a non-homogenous plate (e.g. bi-layered composite plate), the
symmetry of the particle displacement components with respect to the mid-plane of the plate is lost,
unlike the homogeneous isotropic and anisotropic plates. The modes can be considered as quasi-S0
and quasi-A0 (qS0 and qA0) for a limited plate thickness range, while all the other modes can be
generically labelled as ith mode. Unlike the single material plates, the U1 and U3 displacement
components at the free surfaces of the composite plates can be quite different. As an example,
figure 5.1.4a-c shows the field profile of three quasi longitudinal modes, qS0, qL1, and qL2
travelling in AlN(1.4 μm)/SiN(0.2 μm) plate: the corresponding SiN/AlN total thicknesses values
Htotal/ λ = (hAlN + hSiN)/λ values are 0.08, 0.80 and 1.6. As it can be seen, the condition U3 << U1 is
verified on one plate side that is thus the one suitable for contacting a liquid environment [88].
35
Figure 5.1.4: The field profile of the a) qS0, b) qL1, and c) qL2 modes in air [88].
2D COMSOL Multiphysics software was employed to simulate the QL-S0, QL-S1 and QL-S2
modes propagation along the AlN(1.4 μm)/SiN(0.2 μm) composite plate while contacting the liquid
(water) environment from the SiN side of the plate where the U3 = 0 is satisfied, as opposed to the
AlN side of the plate. As it can be seen in figure 5.1.5, the acoustic energy remains confined inside
the plate [88].
36
Figure 5.1.5: The FEM of the field profile for the qS0, qL1 and qL2 modes at (hAlN +hSiN)/λ = 0.08,
0.8 and 1.6, respectively [88].
In reference [89] the sensitivity to liquid viscosity of the QL-S0 mode in wz-BN/c-AlN thin
composite plates is theoretically predicted for different layers thicknesses. As an example, Table
5.1.2 lists the relative velocity shifts and the attenuation of the QL-S0 mode in wBN/c-AlN
composite plate contacting a liquid environment (70% of glycerol in water with ρl =1091.6 kg/m3, η
= 0.003Pa·s, λ = 10 and 100 μm) for four different combinations of wBN and AlN thicknesses.
Table 5.1.2: The relative velocity shifts and the attenuation of the QL-S0 mode in w-BN/c-AlN
composite plate contacting a liquid environment (70% of glycerol in water with ρl =1091.6 kg/m3, η
= 0.003Pa·s) for λ = 10 and 100 μm, for four different combinations of wBN and c-AlN
thicknesses.
QL-S0
h/λwBN
h/λc-AlN
Frequency
(GHz)
λ = 10 μm
Δv/v0
(10-4)
λ = 100 μm
α
Frequency
(dB/cm)
(MHz)
Δv/v0
(10-4)
α
(dB/cm)
0.1
0.2
0.3
0.013
1.6132
-15.6
-4.91
161.32
-4.94621
-0.155
0.01
1.647
-8.70905
-2.74
164.7
-2.75404
-0.086
0.004
1.6664
-5.9056
-1.85
166.64
-1.86751
-0.059
0.325
1E-3
1.871
-4.95934
-1.56
187.1
-1.56828
-0.049
37
5.2 Fundamental antisymmetric mode
Inside the LWs dispersion curves of figure 5.2, the A0 mode is clearly identified by its reducing
velocity as the plate thickness approaches zero. The A0 mode, while being elliptically polarized,
with U3 not null at the plate surfaces, can travel along thin membranes that are in contact with a
liquid if designed to travel at a velocity lower than that of most liquids, which lie in the range from
900 to about 1500 m/s, by choosing the proper plate thickness. At very small thickness-to-
wavelength ratios, the phase velocity of the A0 mode approaches zero; as the thickness increases,
the velocity increases, and reaches asymptotically from below the SAW velocity of the plate
material.
The Scholte mode, not shown in figure 5.1, is an anti-symmetric mode that propagates at the solid-
fluid interface: its name comes from its similarity to the Scholte wave that is widely known in
geophysics. The characteristic equation for the dispersion curve of this mode is obtained as a
solution to the equations of continuity of stress and displacement at the solid-fluid interfaces to be
solved for antisymmetric modes. In the low frequency limit, one solution is the A0 mode and the
other solution is the quasi-Scholte (Q-Sch): their velocity dispersion curves have a linear
dependence with the frequency-thickness product. The dispersion curve of the latter mode is
characterized by an asymptotic behaviour of the phase velocity approaching the sound speed in the
fluid at high frequencies; this non-dispersive branch of the quasi-Scholte mode dispersion curve is
named Scholte mode. The polarisation of the mode is mostly parallel to the interface with a small
out-of-plane displacement component.
Figure 5.2.1 shows the phase velocity and attenuation curves versus the frequency for the S0, A0 and
quasi-Scholte (Q-Sch) modes in a glass plate, 0.15 mm thick, immersed in water.
Figure 5.2.1: Phase velocity and attenuation dispersion curves of S0, A0 and Q-Sch modes for a
glass plate 0.15 mm thick immersed in water.
38
The S0 mode attenuation starts from zero with the frequency and reaches a local maximum value at
2.68 MHz·mm. In the limit as f·h tends to zero, the S0 mode behaves basically as a longitudinally
polarized wave, which explains its weak attenuation. When f·h tends to infinity the attenuation
increases asymptotically with f 2: the mode is essentially concentrated at the surfaces of the plate
and it radiates energy into the surrounding liquid in the same manner as a Rayleigh wave which is
proportional to f2. When the plate thickness becomes comparable with the acoustic wavelength, the
A0 mode behaves in the same way as the mode S0 but it has an important attenuation in the low f ·h
limit, caused by its flexural motion normal to the plate surface. The cut-off of the attenuation curve
is at f·h =0.25 MHz·mm: below this limit, the phase velocity of the A0 mode is smaller than the
sound velocity in the liquid. As a result, no radiation of guided waves is allowed by the Snell law.
The phase velocity of the Q-Sch plate mode rises with frequency from zero and gradually
asymptotes to the velocity of the liquid half-spaces. Its attenuation is affected by the fluid bulk
velocity, viscosity and bulk longitudinal attenuation. Q-Sch mode travels unattenuated in the
direction of the phase velocity (if the fluid has no longitudinal attenuation); as it travels at velocity
lower than the bulk velocity of the fluid, it is consequently evanescent in the direction orthogonal to
the interface. The wave amplitude decays in an exponential manner with distance from the
interface. The extent to which the wave penetrates into the fluid depends on the frequency, as
shown in figure 5.2.2 where U1, U3 and the strain energy density (SED) of the Q-Sch mode
travelling in a glass plate (1 mm thick) immersed in water at f = 107.602 and 454.186 kHz are
shown.
Figure 5.2.2: U1, U3 and the strain energy density (SED) of the Q-Sch mode travelling in a glass
plate (1 mm thick) immersed in water at a) f = 107.602 and b) f = 454.186 kHz.
39
The Q-Sch mode energy distribution between the fluid and the plate depends on the frequency, as
shown in figures 5.2.2 a and b: the out of plane displacement component at 107.602 kHz is almost
constant across the section of the plate and the strain energy density indicates that the energy travels
predominantly in the plate. At frequency 454.186 kHz a relevant part of the energy is travelling in
the fluid. At higher frequencies (> 1MHz·mm) most of the energy travels in the fluid: the
displacements decay away from the surfaces and are a minimum at the centre of the plate. Q-Sch
waves can be used to characterize the fluid properties [90] since the wave attenuation, phase and
group velocity are affected by the viscosity, longitudinal bulk attenuation and bulk velocity of the
fluid. In reference [91] the influence of the waveguide material (steel, aluminium and brass) on the
Q-Sch mode phase and group velocity sensitivities to the liquid parameters (longitudinal velocity
and density) is theoretically studied. The study concluded that higher waveguide material density
leads to higher sensitivities, and higher waveguide acoustic velocities lead to an extended effective
sensing range.
As an example figure 5.2.3 shows Scholte mode phase and group velocity dispersion curves in Al
plate (1 mm thick) immersed in water (ρ = 1000 kg/m3, v =1500 m/s), benzene (ρ = 881 kg/m3, v =
1117 m/s, dynamic viscosity η = 0.65·10-3 Ns/m2) and diesel (ρ = 800 kg/m3, v =1250 m/s ).
Figure 5.2.3: Scholte mode phase and group velocity dispersion curves in Al plate (1 mm thick)
immersed in water (ρ = 1000 kg/m3, v =1500 m/s), benzene (ρ = 881 kg/m3, v = 1117 m/s, dynamic
viscosity η = 0.65·10-3 Ns/m2) and diesel (ρ = 800 kg/m3, v =1250 m/s ).
In reference [92] the sensitivity of the quasi-Scholte mode for fluid characterization was assessed
experimentally by measuring the phase velocity values for the quasi-Scholte mode in distilled water
40
and in different ethanol-water concentrations. In reference [93] a preliminary sensitivity analysis is
performed for application to simultaneous multi-sensing physical quantities of liquids such as
temperature, viscosity and density using interface waves.
Figures 5.2.4a and b show the A0 and S0 mode velocity and attenuation dispersion curves in Si
plate (1mm thick) immersed in glycerol (ρ = 1258 kg/m3, v = 1860 m/s, dynamic viscosity η = 1.49
Ns/m2), water (ρ = 1000 kg/m3, v =1500 m/s), benzene (ρ = 881 kg/m3, v = 1117 m/s, dynamic
viscosity η = 0.65·10-3 Ns/m2) and diesel (ρ = 800 kg/m3, v =1250 m/s ); the longitudinal
attenuation is set to zero.
Figure 5.2.4: a) the A0 mode and b) the S0 mode velocity and attenuation dispersion curves in Si
plate (1 mm thick) immersed in glycerol (ρ = 1258 kg/m3, v = 1860 m/s, dynamic viscosity η = 1.49
Ns/m2), water (ρ = 1000 kg/m3, v =1500 m/s), benzene (ρ = 881 kg/m3, v = 1117 m/s, dynamic
viscosity η = 0.65·10-3 Ns/m2) and diesel (ρ = 800 kg/m3, v =1250 m/s ).
41
The A0 and S0 modes attenuation curves related to benzene and diesel are very close and difficult to
distinguish; the same applies to the phase velocity curves but for all the studied liquids.
The A0 mode attenuation is larger than that of the S0 mode due to its faster decay over the
propagation distance: the maximum attenuation (127.33 Nepers/m) happens when f·h equals 0.15
MHz mm, when the A0 Lamb wave phase velocity is close to the water sound speed. When the A0
phase velocity is less than the water sound speed, there is still attenuation that approaches to zero
slowly as h/λ comes to zero.
In the low viscosity range, the amplitude response of the sensor is also affected by other parameters,
such as temperature, pressure and density which can play more important roles than the viscosity. In
the case of water and diesel, the sensor responses to these two liquids are well distinguishable and
are affected only by the mass density and velocity of the liquids, being the viscosity assumed to be
equal to zero. On the contrary, the sensor responses to benzene and diesel, which have quite similar
ρ and vl but different (and very low) η, are very similar.
The devices based on the A0 mode suffer some limitations, such as the low operating frequency (f =
vph/λ) due to the vph of A0 mode which must be lower than the liquid velocity; thus the A0-based
device is not suitable to achieve high frequencies that is a prerequisite to enhance the sensor
sensitivity [1]. If the device frequency is increased by reducing the IDT width, two technological
problems are met: 1. if the sensor is implemented onto a single crystal piezoelectric substrate, the
plate thickness must be scaled down together with the wavelength, thus increasing the fragility of
the thinned plate; 2. If the sensor is implemented onto a thin suspended piezoelectric membrane, the
layer structural quality imposes a limit to the maximum (and minimum) thickness. Another
limitation is the achievable efficiency of electrical excitation of the acoustic wave. The K2 of A0
mode is dispersive as it depends on the membrane thickness. The theoretical K2 dispersion curve of
the A0 mode is reported in reference [87] for various piezoelectric plates (BN, ZnO, InN, AlN and
GaN),
and
Metal/Substrate/Transducer(MST) configurations. Table 5.2.1 lists the threshold plate thickness for
operation in water (vph ≤ vwater = 1480 m/s) and the corresponding K2 of the ST and MST structures
at the threshold [87].
Substrate/Transducer
structures:
(ST)
for
two
coupling
the
Table 5.2.1: the A0 mode h/λtreshold for operation in water (vph ≤ vwater = 1480 m/s) and the
corresponding K2 of the ST and MST structures at h/λtreshold.
material
h/λtreshold
2 (%)
𝐾𝑆𝑇
2
𝐾𝑀𝑆𝑇
(%)
ZnO
BN
AlN
GaN
InN
0.19
0.05
0.087
0.12
0.2
2.63
0.011
0.25
0.35
1.77
0.7
0.0
0.02
0.06
0.58
42
Even though the ZnO K2 is relatively high, it is significantly smaller than that of the QL-S0 mode
(8.5% for MST structure).
The experimental result of fluid loading of a Lamb wave sensor employing A0 mode was firstly
reported by R.M White and S.W. Wenzel in 1988 [94]. Furthermore, the same group reported the
experimental result of viscosity and density sensing using the same device [95] consisting of a
composite SiN and ZnO membrane with the thicknesses range from 2.8 um to 6.0 μm and the IDT
periodicity of 100 μm. In this configuration, the membrane normalized thickness is thin enough to
obtain an A0 phase velocity lower the sound velocity in water. For h/λ = 0.06, they obtained A0 and
S0 modes with velocity of 470 m/s and 7850 m/s respectively. The effect of viscous fluid loading on
A0 mode is reported which show linear relationship between the attenuation loss and the square root
of the product of fluid mass density and viscosity. Moreover, the authors demonstrated that
simultaneous measurement of frequency shift and attenuation loss allows a fluid's viscosity and
density to be determined.
In reference [96] experimental results on A0-based sensor on PZT are described. The device was
fabricated by deposition of low pressure chemical-vapor deposition (LPCVD) silicon nitride, 1 μm
thick, a metal ground plane of Ta/Pt (10 nm/150 nm), and a 750-nm-thick layer of sol-gel-derived
PZT on silicon wafer, followed by the lift off process of Ta/Pt (10 nm/150 nm) IDTs. The KOH
was used for anisotropic etching of the back side of the silicon wafer to release the composite
membranes structure. They obtained A0 mode phase velocities in between 295 -- 312 m/s and group
velocities in between 414 -- 454 m/s. A frequency shift of 850 kHz and an insertion loss as low as 3
dB are observed when the back side of the membrane is in contact with a column of 15-mm height
of deionized water.
The theoretical study based on the Rayleigh's perturbation approach to compare the A0 and S0 Lamb
wave sensors sensitivity in liquid is reported in reference [97]: it is shown that the sensitivity of the
A0 mode is much greater than that of S0 mode. In reference [98] the experimental test of the A0, S0
and SH0 sensors demonstrates that the A0 mode frequency shift caused by the presence of liquid is
quite larger than that of the S0 and SH0 modes.
The most recent experimental result of S0 mode to measure the mechanical and electrical liquid
properties is reported by Miera et. Al. [99]: using two AlN-based S0 sensor topologies, with floating
bottom metallic layer and without, the influence of mechanical and electrical properties of different
aqueous mixtures was experimentally assessed.
6. Discussion and conclusions
All acoustic wave devices behave like sensors since they are highly sensitive to any change in
the boundary conditions that result in a wave velocity and/or propagation loss change. If the
acoustic wave path is covered by a sensitive coating that is able to absorb only specific chemical
vapors or specific biological chemicals in liquids, the sensor becomes a chemical sensor or a
biosensor. All the acoustic wave sensors are able to work in gaseous environments, but only a
subset of them can operate in contact with liquids. The waves that are predominantly in-plane
polarized do not radiate appreciable energy into liquids contacting the device surface, as opposed to
those waves with a substantial out-of-plane displacement component, which radiates compressive
43
waves into the liquid, thus causing excessive damping. An exception to this rule occurs for devices
utilizing waves that propagate at a velocity lower than the sound velocity in the liquid.
The quartz crystal microbalance (QCM) is one of the most common shear horizontal mode
sensor: it typically consists of a thin disk of AT-cut quartz with parallel metal electrodes patterned
on both sides. When a voltage is applied between these electrodes, the plate undergoes a shear
deformation. The QCM resonant frequency, 𝑓 = 𝑣𝐵𝐴𝑊/2ℎ , is inversely proportional to the quartz
plate thickness h (few hundreds of micrometers): its value is typically between 5 -- 30 MHz. When
the QCM contacts a Newtonian liquid, it results in a resonance frequency shift proportional to the
square root of the liquid viscosity-density product, ∆𝑓 = −𝑓0
3/2√
𝜌𝑙𝜂𝑙 𝜌𝑞𝜇𝑞
⁄
, as described by
Kanazawa and Gordon [100]; ρL and ηL denote the density and viscosity of the liquid, ρQ is the
density of quartz, μQ is the elastic constant of the piezoelectrically stiffened quartz, 𝑓0 is the QCM
resonant frequency, respectively. The fluid is entrained within a distance from the quartz surface
that is equal to the wave penetration depth, = √
2𝜂𝑙 𝜔𝜌𝑙
⁄
. An increased QCM sensitivity requires a
higher resonant frequency: this effect can be achieved by thinning the quartz plate that thus
becomes more fragile and difficult to manufacture and handle.
An alternative to the QCM are the surface-generated acoustic wave sensors whose operating
frequency is determined by the periodicity of the IDTs and the mode velocity, 𝑓 = 𝑣
𝜆⁄ . Moreover
their acoustic energy is trapped near the surface where the sensing phenomena take place: the more
acoustic energy is concentrated at the surface, the higher the sensitivity to surface perturbations.
PSAW, HVPSAW, Love waves, and Lamb waves belong to this group. SHAPMs modes do not
belong to this group since the acoustic energy is distributed throughout the bulk of the substrate,
even if the waves propagation is excited and detected by means of the IDTs, as well as for the
SAWs. The sensitivity of the SHAPM sensors depends strongly on the thickness of the substrate: a
higher sensitivity can be obtained by thinning the substrate, at the cost of lack of substrate
robustness. These sensors are more sensitive than the QCM, but less sensitive than the other
surface-generated acoustic wave sensors since the energy of the wave is distributed throughout the
bulk of the substrate and not trapped at the surface where the sensing phenomena take place. An
advantageous feature of the SHAPM sensors is that these devices are sensitive on both sides of the
substrate so that the back side can be used for sensing while having the front protected from the
liquid [86]. This simple technology is opposed to that required by the Lamb Wave sensors, that
requires the thinning of the bulk piezoelectric substrate or the release of a thin suspended
membrane, or to that of the Love wave sensors, that require the growing of a guiding layer on top of
the piezoelectric half-space.
Thin piezoelectric suspended membranes are
the silicon
micromachining techniques. By using the thin piezoelectric film technology, high-frequency
devices can be designed that offer an important advantage: the surrounding electronic circuitry
(amplifier) can be integrated with the acoustic device, thus offering the possibility for on-chip
compensation of disturbing effects such as temperature or pressure variation. The fundamental anti-
symmetric Lamb mode travelling along a very thin membrane (thickness typically 5% or less of the
fabricated by employing
44
wavelength) has a phase velocity lower than the sound velocity of the loading liquid 𝑣𝑙 (typically
from 900 to 1900 m/s). Therefore, the plate functions as a wave guide yielding no radiation losses
despite the wave has a non-null shear vertical particle displacement component. Due to the low
phase velocity of the A0 mode, the operating frequency is typically in the range of 5-20 MHz. The
fundamental symmetric Lamb mode, S0, is predominantly longitudinally polarized just for small
plate thickness values: as its phase velocity is much higher than that of the A0 mode, it can reach
higher sensibility.
The higher order quasi-longitudinal Lamb modes are promising candidates for liquid sensing
applications: they have high velocity (close to that of the longitudinal BAW in the plate material)
and require a fabrication technology simpler than that of the other Lamb wave sensors since they
correspond a thicker plate thickness-to-wavelength ratio. For example, h/ = 0.07 is the upper limit
of the ZnO plate thickness that allows the propagation of longitudinally polarized S0 mode, with K2
= 9%. Higher order quasi-symmetric modes S1, S2 S3, and S4, propagate for h/λ ~ 0.65, 1.24, 1.86,
2.48. Their K2 is equal to 0.42%, 0.22%, 0.15%, and 0.10%, respectively, much lower than that of
the S0 mode. For 10 μm thick ZnO membrane, the S1, S2, S3 and S4-based devices operating
frequencies are 394 MHz, 754 MHz and 1125 MHz and 1507 MHz, respectively, as opposed to that
(38 MHz) estimated for the S0 mode [101]. The thin suspended membrane devices may be only a
few micrometers thick, thus the mass per unit area of the thin plate can be increased significantly by
the mass-loading effect produced by the changes in the density of a fluid, or by the attachment of
protein molecules, cells, and bacteria from a liquid onto the suspended membrane surface.
Moreover, these sensors can be fabricated by techniques compatible with planar integrated circuit
technology thus allowing the low cost and small-sized sensors fabrication and the integration of the
device with the surrounding electronic circuitry.
Love wave devices include a wave-guiding film deposited on top of a substrate [102]. The
sensitivity of the Love wave sensors depends on the layer and half-space materials type and their
crystallographic orientation, and on the layer thickness. There is a relationship between the slope of
the dispersion curve to the mass sensitivity of a Love wave sensor that allows to find the layer
thickness corresponding to the optimal sensor sensitivity. If the Love wave sensor consists in a
piezoelectric layer covering a non-piezoelectric half-space, then the IDTs can be buried under the
guiding layer that recovers the additional role of shielding the IDTs from aggressive liquid
environment. The Love wave sensor can also include a sensing coating deposited onto the
waveguiding layer: if this coating layer has the proper elastic properties, then it can also cover the
role of waveguiding layer, with the advantage of avoiding a step of the device manufacturing
process.
If the cut of the piezoelectric crystal substrate is properly rotated, the wave propagation mode
changes from an elliptically polarized SAW to in-plane polarized SAWs, that are the PSAW and
HVPSAW. This dramatically reduces the losses when liquids contact the propagating medium, thus
allowing these devices to operate as biosensors. The PSAW and HVPSAW phase velocities are
higher than the generalized SAW phase velocity, thus allowing the fabrication of higher frequency
devices for the same photolithographic resolution, which also increases the sensor sensitivity.The
PSAW and HVPSAW-based sensors have many advantages, such as a simple structure, highly
45
reproducible large-scale fabrication technology, straightforward integration with microfluidic
channels; moreover, their operating frequencies are higher than the QCM counterpart. However, a
disadvantage of such devices is that the IDTs are located on the same side of the substrate that
contacts the liquid: as a consequence, the test cell must have a reduced size and must be sealed
within the propagation path.
Provided that the LWs, SHAPMs, the Love modes and the PSAW-based devices are good
candidates for biosensing and chemical sensing in liquids, it is worth noting to underline that the
design parameters (such as materials type, crystallographic orientation, electrical boundary
conditions, layers thickness, etc.) of any electroacoustic device highly affect the sensors
performances. It has been proven that numerical calculations and FEM analysis yield an in-depth
study of the SAW characteristics to obtain very useful description of the sensor's performances and
a deep assessment of the device sensitivity.
The present study wants to point the reader's attention towards the features of different kinds of
acoustic waves and modes, as well as optimal combinations of materials and electrode structures,
that can be exploited to design electro-acoustic devices suitable for chemical and biochemical
sensing applications. A promising application of the SAW-based devices consist of a fully
integrated platform that uses the SAWs to develop several processes, from sample handling to
detection and measure. SAW-directed transport of analytes can be coupled to the SAW sensing
functionalities. The SAW microfluidic systems can move and remove liquid targets to and from
different sensing modules on the same chip; SAW sensors can make analysis of fluids as well as
separation and fractionation of cells. However, it is important to point out that the future
developments in acoustic-type chemical and biological sensors is also strongly related to the
development of sensitive membranes which are responsible for the selectivity of the sensors: at this
aim, coordinated efforts between researchers in diverse disciplines (such as electrical engineering,
physics, chemistry, microbiology and medicine) are needed.
Acknowledgment
This project has received funding from the European Union's Horizon 2020 research and
innovation programme under the Marie Sklodowska-Curie Grant Agreement No. 642688.
ORCID iDs
C Caliendo https://orcid.org/0000-0002-8363-2972
M Hamidullah https://orcid.org/0000-0003-2131-4335
References
1.
Acoustic Wave Sensors: Theory, Design and Physico-Chemical Applications, Series
Editors: Moises Levy Richard Stern, Authors: D. Ballantine, Jr. Robert White S. Martin Antonio
Ricco E. Zellers G. Frye H. Wohltjen, Academic Press 1997.
46
C. Caliendo, A. D'Amico, P. Verardi, E. Verona, Surface acoustic wave H2 sensor on silicon
2.
substrate, 1988 IEEE Ultrasonics Symposium Proc., Chicago (IL), 2-5 Oct. 1988, p. 569-574, vol.1.
3.
C. Caliendo, A. D'Amico, E. Verona, Surface acoustic gas sensors, in Gas sensors:
Principles, Operation and Developement, cap: VIII, Ed. G. Sberveglieri, (Kluver Academic
Publisher, Dordrecht 1992), pp. 281-306.
4.
A.J. Ricco, S.J. Martin, T.E. Zipperian, Surface acoustic wave gas sensor based on film
conductivity changes, Third International Conference on Solid-State Sensors and Actuators
(Transducers '85), Philadelphia, PA, U.S.A., June 11 - 14, 1985, https://doi.org/10.1016/0250-
6874(85)80031-7.
5.
R. Arsat, M. Breedon, M. Shafiei, K. Kalantar-zadeh, W. Wlodarski, P. G. Spizziri, S. Gilje,
R. B. Kaner , Graphene-like nano-sheets for surface acoustic wave gas sensor applications,
Chemical Physics Letters, ISSN: 0009-2614, Vol: 467, Issue: 4, Page: 344-347, 2009,
https://doi.org/10.1016/j.cplett.2008.11.039.
Adnan Mujahid and Franz L. Dickert, Surface Acoustic Wave (SAW) for Chemical Sensing
(Basel). 2017 Dec; 17(12): 2716, doi:
6.
Applications of Recognition Layers, Sensors
10.3390/s17122716.
7.
C. Caliendo, A. D'Amico, A. Furlani, G. Iucci, M.V. Russo, E. Verona, A new surface
acoustic wave humidity sensor based on a polyethynyl-fluorenol membrane, Sensors and Actuators,
B 18, 82 (1994).
Li, Z., Jones, Y., Hossenlopp, J., Cernosek, R., and Josse, F. (2005) Analysis of liquid-phase
8.
chemical detection using guided shear horizontal-surface acoustic wave sensors. Anal. Chem., 77:
4595 -- 4603.
X.C.Zhou, S.C.Ng, H.S.O.Chan, S.F.Y.Li, Piezoelectric sensor for detection of organic
9.
amines in aqueous phase based on a polysiloxane coating incorporating acidic functional groups,
Analytica Chimica Acta, Volume 345,
June 1997, Pages 29-35,
https://doi.org/10.1016/S0003-2670(97)00070-6.
Issues 1 -- 3, 20
10.
Steffen Rosler, Ralf Lucklum, Ralf Borngrab, Sensor system for the detection of organic
pollutants in water by thickness shear mode resonators, 1998 Sensors and Actuators BChemical
48(1):415-424, DOI: 10.1016/S0925-4005(98)00079-3. Ù
11.
Berkenpas E., Millard P, Pereira da Cunha M., Detection of Escherichia coli O157:H7 with
langasite pure shear horizontal surface acoustic wave sensors, Biosens Bioelectron. 2006 Jun
15;21(12):2255-62. Epub 2005 Dec 13.
Howe E, Harding G., A comparison of protocols for the optimisation of detection of bacteria
12.
using a surface acoustic wave (SAW) biosensor.. Biosens Bioelectron. 2000;15(11-12):641- 9.
Kerstin Lange & Bastian E. Rapp & Michael Rapp, Surface acoustic wave biosensors: a
13.
review, Anal Bioanal Chem (2008) 391:1509 -- 1519, DOI 10.1007/s00216-008-1911-5
47
Bryant,A. ; Poirier,M. ; Lee,D. L. ; Vetelino,J. F., A Surface Acoustic Wave Gas Detector,,
IL, USA, DOI:
IEEE Ultrasonics Symposium, 14-16 Oct. 1981
, Chicago,
14.
1981
10.1109/ULTSYM.1981.197604, p. 171.
Caliendo C., Verona E., D'Amico A. (1992) Surface Acoustic Wave (SAW) Gas Sensors.
15.
In: Sberveglieri G. (eds) Gas Sensors. Springer, Dordrecht, DOI 10.1007/978-94-011-2737- 0_8.
16.
A.Venema, E.Nieuwkoop, M.J.Vellekoop, M.S.Nieuwenhuizen, A.W.Barendsz, Design
aspects of saw gas sensors, , Sensors and Actuators, Volume 10, Issues 1 -- 2, 10 September 1986,
Pages 47-64, https://doi.org/10.1016/0250-6874(86)80034-8.
17. Wen Wang, Yana Jia, Xinlu Liu, Yong Liang, Xufeng Xue and Zaofu Du , Enhanced
sensitivity of temperature-compensated SAW-based current sensor using the magnetostrictive
effect, Smart Mater. Struct. 26 025008, https://doi.org/10.1088/1361- 665X/aa5137.
H. Nakamura, H. Nakanishi, J. Fujivara, T. Tsurunari, A review of SiO2 thin film
18.
technology for temperature compensated SAW devices, Proc. Of the Sixth International
Symposium on Acoustic Wave Devices for future mobile communication systems, 24-25 Nov.
2015, Keyaki Hall, Chiba University, Japan, pp. 67-72.
19.
Chih-Ming Lin, Ting-Ta Yen, Yun-Ju Lai, Valery V. Felmetsger, Matthew A. Hopcroft, Jan
h. Kuypers, and Albert P. Pisano, Temperature-Compensated Aluminum Nitride Lamb Wave
Resonators, IEEE Trans. on Ultr., Ferroel., and Freq. Control, vol. 57, no. 3, 2010.
20. White, R. M.; Voltmer, F. W. Appl. Phys. Lett. 7, 314 (1965)].
Ken-Ya Hashimoto, Surface Acoustic Wave Devices in Telecommunications: Modelling
21.
and Simulation, chapter 3, Springer-Verlag Berlin Heidelberg, 2000, doi 10.1007/978-3-662-04223-
6
Kenji Uchino, Introduction to Piezoelectric Actuators and Transducers, Public report from
22.
International Center for Actuators and Transducers, Penn State University University Park, PA
16802, 2003.
Colin K. Campbell, Surface Acoustic Wave Devices for Mobile and Wireless
23.
Communications. Academic Press: Boston, 1998
24.
O. Elmazria and T. Aubert, Wireless SAW sensor for high temperature applications:
Material point of view, Smart Sensors, Actuators, and MEMS V, edited by Ulrich Schmid, Jose
Luis Sanchez-Rojas, Monika Leester-Schaedel, Proc. of SPIE Vol. 8066, 806602-1, 2011 SPIE
doi: 10.1117/12.889165.
25.
C. Caliendo, P.M. Latino, Characterization of Pt/AlN/Pt-based structures for high
temperature microwave electroacoustic devices applications, Thin Solid Films, 519 (19), 6326-
6329, 2011.
26.
Cinzia Caliendo, Fabio Lo Castro, CHAPTER 17 - Advanced bulk and thin film materials
for harsh environment MEMS applications, Editor(s): Mahmood Aliofkhazraei, Anti-Abrasive
Nanocoatings, Woodhead Publishing, 2015, Pages 429-453, doi.org/10.1016/B978-0-85709-211-
3.00017-0.
48
C.Caliendo, Surface Acoustic Wave Devices for Harsh Environment, Acoustic Waves -- From
27.
Microdevices to Helioseismology, InTech, Croatia (2011), pp. 579-594, doi. 10.5772/21665.
28.
Advances in surface acoustic wave technology, systems and applications, Vol.2, Ed. C. W.
Ruppel and Tor A. Fjeldly, 2001 World Scientific Singapore, book chapter "Pseudo and High
Velocity Pseudo SAWs", M. P. da Cunha, pp. 203-244.
29.
F. S. Hickernell, E. L. Adler, "The experimental and theoretical characterization of SAW
modes on ST-X quartz with a zinc oxide film layer", Proc. IEEE Int. Freq. Contr. Symp., pp. 852-
857, 1997-May.
Slie, W.M.; Donfor, A.R., Jr.; Litovitz, T.A. Ultrasonic Shear and Longitudinal
30.
Measurements in Aqueous Glycerol. J. Chem. Phys. 1966, 44, 3712.
31.
H. Tarbague, J.-L. Lachaud, S. Destor, L. Vellutini, J.-P. Pillot, B. Bennetau, E. Pascal, D.
Moynet, D.Mossalayi, D. Rebiere, C. Dejous, PDMS (Polydimethylsiloxane) Microfluidic Chip
Molding for Love Wave Biosensor, Journal Integrated Circuits and Systems 2010; v.5 / n.2:125-13.
32.
L. A. Francis, J.-M. Friedt, C. Bartic and A. Campitelli, A SU-8 liquid cell for surface
acoustic wave biosensors, Proc. SPIE 5455, MEMS, MOEMS, and Micromachining, (16 August
2004); doi: 10.1117/12.544779; https://doi.org/10.1117/12.544779.
33. M. P. da Cunha, Pseudo and High velocity Pseudo SAW, in Advances in Surface Acoustic
Wave Technology, Systems and Applications, Volume 2, Editors Clemens C. W. Ruppel and Tor
A. Fjeldly, 2001 World Scientific Publishing, Singapore, New Jersey, London, Hong Kong, pp.
203-243.
34. M. Pereira da Cunha, PSAW and HVPSAW behaviour in layered structures, Proceedings of
the International Conference Microwave and Optoelectronics 1997 SBMO/IEEE MTT-S 11- 14
Aug., Natal, Brazil, DOI: 10.1109/SBMOMO.1997.646835.
35.
M. Pereira da Cunha, Effects of layer thickness for GSAW, PSAW and HVPSAW devices,
1997 IEEE Ultrasonics Symposium Proceedings, (Cat. No.97CH36118) 1997, Volume: 1, Pages:
239 -- 244, DOI: 10.1109/ULTSYM.1997.663018.
Eric L. Adler , SAW, pseudo SAW, and hvpseudo SAW in langasite, 1998 Proceedings of
36.
the IEEE Ultrasonics Symposium pp. 307 - 310 vol.1, DOI10.1109/ULTSYM.1998.762152.
37. Michio Kadota, Jun Nakanishi, Takeshi Kitamura and Makoto Kumatoriya, Properties of
Leaky, Leaky Pseudo, and Rayleigh Surface Acoustic Waves on Various Rotated Y-cut Langasite
Crystal Substrates, 1999 The Japan Society of Applied Physic, Japanese Journal of Applied
Physics, Volume 38, Part 1, Number 5B .
38. M. Pereira da Cunha ; D.C. Malocha ; D.W. Puccio ; J. Thiele ; T.B. Pollard, LGX pure
shear horizontal SAW for liquid sensor applications, IEEE Sensors Journal, Vol. 3, Issue: 5, Oct.
2003, pp. 554 -- 561, 2003, DOI: 10.1109/JSEN.2003.817163.
J. Hechner , W. Soluch, Pseudo surface acoustic wave dual delay line on 41◦YX LiNbO3 for
39.
liquid sensors, Sensors and Actuators B 111 -- 112 (2005), pp. 436 -- 440.
49
Jun Kondoh, Shohei Tabushi, Yoshikazu Matsui, Showko Shiokawa, Development of
40.
methanol sensor using a shear horizontal surface acoustic wave device for a direct methanol fuel
cell, Sensors and Actuators B 129 (2008) 575 -- 580.
41. Marshall S. Smith ; Donald C. Malocha, High frequency characterization of leaky waves for
liquid delay
IEEE, 30 Oct.-3 Nov. 2016, DOI:
10.1109/ICSENS.2016.7808402.
sensors, SENSORS, 2016
line
42.
F. S. Hickemell, E. L. Adler, Pseudo-SAW Propagation on Layered Piezo-
Substrates:Experiments and Theory Including Film Viscosity, Ultrasonics Symposium, 1996.
Proceedings., 1996
IEEE, 3-6 Nov. 1996, San Antonio, TX, USA, USA, DOI:
10.1109/ULTSYM.1996.583805.
Showko Shiokawa and Toyosaka Moriizumi, Design of SAW Sensor in Liquid, 1988
43.
Japanese Journal of Applied Physics, Volume 27, Supplement 27-1.
44.
X.J. Tong ; D. Zhang ; Y.C. Li ; Y.A. Shui, Investigation of longitudinal leaky surface
acoustic wave on lithium niobate and quartz, Ultrasonics Symposium, 1997. Proceedings., 1997
IEEE 5-8 Oct. 1997, DOI: 10.1109/ULTSYM.1997.663000.
45.
X.J. Tong ; D. Zhang, Novel propagation direction of quasi-longitudinal leaky surface
acoustic wave on quartz and its potential as liquid sensors, Sensors and Actuators A Physical 78(2-
3):160-162 DOI 10.1016/S0924-4247(99)00221-6.
46.
Yingmin Fan, Zhenghui Liu, Gengzhao Xu, Haijian Zhong, Zengli Huang, Yumin Zhang,
Jianfeng Wang, and Ke Xu, Surface acoustic waves in semi-insulating Fe-doped GaN films grown
by
(2014);
https://doi.org/10.1063/1.4893156.
epitaxy, Appl.
Phys. Lett.
hydride
vapor
phase
105,
062108
47.
Fumiya Matsukura, Masato Uematsu, Keiko Hosaka and Shoji Kakio, Longitudinal-Type
Leaky Surface Acoustic Wave on LiNbO3 with High-Velocity Thin Film, 2013 The Japan Society
of Applied Physics, Japanese Journal of Applied Physics, Volume 52, Number 7S .
Xiaojun Tong and De Zhang, Novel propagation direction of quasi-longitudinal leaky
48.
surface acoustic wave on quartz and its potential as liquid sensors, Sensors ans Actuators, 78 (1999)
160-162],
F. Josse F. Bender ; R.W. Cernosek ; K. Zinszer, Guided SH-SAW sensors for liquid-phase
49.
detection, Proceedings of the 2001 IEEE International Frequncy Control Symposium and PDA
Exhibition, Seattle, WA, USA, DOI: 10.1109/FREQ.2001.956273.
Zeinab Ramshani, Avuthu S.G. Reddy, Binu B. Narakathu, Jared T. Wabeke, Sherine O.
50.
Obare, Massood Z. Atashbar, SH-SAW sensor based microfluidic system for the detection of heavy
metal compounds in liquid environments, Sensors and Actuators B: Chemical, Volume 217, 2015,
Pages 72-77, ISSN 0925-4005, doi:10.1016/j.snb.2014.12.026.
51.
JUN KONDOH, A Liquid-Phase Sensor Using Shear Horizontal Surface Acoustic Wave
Devices, Electronics and Communications in Japan, Vol. 96, No. 2, 2013 Translated from Denki
Gakkai Ronbunshi, Vol. 131-C, No. 6, June 2011, pp. 1094 -- 1100.
50
J. L. Vivancos, Z. Racz, M. Cole, J. Soto, J. W. Gardner , Detergents sensing system based
52.
on SH-SAW devices, December 2011, Procedia Engineering 25:1125-1128, DOI:
10.1016/j.proeng.2011.12.277
53.
N. Gozde Durmus, Richard L. Lin, Mariel Kozberg, Deniz Dermici, Ali Khademhosseini
and Utkan Demirci, Acoustic-Based Biosensors, Encyclopedia of Microfluidics and Nanofluidics,
DOI 10.1007/978-3-642-27758-0_10-2, Springer Science+Business Media New York 2014.
54.
E. Gizelli, A.C. Stevenson, N.J. Goddard, C.R. Lowe, Surface skimming bulk waves: a
novel approach to acoustic biosensors, TRANSDUCERS '91: 1991 International Conference on
Solid-State Sensors and Actuators. Digest of Technical Papers, 24-27 June 1991 San Francisco, CA,
USA, USADOI: 10.1109/SENSOR.1991.148975.
55. Maria Isabel Rocha Gaso, Yolanda Jimenez, Laurent A. Francis and Antonio Arnau, Love
wave Biosensors: A Review, Intech 2013, chapter 11, doi.org/10.5772/53077.
E. L. Adler, J. K. Slaboszewics, G. W. Farnell, and C. K. Jen, IEEE Trans. Ultrason.
56.
Ferroelectr. Freq. Control 37, 215 (1990).
57.
Fabrice Martin, Glen McHale, and Michael I. Newton, Experimental Study of Love wave
Sensor Response by Phase and Group Velocity Measurement, IEEE SENSORS JOURNAL, VOL.
4, NO. 2, APRIL 2004, pp. 216-220.
58.
Cinzia Caliendo, Muhammad Hamidullah, A Theoretical Study of Love wave Sensors Based
on ZnO -- Glass Layered Structures for Application to Liquid Environments, Biosensors (Basel).
2016 Dec; 6(4), 59.
59.
Z.N. Danoyan, G.T. Piliposian, Surface electro-elastic Love waves in a layered structure
with a piezoelectric substrate and a dielectric layer, International Journal of Solids and Structures,
44 (2007), pp. 5829-5847.
60.
Kovacs G, Lubking GW, Vellekoop MJ, Venema A. Love waves for (bio)chemical sensing
in liquids. In: proceedings of the IEEE Ultrason.Symp., Tucson, USA. 1992, 20-23 Oct. 1992
Tucson, AZ, USA, USA DOI: 10.1109/ULTSYM.1992.275995.
Gizeli E, Goddard NJ, Lowe CR, Stevenson AC. A Love plate biosensor utilising a polymer
61.
layer. Sens. Actuators, B 1992;6 131-137, https://doi.org/10.1016/0925-4005(92)80044-X
Shyam Trivedi Harshal B. Nemade, Simulation of a Love wave device with ZnO nanorods
62.
for high mass sensitivity, Ultrasonics, Volume 84, March 2018, pp. 150-161.
Sheng-Yuan Chu, Walter Water, Jih-Tsang Liaw, An investigation of the dependence of
63.
ZnO film on the sensitivity of Love mode sensor in ZnO/quartz structure, Ultrasonics, Volume 41,
Issue 2, March 2003, pp. 133-139, https://doi.org/10.1016/S0041- 624X(02)00430-4.
64. Mihaela Puiu, Ana-Maria Gurban, Lucian Rotariu, Simona Brajnicov, Cristian Viespe, and
Camelia Bala, Enhanced Sensitive Love wave Surface Acoustic Wave Sensor Designed for
Immunoassay Formats, Sensors (Basel). 2015; 15(5): 10511 -- 10525, doi: 10.3390/s150510511.
51
David A. Powell, Kourosh Kalantar-zadeh, Samuel Ippolito and Wojtek Wlodarski , A
65.
Layered SAW Device Based on ZnO/LiTaO3 for Liquid Media Sensing Applications, 2002 IEEE
Ultrasonics Symposium, pp. 493-496.
66.
Feng-mei Zhou, Zhe Li, Li Fan, Shu-yi Zhang, Xiu-jiShui, Experimental study of Love-
wave immunosensors based on ZnO/LiTaO3 structures, Ultrasonics, Volume 50, Issue 3, March
2010, Pages 411-415, https://doi.org/10.1016/j.ultras.2009.09.024.
67.
Santanu Manna, Santimoy Kundu and Shishir Gupta, Propagation of Love type wave in
piezoelectric layer overlying non-homogeneous half-space, Electronic Journal of Mathematics and
Technology, vol. 8, no. 1, 2014.
Cinzia Caliendo, Smail Sait and Fouad Boubenider, Love-Mode MEMS Devices for Sensing
68.
Applications in Liquids, https://doi.org/10.3390/mi7010015, Micromachines 2016, 7(1), 15;
69. M. Hamidullah, C. Caliendo, F. Laidoudi, Love wave sensor based on PMMA/ZnO/Glass
structure for liquids sensing, in: Proceedings of the Third Int. Electron. Conf. Sens. Appl., 15-30
November 2016; Sciforum Electronic Conference Series, 3, 2017, p. 20, doi:10.3390/ecsa-3-
C005.5pp.
70. Maria Isabel Rocha Gaso, Yolanda Jimenez, Laurent A. Francis and Antonio Arnau, Love
wave Biosensors: A Review, "State of the Art in Biosensors - General Aspects", book edited by
Toonika Rinken, ISBN 978-953-51-1004-0, Published: March 13, 2013 under CC BY 3.0 license.
71. Wen Wang and Shitang He, A Love wave Reflective Delay Line with Polymer Guiding
Layer for Wireless Sensor Application, Sensors (Basel). 2008 Dec; 8(12): 7917 -- 7929, doi:
10.3390/s8127917.
72. Wallnofer, W.; Krempl, P.W.; Asenbaum, A. Determination of the elastic anti photoelastic
constants of quartz-type GaPO4 by Brillouin scattering. Phys. Rev. B 1994, 49, 10075 -- 10080
Reiter, C.; Krempl, P.W.; Wallnofer, W.; Leuprecth, G. GaPO4: A Critical Review of
73.
Material Data. In Proceedings of the 9th EFTF, Besancon, France, 8 -- 10 March 1995.
74. C. Caliendo, M. Hamidullah, GaPO4-based SHAPM Sensors for Liquid Environments, book
chapter in Sensors and Applications in Measuring and Automation Control Systems, Book Series:
Advances in Sensors: Reviews, Vol. 4, 2017.
75.
J. C. Andle; M. G. Schweyer; L. A. French; J. F. Vetelino, Acoustic plate mode properties of
rotated Y-cut quartz, 1996 IEEE Ultrasonics Symposium. Proceedings, 1996, Volume: 2, Pages:
971
San Antonio, TX, USA, USA DOI:
10.1109/ULTSYM.1996.584153.
3-6 Nov.
vol.2
-
976
1996
76.
C. Jeffrey; M. G. Schweyer; J. F. Vetellino, Acoustic plate mode properties of rotated Y-cut
quartz , Proceedings of 1996 IEEE International Frequency Control Symposium , 1996 Pages: 532
-- 539 5-7 June 1996 Honolulu, HI, USA, USA DOI: 10.1109/FREQ.1996.559922.
C Caliendo, A D'Amico, P Verardi, E Verona, K+ detection using shear horizontal acoustic
77.
modes, Ultrasonics Symposium, 1990. Proceedings., IEEE 1990, 383-387]
52
Corinne Dejous, Michel Savart, Dominique Rebiere, Dominique Rebiere, J. Pistre, J. Pistre,
78.
A shear-horizontal acoustic plate mode (SH-APM) sensor for biological media, 1995, Sensors and
Actuators B Chemical 26(1-3), pp. 452-456.
79.
N. Pavlakovic, M. J. S. Lowe, D. N. Alleyne, and P. Cawley, Disperse: A general purpose
program for creating dispersion curves, in D. O. Thompson and D. E. Chimenti editors, Review of
Progress in Quantitative NDE, vol.16, pp. 185-192, Plenum Press, New York, (1997).
80. Martin, A. Ricco, T. Niemczyk, and G. Frye, "Characterization of SH acoustic plate mode
liquid, J. Appl. Phys. 64 (10), 5002-8 .1988;
81. M.GSchweyer , J.Candle , D.JMcAllister, J.FVetelino , An Acoustic Plate Mode Sensor for
Aqueous Mercury, Sensors and Actuators B: Chemical, Volume 35, Issues 1 -- 3, September 1996,
Pages 170-175, https://doi.org/10.1016/S0925-4005(97)80049-4.
J. Andle, M. Schweyer, J. Munson, R. Roderick, D. McAllister, L. French, J. Vetelino, C.
82.
Watson, J. Foley, A. Bruce, M. Bruce, Electrochemical Piezoelectric Sensors for trace Ionic
Contaminants, IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, Vol. 45,
Issue: 5, 1998, Page(s): 1408 -- 1415, DOI: 10.1109/58.726469.
83.
F.G. Tseng ; K.C. Leou ; L.C. Pan ; Y.Y. Lai ; Y.C. Liang ; L.D. Chen , Acoustic plate
mode tissue sensor, SENSORS, 2002 IEEE, 12-14 June 2002, Orlando, FL, USA, USA, DOI:
10.1109/ICSENS.2002.1037099.
84.
Viktorov, I., A. Rayleigh and Lamb Waves. Plenum Press New York, 1967.
85.
J.L. Rose, Ultrasonic Waves in Solid Media, Cambridge University Press, 1999.
I. V. Anisimkin, A new type of acoustic modes of vibration of thin piezoelectric plates:
86.
https://doi.org/10.1134/1.1776212.
C. Caliendo, Longitudinal Modes along Thin Piezoelectric Waveguides for Liquid Sensing
87.
Applications, Sensors 2015, 15, 12841-12856; doi:10.3390/s150612841.
88.
Caliendo, C.; Giovine, E.; Hamidullah, M. Theoretical Study of Quasi-Longitudinal Lamb
Modes in SiN/c-AlN Thin Composite Plates for Liquid Sensing Applications. Proceedings 2018, 2,
134.
Cinzia Caliendo, Analytical Study of the Propagation of Fast Longitudinal Modes along wz-
89.
BN/AlN Thin Acoustic Waveguides, Sensors (Basel). 2015 Feb; 15(2): 2525 -- 2537, doi:
10.3390/s150202525.
Cegla, Frederic Bert, Ultrasonic waveguide sensors for fluid characterisation and remote
90.
sensing, Thesis (Ph.D.), Imperial College London, London UK. (2006), p. 248.
91.
Onursal Onen, Dispersion and Sensitivity Analysis of Quasi-Scholte Wave Liquid Sensing
by Analytical Methods, Hindawi, Journal of Sensors, Volume 2017, Article ID 9876076, 9 pages,
https://doi.org/10.1155/2017/9876076.
92.
A. Takiy, S. Granja, R. Higuti, C. Kitano, L. Elvira, O.Martinez-Graullera and F. Montero
de Espinosa, et al., Theoretical Analysis and Experimental Validation of the Scholte Wave
Propagation in Immersed Plates for the Characterization of Viscous Fluids, 2013 Ieee International
53
Ultrasonics Symposium, 21-25 July 2013, Prague, Czech Republic, pp.1610 -- 1613, DOI:
10.1109/ULTSYM.2013.0411.
93.
Onursal Onen,Yusuf Can Uz, Investigation of Scholte and Stoneley Waves in Multi-layered
Systems, Physics Procedia Volume 70, 2015, Pages 217-221, part of Special Issue Proceedings of
the 2015 ICU International Congress on Ultrasonics, Metz, France, Edited by Nico F. Declercq,
doi.org/10.1016/j.phpro.2015.08.138.
94. White, R. M., and S. W. Wenzel. "Fluid loading of a Lamb‐wave sensor." Applied Physics
Letters 52.20 (1988), pp. 1653-1655.
95. Martin, Bret A., Stuart W. Wenzel, and Richard M. White. "Viscosity and density sensing
with ultrasonic plate waves." Sensors and Actuators A: Physical 22.1-3 (1990), pp. 704-708.
96.
Luginbuhl, Philippe, et al. "Microfabricated Lamb wave device based on PZT sol-gel thin
film for mechanical transport of solid particles and liquids." Journal of Microelectromechanical
systems 6.4 (1997), pp.337-346.
97. Wu, Junru, and Zhemin Zhu. "Sensitivity of Lamb wave sensors in liquid sensing." IEEE
transactions on ultrasonics, ferroelectrics, and frequency control 43.1 (1996): 71-72.
98.
Laurent, Thierry, et al. "Lamb wave and plate mode in ZnO/silicon and AlN/silicon
membrane: Application to sensors able to operate in contact with liquid." Sensors and Actuators A:
Physical 87.1 (2000), pp. 26-37.
99. Mirea, Teona, et al., Influence of liquid properties on the performance of S 0‐mode Lamb
wave sensors II: Experimental validation, Sensors and Actuators B: Chemical 229 (2016): 331-337.
100. Kanazawa K K and Gordon J G II 1985 Frequency of a quartz microbalance in contact with
liquid Anal. Chem. 57 1770-1
101. Vivian Aubert, Regis Wunenburger, Tony Valier-Brasier, David Rabaud, Jean-Philippe
Kleman and Cedric Poulain, A simple acoustofluidic chip for microscale manipulation using
evanescent Scholte waves, Lab Chip, 2016,16, 2532-2539 , DOI: 10.1039/c6lc00534a.
102. B. Jakoby. M.J. Vellekoop, Viscosity sensor based on Love waves. Proc. Eurosensors XI,
Warsaw, Poland, 1997, p. 919.
54
|
1910.06064 | 1 | 1910 | 2019-09-17T13:42:01 | The mechanics of solid-state nanofoaming | [
"physics.app-ph"
] | Solid-state nanofoaming experiments are conducted on two PMMA grades of markedly different molecular weight using CO$_2$ as the blowing agent. The sensitivity of porosity to foaming time and foaming temperature is measured. Also, the microstructure of the PMMA nanofoams is characterised in terms of cell size and cell nucleation density. A one dimensional numerical model is developed to predict the growth of spherical, gas-filled voids during the solid-state foaming process. Diffusion of CO$_2$ within the PMMA matrix is sufficiently rapid for the concentration of CO$_2$ to remain almost uniform spatially. The foaming model makes use of experimentally calibrated constitutive laws for the uniaxial stress versus strain response of the PMMA grades as a function of strain rate and temperature, and the effect of dissolved CO$_2$ is accounted for by a shift in the glass transition temperature of the PMMA. The maximum achievable porosity is interpreted in terms of cell wall tearing and comparisons are made between the predictions of the model and nanofoaming measurements; it is deduced that the failure strain of the cell walls is sensitive to cell wall thickness. | physics.app-ph | physics | Preprint of http://doi.org/10.1098/rspa.2019.0339
The mechanics of solid-state nanofoaming
Frederik Van Loock1, Victoria Bernardo2, Miguel Angel Rodríguez Pérez2,
Norman A. Fleck1*
1. Engineering Department, University of Cambridge, Trumpington Street, CB2 1PZ
Cambridge, United Kingdom
2. Cellular Materials Laboratory (CellMat), Condensed Matter Physics Department,
University of Valladolid, Paseo de Belen 7, 47011, Valladolid, Spain
*Corresponding Author
E-mail address: [email protected] (N.A. Fleck)
Abstract
Solid-state nanofoaming experiments are conducted on two PMMA grades of markedly
different molecular weight using CO2 as the blowing agent. The dependence of porosity of the
nanofoams upon foaming time and foaming temperature is measured. Also, the microstructure
of the PMMA nanofoams is characterized in terms of cell size and cell nucleation density. A
one dimensional numerical model is developed to predict the growth of spherical, gas-filled
voids during the solid-state foaming process. Diffusion of CO2 within the PMMA matrix is
sufficiently rapid for the concentration of CO2 to remain almost uniform spatially. The foaming
model makes use of experimentally calibrated constitutive laws for the PMMA grades, and the
effect of dissolved CO2 is accounted for by a shift in the glass transition temperature of the
PMMA. The observed limit of achievable porosity is interpreted in terms of cell wall tearing;
it is deduced that the failure criterion is sensitive to cell wall thickness.
Keywords: solid-state foaming, PMMA nanofoams, molecular weight, void growth model,
porosity limit, deformation maps
1. Introduction
Polymeric nanofoams are polymer foams with an average cell size of below 1 micrometer [1].
This relatively new class of porous solids has the potential to offer unique and attractive
combinations of thermal, mechanical, and optical properties [2 -- 4]. For example, the thermal
1
Preprint of http://doi.org/10.1098/rspa.2019.0339
conductivity of polymeric nanofoams can be lower than that of air (= 0.025 W m-1K-1). When
the average cell size is in the order of the mean free path of the gas molecules in the cells (close
to 70 nm for air at standard conditions), the thermal conductivity of the gas in the foam is
significantly reduced due to the Knudsen effect [5,6].
Wang et al. [7] calculated that a polymeric nanofoam has a thermal conductivity close to or
below 0.025 W m-1K-1 when the average cell size l is below 200 nm and the porosity
exceeds
0.85. To achieve this morphology, the cell nucleation density
must exceed 1021 m-3 [1]. A
large number of experimental studies focus on the effect of processing conditions and the
choice of polymer precursor upon the cell nucleation density
, the void size l and the
porosity of polymeric nanofoams, as reviewed by Costeux [1]. Many of these studies make
use of the solid-state foaming method in which a physical blowing agent (e.g. CO2) is used to
nucleate and grow cells in a polymer matrix such as polymethyl methacrylate (PMMA) [8,9].
The review by Costeux [1] on experimental studies of solid-state nanofoaming discusses the
trade-off between porosity and cell size of a polymeric nanofoam. Polymeric nanofoams of l <
200 nm are reported for a nucleation density above 1021 m-3, but their porosity is limited to
close to 0.85, see, for example, Aher et al. [10] and Costeux and Zhu [11].The observed
porosity limit for nanofoams with a nucleation density above 1021 m-3 may be due to the fact
that the walls between the nano-sized cells are limited by the end-to-end distance of the
individual polymer chains [1,12]. Polymeric nanofoams of porosity on the order of 0.8 to 0.9
have been produced, but their cell size is well above 200 nm (and
<< 1021 m-3 ) [1,9,13].
The microstructural requirement for polymeric nanofoams with a thermal conductivity lower
than the thermal conductivity of air is currently beyond the practical limit of the state-of-the-
art solid-state nanofoaming process [7].
The final porosity and final cell size in solid-state nanofoaming can be predicted by simulating
void growth. In contrast to the substantial body of experimental work on polymeric nanofoams
produced by solid-state foaming, as reviewed by Costeux [1], and the development of cell
growth models for liquid state foaming processes [14 -- 16], theoretical studies on cell growth
during solid-state nanofoaming are limited. Costeux and co-workers [13,17] simulated void
nucleation and void growth during the solid-state nanofoaming of acrylate co-polymers by
making use of the model of Shafi et al. [18]. They conducted a series of nanofoaming
2
fdNdNfdN Preprint of http://doi.org/10.1098/rspa.2019.0339
experiments but their model overestimates the measured final porosity of the nanofoams. The
mismatch between the simulated and the measured porosity of acrylic nanofoams may be due
to (i) the assumption that cell growth continues until the foaming temperature attains the glass
transition temperature of the polymer-gas solid and/or (ii) the assumption that the polymer-gas
solid surrounding the cell is in a liquid (viscous) state throughout the solid-state foaming
process. In reality, void growth occurs at a temperature above and below the glass transition
temperature of the solid surrounding the void. This is addressed in detail in the present study.
Scope of study
PMMA nanofoams are produced from two PMMA grades of widely different molecular
weight; a solid-state foaming process is used with CO2 as the blowing agent. We characterize
the microstructure of the nanofoams in terms of porosity
, cell size
, and cell nucleation
density
. In addition, we develop a void growth model, based on the constitutive law of
PMMA grades close to the glass transition temperature, by building on the recent study of Van
Loock and Fleck [19]. Both predicted and measured final porosities are obtained as a function
of foaming time and foaming temperature; also, cell wall tearing mechanisms are considered
in order to account for the observed limit in final porosity.
2. Nanofoaming experiments
2.1 Materials
Foaming experiments were conducted on two PMMA grades: pelletized PMMA (Altuglas
V825T) of average molecular weight1
= 92 500 g mol-1 and cast PMMA sheets (Altuglas
CN with sheet thickness close to 3 mm) with
= 3 580 000 g mol-1. We shall refer to the
Altuglas V825T and Altuglas CN grades as 'low
PMMA' and 'high
PMMA',
respectively. Both grades have a density
equal to 1 190 kg m−3 (as measured at 23 °C and
at 50% relative humidity). The glass transition temperature (
= 114.5 C) of the low
1 The average molecular weight was measured by gas permeation chromatography (GPC)
with an Agilent Technologies PL GPC220 (USA) instrument with a nominal flow rate equal
to
at a test temperature equal to 30 C.
3
fldNwMwMwMwMpgTwM-511.6710 l s− Preprint of http://doi.org/10.1098/rspa.2019.0339
PMMA is close to the glass transition temperature (
= 116.5 C) of the high
PMMA as
measured by differential scanning calorimetry (DSC) using a heating rate of 10 C min-1.
2.2 Solid-state nanofoaming experiments
Foaming precursors of the low
and high
PMMA grades were made as follows. The
low
PMMA pellets were heated to 250 C for 450 s, then compressed for 60 s between two
heated plates at a pressure equal to 17 MPa. The resulting sheet was cooled to room temperature
with the pressure of 17 MPa maintained. Cuboid precursors of dimension 20x10x3 mm3 were
machined from the low
PMMA sheet and from the as-received high
PMMA sheet.
Foaming experiments were performed using a pressure vessel2 with a pressure controller3 and
temperature controller4. Medical grade CO2 (> 99.9% purity) was used as the blowing agent
for the foaming experiments. The two step solid-state foaming process was employed, as
detailed in the study of Martin-de León et al. [9]. The precursor samples were held in the
pressure vessel at a constant CO2 saturation pressure equal to 31 MPa, and at a constant
temperature equal to 25 C for 24 hours in order to ensure saturation of the CO2 into the PMMA.
The mass concentration5 C, at equilibrium, is close to 24 wt% for both the low and high
PMMA, according to the measurement procedure detailed by Martin-de León et al. [9]. Next,
the pressure was progressively released to atmospheric pressure with an instantaneous pressure
drop rate close to 100 MPa s-1. The samples were then foamed in a foaming bath6 at selected
foaming temperatures (25 C, 40 C, 60 C, 80 C, 100 C) and selected foaming times7 (60 s, 180
s, 300 s, and 600 s). It is assumed in the remainder of the study that the foaming times are
sufficiently long for the temperature to be spatially uniform8 within the sample.
2 Pressure vessel model PARR 4681 of Parr Instrument Company (USA).
3 Pressure controller pump SFT-10 of Supercritical Fluid Technologies Inc (USA).
4 Temperature controller CAL 3300 of CAL Controls Ltd (UK).
5 We define the mass concentration C of CO2 in PMMA with respect to the total mass of the
PMMA-CO2 mixture. Note that the definition of CO2 solubility (with respect to the mass of
the PMMA absent CO2) is used in the work of Martin-de León et al. [9].
6 Thermal bath J.P. Selecta Model 6000685 of Grupo Selecta (Spain). The time between the
pressure release and the start of foaming was close to 120 s.
7 Samples were immersed in a water bath at a temperature close to 10 C at the end of the
foaming time.
8 The justification for this assumption is as follows. Immersion of the sample in water or oil
provides excellent heat transfer at the surface of the sample. The time constant
≈
4
gTwMwMwMwMwMwMwM2/x= Preprint of http://doi.org/10.1098/rspa.2019.0339
2.3 Characterization of the PMMA nanofoams
Porosity
The density
of the foamed samples was determined by the water-displacement method with
a weight balance9. A surface layer of depth 200
was removed by polishing10 to ensure that
the solid skin (of thickness below 100
) was absent before the density measurements were
made. The porosity
of the samples is obtained by:
(1)
where
( = 1 190
) is the density of solid PMMA.
Microstructure
Foamed samples were cooled in liquid nitrogen and then fractured. The fracture surfaces were
coated with a layer of gold by sputtering11, and micrographs of the coated fracture surfaces
were taken by a scanning electron microscope12 (SEM). The cellular structure of each material
was characterised by analysing the micrographs with dedicated in-house software based on
ImageJ/FIJI [20]. Microstructural parameters such as the average cell size l, standard deviation
of the observed cell sizes, and cell nucleation density
, using the method as suggested by
Kumar and Suh [21], were obtained13.
Open cell content
The open cell content of the foamed samples was measured by gas pycnometry14 with nitrogen
in accordance with the ASTM D6226-15 standard [23]. The open cell content ratio
is
= 1.5 mm is the half-thickness of the PMMA sample and
20 s where
is the thermal diffusivity of PMMA at room temperature [22].
9 Analytical balance AT261 of Mettler-Toledo (USA).
10 Grinding and polishing system LaboPOl2-LaboForce3 of Struers (USA).
11 Sputter coater SDC 005 of Balzers Union (Liechtenstein).
12 Scanning electron microscope QUANTA 200 FEG of Thermo Fisher Scientific (USA).
13 At least 200 cells were analysed from multiple micrographs per foamed sample.
14 Gas pycnometer (USA) AccuPyc II 1340 of Micromeritics (USA).
m2 s-1
5
fμmμmffp1f=−p-3kg msdNvOx71.110−= Preprint of http://doi.org/10.1098/rspa.2019.0339
defined as the ratio of the volume of open pores to the total pore volume of a foam, and is
obtained by:
(2)
where
is the geometric volume of the foam,
is the pycnometer volume and
is a
penalty volume to account for the volume of the cells at the surface of the foam. The penalty
volume
is assumed to be close to zero in the case of nanofoams. The geometric volume
is measured by the water-displacement method as detailed above. Foamed samples were
subjected to a pressure scan from 0.02 MPa to 0.13 MPa in the gas pycnometer. The
pycnometer volume initially decreases as the gas pressure increases until the interconnected
open cells are completely filled with gas and the pycnometer volume remains constant at
increased pressures. We take this constant value of pycnometer volume
in order to calculate
via Eq. (2).
3. Results nanofoaming experiments
The measured porosity
, average observed cell size , standard deviation
of observed cell
sizes, and cell nucleation density
of the nanofoams are reported in Tables 1 and 2 for the
low
and high
grades of PMMA, respectively. In addition, a representative series of
SEM micrographs of the nanofoams are shown in Figure 1. The low
and the high
nanofoams have contrasting microstructures and the cell nucleation density of the low
nanofoams (
) is an order of magnitude less than that of the high
nanofoams (
). The average cell size of the high
nanofoams ranges from
20 nm to 50 nm, and is an order of magnitude smaller than the average cell size of the low
nanofoams (of size 200 nm to 350 nm). These values of
and
for the low
nanofoams are consistent with the results of Martin-de León et al. [9] who conducted solid-
state foaming experiments with an identical low
PMMA grade. The measured average
cell size
of the low
and the high
nanofoams, as a function of foaming time
for
= 60 C, is plotted in Figure 2a. Void growth typically occurs over a foaming time period of
6
gpsvgVVVOfV−−=gVpVsVsVgVpVvOflsdNwMwMwMwMwM203d210mN−wM213d210mN−lwMwMldNwMwMlwMwMftfT Preprint of http://doi.org/10.1098/rspa.2019.0339
60 s to 180 s, followed by arrest. There is a mild dependence of the foaming temperature
upon the final value for , see Tables 1 and 2.
The measured porosity
of the nanofoams is plotted as a function of
in Figure 2b for
= 60 C and for
= 100 C. Consistent with the versus
curves for
= 60 C, as presented
in Figure 2a, the porosity increases over a foaming period of 60 s to 180 s until a stable (
-
independent) value of final porosity is achieved. The highest observed porosity of the low
PMMA nanofoams (
) is approximately 25% higher than that of the high
PMMA nanofoams (
). At a foaming temperature of
= 100 C, the porosity
decreases with increasing foaming time beyond
= 60 s, and this is due to collapse of the
foamed structure. This behavior is also illustrated in plots of f versus
, over the explored
range of foaming times, see Figures 2c and 2d for the low
and high
PMMA
nanofoams, respectively.
The measured open cell content
is plotted as a function of the measured porosity
in
Figure 3a (low
) and in Figure 3b (high
) for 20 C
80 C. Nanofoams with
porosities well below the highest observed porosity
are closed-cell in nature. An abrupt
transition to an open-celled structure occurs close to
. The observed collapse of the foam
at
= 100 C is preceded by cell wall failure for the low
nanofoams (see Figure 1b) and
by the formation of cracks interconnecting the nano-sized pores in for the high
nanofoams
(see Figure 1d).
4. Void growth model
A void growth model is now developed to predict porosity as a function of foaming time and
foaming temperature for the PMMA nanofoams. The expansion of a pre-existing as-nucleated
cavity during solid-state nanofoaming is simulated by means of a one dimensional single cell
growth model [15,24]. A finite shell surrounds the void in order to account for void-void
interaction in an approximate manner. More sophisticated models of an array of voids (such as
periodic cell models) could be adopted but the intent here is to emphasize the strong role of the
evolving constitutive response. Consider a polymer-gas solid with equisized spherical voids.
A cross-section of the undeformed (reference) configuration of the spherical void, with initial
7
fTlfftfTfTlftfTftwMmax0.75f=wMmax0.60f=fTftfTwMwMvOfwMwMf TmaxfmaxffTwMwM Preprint of http://doi.org/10.1098/rspa.2019.0339
radius
and initial outer radius
, along with the adopted spherical coordinate system
, is shown in Figure 4. Assume that the initial gas pressure
in the as-nucleated void
equals the saturation pressure during the saturation phase prior to nucleation of the voids. The
deformed configuration for the void of inner radius a and outer radius
at time t is shown in
Figure 4.
Kinematics
Assume that the void remains spherical during growth and that the solid surrounding the void
is incompressible. Then a material point, initially at radius R, is displaced to a radius r such
that:
by incompressibility. For later use, this relation is re-arranged to the form:
(3)
(4)
Note that
is a function of the time-like variable (
) and of the Lagrangian position
variable
. The von Mises effective strain
is defined in the usual manner as
, giving:
where
is the hoop strain. Now insert Eq. (4) into Eq. (5) to obtain:
and take the time derivative of r in Eq. (3) to give:
(5)
(6)
(7)
8
0a0b(,,)r0pb33330raRa−−=3330011araRRa=+−/rR0/aa0/Rae2e23ijij=e22lnrR==330e02l1n13aaRa+=−2rrrvaa== Preprint of http://doi.org/10.1098/rspa.2019.0339
where
is the radial velocity of a material element at r. Consequently, the effective strain rate
reads:
Equilibrium
(8)
Write (
,
,
) as the active stress components in the spherical coordinate system.
Radial equilibrium dictates that [25]:
Due to symmetry,
and Eq. (9) simplifies to:
(9)
(10)
where
is the von Mises effective stress [26]. Integration of Eq. (10) leads to:
(11)
where
is the gas pressure inside the cavity for a given radius a, and
is the ambient
pressure. Upon making use of Eq. (3), the above integral can be re-phrased as:
(12)
The effective stress
is a function of the effective strain
, the effective strain rate
, and
the normalized temperature
via the constitutive law for the PMMA-CO2 solid, of general
functional form F where:
(13)
9
rve32e32rarRRvar−==rr()120rrrrrr+−−==()22rrerrrrr=−=err=−ead2rarbrppr==−=pap003ae2dRaRbRpRpRr==−=eeeg/TT()eeegF,,/TT= Preprint of http://doi.org/10.1098/rspa.2019.0339
The choice of F is given below. We emphasize that the glass transition temperature
of the
PMMA is a function of CO2 concentration and that the effective strate rate
scales with
as prescribed in Eq. (8). Upon substituting Eq. (13) into Eq. (12) we obtain an expression for
, and we integrate over time in order to determine the time evolution of
.
Mass conservation
At the start of the foaming process, the chemical potential of the CO2 molecules in the nucleated
voids is lower than chemical potential of CO2 molecules in the PMMA-CO2 solid.
Consequently, CO2 gas molecules migrate from the PMMA-CO2 solid into the voids. The
concentration of CO2 gas molecules
at time t and position r (for
) can be
obtained by solving Fick's second law of diffusion [27]:
(14)
in the deformed configuration, where
is the diffusion coefficient for CO2 in PMMA.
Measurements of
at temperatures and pressures typical for solid-state nanofoaming of
PMMA by CO2 are available in the literature as follows. Guo and Kumar [28] measured D
based on desorption measurements and found that D ranges from
to
for temperatures ranging from -30C to 100 C at a CO2 pressure equal
to 5 MPa. Li et al. [29] measured
by a sorption technique and found that D is in the range
of
to
for temperatures ranging from 30C to 70 C, and
pressures ranging from 6 MPa to 18 MPa. Now, introduce a characteristic diffusion time
:
(15)
where
is a diffusion length which is approximated for the void growth problem by:
(16)
Observations of cell nucleation densities of PMMA nanofoams (
m-3) imply
<
133 nm [1]. Upon assuming
, we obtain
20 ms via Eq. (15), which is two
10
gTeaa0/aa(,)Crtarb22CDCrtrrr=DD122-12.510 m sD−=112-13.6510 m sD−=D112-1610 m s−112-19.510 m s−D2D()DLD=DL0d1343DNLb20d10NDL12210 m/sD−=D Preprint of http://doi.org/10.1098/rspa.2019.0339
orders of magnitude lower than typical observed cell growth times for solid-state nanofoaming
of PMMA by CO2 as reported by Martín-de León et al. [9] and reported above. We conclude
that the CO2 concentration profile
is spatially uniform at all times:
.
Consequently, we do not need to solve the diffusion equation to predict void growth during
solid-state nanofoaming of PMMA by CO2.
We shall assume that the mass of gas molecules in the voids and in the surrounding solid is
constant; leakage of gas molecules to neighbouring voids or the sample's environment is
neglected. The resulting mass conservation statement for CO2 reads:
(17)
where
is the density15 of the PMMA-CO2 solid and
is the density of the CO2 in the voids.
The relations Eq. (12) and Eq. (17) form a coupled pair of equations which can be solved to
obtain
as a function of cavity expansion
. Additional assumptions are detailed below.
Depression of the glass transition temperature by CO2
The dissolution of CO2 into a linear, amorphous polymer such as PMMA reduces the glass
transition temperature
of the PMMA-CO2 solid. This plasticization effect is attributed to
the increased mobility of PMMA chains due to lubrication by the CO2 molecules, and the
decrease of the intermolecular bond strength, as the CO2 molecules increase the spacing
between the PMMA chains [30,31]. A range of experimental techniques have been used in the
literature to determine the glass transition temperature Tg of PMMA as a function of CO2 mass
concentration
. Chiou et al. [32] made use of DSC to measure
as a function of
,
where
. Likewise, Wissinger and Paulaitis [33] measured the dependence of
upon
via creep compliance measurements. Guo and Kumar [28] made use of solid-
state foaming experiments to observe the relation between
and CO2 for a PMMA-CO2
15 We assume that the density of the PMMA-CO2 solid is equal to the density of PMMA
absent CO2 at standard conditions (i.e.
) based on the measurements of
Pantoula and Panayiotou [40] and Pantoula et al. [41] who observed that the relative increase
in volume of a PMMA-CO2 mixture is close to the relative increase of the mass of a PMMA-
CO2 mixture for a CO2 pressure up to 30 MPa.
11
(,)CRt(,)()CRtCt=()()p33g3p33g300000 + aCbCbaaa−=−+pgp0/aagTC0gg/TTC0gg(0)TTC==0gg/TTC0gg/TT-31190 kg mp= Preprint of http://doi.org/10.1098/rspa.2019.0339
mixture. The measured
versus
data, for PMMA-CO2, as reported by Chiou et al.
[32], Wissinger and Paulaitis [33], and Guo and Kumar [28] are shown in Figure 5. Chow [34]
used statistical thermodynamics to predict
as a function of
and introduced a
parameter where:
(18)
Here,
is the molecular weight of the polymer repeat unit (
= 100.12 g mol-1 for a
methyl methacrylate monomer),
is the molecular weight of the gas (
= 44.01 g mol-1
for CO2), and z is a lattice coordination number equal to 2, as suggested by Chow [34]. In
addition, Chow [34] defined a parameter
:
(19)
where
is the universal gas constant and
is the change in specific heat capacity of the
polymer at the glass transition temperature at constant pressure. The normalized glass transition
temperature is then predicted by:
(20)
Equation (20) is curve fitted to the measured
versus C data shown in Figure 5 by a
suitable choice of
. The fitted value for
equals 355
which is slightly
higher than the value of
for PMMA as measured by DSC, see Chiou et al. [32] and Li et
al. [35].
Constitutive model for the PMMA-CO2 solid
We assume that the effective stress
of the PMMA-CO2 solid at a given strain
, strain rate
and normalized temperature
is the same as that given by PMMA in the absence of
CO2: the effect of CO2 is accounted for by a shift in the value for Tg. The deformation
mechanisms for PMMA in uniaxial tension close to the glass transition temperature have been
12
0gg/TTC0gg/TTCgpww1MCzMC=−pwMpwMgwMgwMpwp CzRM=RpC()()()g0gexp1ln1lnTT=−−+0/ggTTpCpC11J kg K−−pCeeeg/TT Preprint of http://doi.org/10.1098/rspa.2019.0339
reviewed recently by Van Loock and Fleck [19] an deformation mechanism maps were
constructed by performing a series of uniaxial tension tests on the high
PMMA over a
range of temperatures near the glass transition and over two decades of strain rate. The
operative deformation mechanism depends upon the temperature
, the strain rate
, and
strain
. We shall make use of the constitutive models as calibrated by Van Loock and Fleck
[19] for the high
PMMA: the Ree-Eyring equation and a rubbery-flow model. For the low
PMMA it is necessary to construct an alternative deformation mechanism map. This is
reported in the Appendix. For this grade, the relevant deformation mechanisms are Ree-Eyring
and viscous flow.
The Ree-Eyring equation relates
in the glassy and glass transition regime to temperature
and strain rate
:
(21)
where
is a reference strain rate,
is an activation energy,
is an activation volume, and
k is Boltzmann's constant. Visco-elastic effects are neglected in this finite strain regime. Van
Loock and Fleck [19] also fitted an empirical equation to relate
to
and
in the
rubbery regime for the high
PMMA:
(22)
where
is a reference modulus,
is a temperature sensitivity coefficient,
a reference
strain rate, and
a strain rate sensitivity coefficient.
Note that the rubbery regime above the glass transition is absent for PMMA grades of relatively
low molecular weight, i.e.
< 150
[36]. Instead, a linear, viscous flow rule can be
used to describe the constitutive behavior of a low
PMMA for
:
where
is a temperature-dependent viscosity [37,38]:
(23)
13
wMg/TTeewMwMeg/TTeee0-sinhexpvqkTkT=0qveg/TTewMee0RegR1nTET=−0RERRnwM-1kg molwMg/1TTee3= Preprint of http://doi.org/10.1098/rspa.2019.0339
(24)
in terms of a reference viscosity
at
= 1;
and
are fitting constants.
The dependence of the effective stress
upon normalized temperature
and strain rate
is assumed to be governed by Eq. (21) and Eq. (22) for the high
PMMA and by Eq.
(21) and Eq. (23) for the low
PMMA. The fitted parameters for the constitutive laws for
the high
PMMA are taken from Van Loock and Fleck16 [19] and are summarized in Table
3. An additional series of tensile tests have been performed on the low
PMMA at
temperatures close to the glass transition in order to calibrate Eq. (21) and Eq. (23) for the low
PMMA as detailed in the Appendix. The resulting calibrated parameters for Eq. (21) and
Eq. (23) for the low
PMMA are included in Table 3.
Gas laws
The equilibrium concentration C of CO2 in PMMA is a function of CO2 pressure p and of
temperature. Here, we assume that Henry's law suffices such that [39 -- 42]:
(25)
where Henry's law coefficient
is assumed to be independent of both temperature and
pressure. Assume that the concentration of CO2 at the surface of the cavity (
) is in
equilibrium with the CO2 pressure within the void via Eq. (25). Take
for
both the low
and the
PMMA grades, based on the measured C = 0.24 equilibrium
concentration of CO2 in PMMA at a pressure p equal to 31 MPa and temperature T = 25 C, as
detailed in section 2.2. Also, assume that the CO2 gas in the void statisfies the ideal gas law:
(26)
Temperature-time profile during void growth
16 We assume that the dependence of the effective stress
of the PMMA-CO2 solid upon
pressure is small as a first order approximation for the void growth problem.
14
1g02gg-(/-1)ex-p//1TCTTCTT=+0g/TT1C2Ce/gTTewMwMwMwMwMwMHCKp=HK0Ra=91H7.7410 PaK−−=wMwMggwRTpM=e Preprint of http://doi.org/10.1098/rspa.2019.0339
During the rapid release of pressure at the end of the saturation phase, the samples cool down
from the saturation temperature equal to 25 C to a temperature17
= -15 C due to adiabatic
cooling of the expanding gas. The samples are subsequently placed in a thermal bath at a
maintained foaming temperature
. Upon submersion in the foaming bath, assume that the
temperature profile
is of the form:
(27)
where
is a time constant associated with the heat conduction into the PMMA, as measured
by a thermocouple.
Void growth simulations
Void growth during solid-state foaming is simulated by solving the equilibrium Eq. (12) and
the mass conservation statement Eq. (17) simultaneously, with due account of the dependence
of
upon C via Eq. (20), the dependence of the effective stress
of the PMMA-CO2 solid
upon
,
and
(Eqs. (21) to (23)), the gas laws (Eqs. (26) and (25)), and the time-
temperature profile as captured by Eq. (27). The resulting system of equations is solved by
numerical integration18. The values of the processing parameters and the material properties
are summarized in Table 4. Note that the initial porosity
is:
(28)
and is estimated19 to equal
for both the low
and high
PMMA nanofoams. The
initial void radius
is estimated by:
17 Measured by placing a thermocouple on the sample after pressure release at the end of the
saturation phase.
18 The numerical integration was conducted within the Matlab computing environment by
means of the ode15s function.
19 The initial porosity
is estimated by saturating low
and high
PMMA precursors
with CO2 at p = 31 MPa and T = 25 °C. Upon release of the pressure to atmospheric pressure,
the samples were immediately immersed in liquid nitrogen to prevent the growth of the
nucleated voids. The porosity of the samples was measured by the method detailed in section
2 after the CO2 was completely desorbed. The measured porosity was assumed to be
representative for
.
15
0TfT()Tt()()()00f1exp-/TTtTT−=+−gTeeeg/TT0f3000fab=310−wMwM0a0fwMwM0f Preprint of http://doi.org/10.1098/rspa.2019.0339
(29)
where the cell nucleation density
equals
for the low
PMMA nanofoams
(see Table 1) and
equals
for the high
PMMA nanofoams (see Table 2).
5. Results and discussion of the void growth predictions
Consider the deformation mechanism maps for the low
PMMA (see Figure 6a) and for the
high
PMMA (see Figure 6b). We superpose the predicted trajectory of the effective stress
at the surface of the cavity
by the void growth model as a function of
for foaming
temperatures = 25 °C and = 80 °C, and for a foaming time up to 600 s. Note that both the
temperature
and glass transition temperature
evolve in time during foaming. For both the
low
and high
PMMA, at the start of foaming,
equals
and
is close to 0.9;
at this instant
is close to 0.8 MPa for the low
PMMA and
is close to 0.3 MPa for
the high
PMMA. When the temperature increases from
to
,
rises to
almost unity and
rises steeply. The void growth simulations suggest that during solid-state
foaming of PMMA, the normalized temperature
remains between 0.9 and 1 and
consequently void growth does not occur within either the viscous regime (low
PMMA)
or within the rubbery regime (high
PMMA).
The measured porosity
is plotted as a function of foaming time
for = 25 °C to = 80
°C, and compared with the predicted
versus
curves for the low
and high
nanofoams, in Figure 7a and Figure 7b, respectively. There is reasonably good agreement
between the measured and the predicted
-
curves for = 25 °C and = 40 °C. The void
growth model overestimates the porosity at = 60 °C and at = 80 °C, where porosities close
to
are observed. Observations of SEM micrographs suggest that cell walls tear, leading to
open-celled microstructures. This is confirmed by open cell content measurements by gas
pycnometry: nanofoams with the highest observed porosities have predominantly open-celled
16
00d1334faNdN203210 m−wMdN2032010 m−wMwMwMeg/TTfTfTTgTwMwMT0Tg/TTewMewM0TT=fTT=g/TTeg/TTwMwMfftfTfTfftwMwMfftfTfTfTfTmaxf Preprint of http://doi.org/10.1098/rspa.2019.0339
microstructures, see Figure 3. At increased foaming temperatures (i.e. = 100 °C) collapse of
the foamed open-celled microstructure is observed leading to measured porosities below the
maximum observed porosities at = 80 °C, as shown in Figures 2c and 2d.
We proceed to explore two alternative hypotheses for cell wall failure which could lead to
open-celled microstructures as observed for the PMMA nanofoams: (i) achievement of a
critical hoop strain at the void, or (ii) achievement of a minimum (critical) value of ligament
thickness between neighbouring voids.
(i) Critical hoop strain
Assume that the solid surrounding the expanding void is incompressible. Then, by Eq. (3),
(30)
Recall that the initial (as-nucleated) porosity
equals
as defined in Eq. (28) and the
current porosity
equals
. Now, rearrange Eq. (30), to express f as a function of
and the true hoop strain
at the surface of the void, where
:
(31)
Tearing of the cell wall occurs when
equals the
-dependent20 true tensile failure strain
. The critical porosity
corresponding to this ductility-governed failure criterion reads:
(32)
(ii) Critical ligament size
The alternative failure hypothesis assumes that there is a minimum number of confined
polymer chains separating individual cells to prevent rupture of the solid between the cells.
Define the smallest distance between two neighbouring cells h as:
(33)
20 We assume
to be insensitive to strain rate [19,46].
17
fTfT333300--bbaa=0f300(/)abf()3/ab0fs()()s0 = ln/raaa==()()-1-1s01exp-13-ff+=s/gTTfff()()0f-1-1f1exp-13-ff+=()2 - hba=f Preprint of http://doi.org/10.1098/rspa.2019.0339
Then, upon making use of the expressions
,
, and Eq. (31), we obtain:
(34)
Write
as the critical cell wall thickness, and assume that it is independent of the value of
. The corresponding critical value of porosity
is given by Eq. (34) with
.
The ductility-governed porosity limit
as given by Eq. (32) is plotted in Figure 7 based on
the predicted hoop strain
during void growth. Note that we make use of the measured
response of
versus
(Eq. (A.2) for the low
PMMA and Eq. (A.1) for the high
PMMA as detailed in the Appendix) and assume that the initial porosity
equals
. The
measured values of final porosity
and the predictions of the void growth model exceed the
porosity limit as given by
.
We now plot the porosity limit
in Figure 7 via Eq. (34) for
by taking
= 3
(low
PMMA) and
= 4.2 (high
PMMA) in order to match to observed values
of the maximum observed porosity
of the nanofoams. Recall that the initial void size
of the low
PMMA nanofoams is estimated to be close to 10.5 nm, whereas
is close to
5 nm for the high
PMMA nanofoams. Consequently, the estimated corresponding critical
cell wall dimension
equals 32 nm for the low
PMMA nanofoams, whereas
equals
21 nm for the high
PMMA. These values for
are of the same order of magnitude as
root-mean-square end-to-end distance
of the PMMA chains, i.e.
≈ 20 nm for the low
PMMA and
≈ 110 nm for the high
PMMA based on an idealized equivalent
freely jointed chain calculation [43]. This is in agreement with the results of Crosby and co-
workers who conducted a series of uniaxial tensile tests on thin polystyrene (PS) films with
= 136 000 g mol-1 [44,45]. They found that the tensile failure strain
decreases with
decreasing film thickness in the regime = 15 nm to = 77 nm; these values are close to the
estimated value for
= 25 nm of the PS chains.
18
3000(/)fab=()3/fab=1133-1-0-10-1-1- 1 2ffhaf=ch/gTTcfchh=ffsfg/TTwMwM0f310−fffcf3010f−=c0/hawMc0/hawMmaxf0awM0awMchwMchwMcheeReeRwMeeRwMwMfttteeR Preprint of http://doi.org/10.1098/rspa.2019.0339
Concluding remarks
Solid-state nanofoaming experiments are performed with two grades of PMMA of markedly
different molecular weight (
= 92 500 g mol-1 and
= 3 580 000 g mol-1). It was found
that the molecular weight of the PMMA has a profound effect upon the microstructure of the
PMMA nanofoams. When subjected to identical foaming conditions, the observed cell size l
35 nm of the high molecular weight PMMA nanofoams is an order of magnitude less than
that of the low molecular weight PMMA nanofoams,
250 nm. This is consistent with the
observation that the nucleation density,
of the high molecular weight
PMMA nanofoams is an order of magnitude higher than that of the low molecular weight
PMMA nanofoams
. In addition, a limit in attainable porosity
was
observed:
equals 0.65 for the high molecular weight PMMA and
equals 0.75 for the
low molecular weight PMMA. The microstructure of the PMMA nanofoams transitions from
closed-celled to open-celled at a porosity close to
.
A void growth model has been developed to simulate cavity expansion during solid-state
nanofoaming of PMMA by CO2. Experimentally calibrated constitutive laws for the PMMA
grades close to the glass transition are used in the simulations. The predicted porosity of the
nanofoams versus foaming time, at selected foaming temperatures, are in good agreement with
the measured responses for porosities well below the maximum observed porosity. There is
also close agreement between the predicted and observed sensitivity to molecular weight. This
suggests that the observed difference in constitutive response close to the glass transition
between the two PMMA grades leads to the measured difference in porosity. Moreover, cell
wall tearing accounts for the observed limit in final porosity. Our analysis suggests the
existence of a limiting minimum cell wall thickness of magnitude close to that of the end-to-
end distance of the polymer chains. When the cell wall thickness approaches this minimum
value during foaming, rupture of the cell walls occurs; resulting in an open-celled structure,
and to a limit on foam expansion.
Acknowledgements
Financial support from the Engineering and Physical Sciences Research Council (UK) award
1611305 (F. Van Loock) and the FPU grant FPU14/02050 (V. Bernardo) from the Spanish
Ministry of Education is gratefully acknowledged. N. A. Fleck is grateful for additional
financial support from the ERC project MULTILAT. Financial assistance from SABIC,
19
wMwMldN2032010 m−203d210 mN−maxfmaxfmaxfmaxf Preprint of http://doi.org/10.1098/rspa.2019.0339
MINECO, FEDER, UE (MAT2015-69234-R) are acknowledged too. The authors would also
like to thank Dr. Martin van Es from SABIC for the technical assistance.
20
Preprint of http://doi.org/10.1098/rspa.2019.0339
List of table captions
Table 1: Measured porosity
, average cell size
, standard deviation of observed cell size
,
cell nucleation density
, and open cell content
of the low
PMMA nanofoams as a
function of foaming time
and foaming temperature
. Foams collapsed at
= 100
, and
so no open cell content values are reported for nanofoams produced at
= 100
.
Table 2: Measured values for the porosity
, the average observed cell size
, the standard
deviation of the observed cell sizes
, the cell nucleation density
, and the open cell content
of the high
PMMA nanofoams as a function of foaming time
and foaming
temperature
. Foams collapsed at
= 100
, and so no open cell content values are
reported for the nanofoams produced at
= 100
.
Table 3: Fitted parameters for the constitutive laws for the low
PMMA (Eq. (21) and Eq.
(23)) and the high
PMMA obtained from Van Loock and Fleck [18] , see Eq. (21) and Eq.
(22).
Table 4: Summary of the assumed processing parameters and material properties for the void
growth predictions.
21
flsdNvOwMftfTfTCfTCflsdNvOwMftfTfTCfTCwMwM Preprint of http://doi.org/10.1098/rspa.2019.0339
Tables
Table 1: Measured porosity
, average cell size
, standard deviation of observed cell size
,
cell nucleation density
, and open cell content
of the low
PMMA nanofoams as a
function of foaming time
and foaming temperature
. Foams collapsed at
= 100
, and
so no open cell content values are reported for nanofoams produced at
= 100
.
(s)
60
180
300
600
60
180
300
600
60
180
300
600
60
180
300
600
60
180
300
600
)
(
25
25
25
25
40
40
40
40
60
60
60
60
80
80
80
80
100
100
100
100
f
0.45
0.47
0.51
0.51
0.52
0.61
0.64
0.66
0.56
0.66
0.68
0.68
0.72
0.74
0.75
0.73
0.64
0.68
0.62
0.51
(nm)
s
(nm)
)
(
219
228
283
235
262
250
254
233
234
297
279
284
333
288
297
274
297
253
246
291
87
79
112
85
102
125
105
103
89
111
122
109
134
138
125
109
122
110
103
125
1.50
1.50
0.91
1.48
1.22
1.70
1.27
2.11
2.34
1.72
1.76
1.63
1.16
1.83
1.75
2.08
1.21
1.81
1.75
0.76
0.12
0.08
0.08
0.08
0.07
0.02
0.15
0.14
0.07
0.33
0.40
0.36
0.63
0.90
0.78
0.93
-
-
-
-
22
flsdNvOwMftfTfTCfTCftfTCldN03210 m−vO Preprint of http://doi.org/10.1098/rspa.2019.0339
Table 2: Measured values for the porosity
, the average observed cell size
, the standard
deviation of the observed cell sizes
, the cell nucleation density
, and the open cell content
of the high
PMMA nanofoams as a function of foaming time
and foaming
temperature
. Foams collapsed at
= 100
, and so no open cell content values are
reported for the nanofoams produced at
= 100
.
(s)
60
180
300
600
60
180
300
600
60
180
300
600
60
180
300
600
60
180
300
600
)
(
25
25
25
25
40
40
40
40
60
60
60
60
80
80
80
80
100
100
100
100
f
0.22
0.28
0.29
0.31
0.33
0.42
0.45
0.47
0.45
0.55
0.57
0.57
0.58
0.60
0.60
0.59
0.59
0.53
0.50
0.45
(nm)
s
(nm)
)
(
36
23
30
36
28
32
37
45
37
39
40
41
39
39
38
44
34
27
37
34
14
10
12
18
13
16
14
29
14
17
17
19
20
19
19
22
15
14
18
12
14.9
40.0
9.0
6.9
54.2
32.3
7.8
26.0
20.4
24.0
31.8
25.8
21.8
27.8
36.6
46.6
35.4
80.4
24.9
32.6
0.30
0.22
0.28
0.21
0.19
0.07
0.08
0.09
0.08
0.03
0.28
0.03
0.51
0.73
0.95
0.88
-
-
-
-
23
flsdNvOwMftfTfTCfTCftfTCldN03210 m−vO Preprint of http://doi.org/10.1098/rspa.2019.0339
Table 3: Fitted parameters for the constitutive laws for the low
PMMA (Eq. (21) and Eq.
(23)) and the high
PMMA obtained from Van Loock and Fleck [19] , see Eq. (21) and Eq.
low
high
PMMA
PMMA
2.5
1.8
3.2
17.3
-
-
-
-
-
-
-
65.8
0.8
1.58
0.173
(22).
Table 4: Summary of the assumed processing parameters and material properties for the void
growth predictions.
(MPa)
(MPa)
f0
(nm)
low
high
PMMA
PMMA
31
0.1
20
1190
114.5
10-3
10.5
31
0.1
20
1190
116.5
10-3
5
24
wMwMwMwM3 (nm)v− (J)q197.3110−197.3110−-10 (s)561.510561.5100 (Pa s)62.8101C2 (K)C0R (MPa)ER1R (s)− nwMwM0pap (s)p3 (kg m)−°g) (CT0a Preprint of http://doi.org/10.1098/rspa.2019.0339
List of Figure captions
Figure 1: SEM micrographs of the low
nanofoams at (a)
, (b)
of the high
nanofoams at (c)
and (d)
.
and
Figure 2: Nanofoaming experiments on the low
and high
PMMA grades: (a)
measured average cell size
versus foaming time
for
= 60 °C, (b) measured porosity f
versus foaming time
for
= 60 °C and
=100 °C, (c) measured porosity f versus foaming
temperature
for the range of explored foaming times ( = 60 s to = 600 s) for the low
nanofoams, and (d) measured f versus
for the range of explored foaming times ( = 60 s to
= 600 s) for the high
nanofoams.
Figure 3: Measured open cell content
as a function of porosity f for (a) the low
PMMA
nanofoam and (b) high
PMMA nanofoam.
Figure 4: Spherical void in (a) undeformed configuration with initial radius
and initial outer
radius
and (b) deformed configuration at time
of the void with radius
, outer radius
and gas pressure p.
Figure 5: The normalized glass transition temperature
of PMMA as a function of CO2
mass concentration
, as reported by Chiou et al. [30], Wissinger and Paulaitis [31], and Guo
and Kumar [26]. The
versus
curve is given by the calibrated version of Eq. (20).
Figure 6: Deformation mechanism maps for (a) low
PMMA and (b) high
PMMA
(for a reference strain
), for contours of effective strain rate
. The predicted
effective stress at the surface of the cavity
is plotted as a function of
for foaming
temperatures = 25 °C and = 80 °C and for a foaming time up to 600 s.
Figure 7: Predicted and measured porosity versus foaming time
, for = 25 °C to = 80
°C for (a) low
nanofoams and (b) high
PMMA nanofoams. The ductility-governed
porosity limit
is plotted via Eq. (32) for an initial porosity
. The minimum cell
wall thickness-governed porosity limit
is plotted via Eq. (34) for
and
= 3
(low
PMMA) and
= 4.2 (high
PMMA).
25
wMf C60T=f10C0 T=wMf C60T=f10C0 T=wMwMlftfTftfTfTfTftftwMfTftftwMvOwMwM0a0btab0gg/TTC0gg/TTCwMwMref0.05=ee/gTTfTfTfftfTfTwMwMff3010f−=cf3010f−=c0/hawMc0/hawM Preprint of http://doi.org/10.1098/rspa.2019.0339
Figures
Figure 1: SEM micrographs of the low
nanofoams at (a)
, (b)
and of the
high
nanofoams at (c)
and (d)
.
26
wMf C60T=f10C0 T=wMf C60T=f10C0 T= Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure 2: Nanofoaming experiments on the low
and high
PMMA grades: (a) measured average cell
size versus foaming time
for
= 60 °C, (b) measured porosity f versus foaming time
for
= 60 °C
and
=100 °C, (c) measured porosity f versus foaming temperature
for the range of explored foaming
times ( = 60 s to = 600 s) for the low
nanofoams, and (d) measured f versus
for the range of
explored foaming times ( = 60 s to = 600 s) for the high
nanofoams.
27
wMwMlftfTftfTfTfTftftwMfTftftwM Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure 3: Measured open cell content
as a function of porosity f for (a) the low
PMMA
nanofoam and (b) high
PMMA nanofoam.
28
vOwMwM Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure 3: Spherical void in (a) undeformed configuration with initial radius
and initial outer radius
and (b) deformed configuration at time of the void with radius
, outer radius
and gas pressure
p.
29
0a0btab Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure 5: The normalized glass transition temperature
of PMMA as a function of CO2
mass concentration
, as reported by Chiou et al. [32], Wissinger and Paulaitis [33], and Guo
and Kumar [28]. The
versus
curve is given by the calibrated version of Eq. (20).
30
0gg/TTC0gg/TTC Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure 6: Deformation mechanism maps for (a) low
PMMA and (b) high
PMMA
(for a reference strain
), for contours of effective strain rate
. The predicted
effective stress at the surface of the cavity
is plotted as a function of
for foaming
temperatures = 25 °C and = 80 °C and for a foaming time up to 600 s.
31
wMwMref0.05=ee/gTTfTfT Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure 4: Predicted and measured porosity versus foaming time
, for = 25 °C to = 80 °C for (a) low
nanofoams and (b) high
PMMA nanofoams. The ductility-governed porosity limit
is plotted
via Eq. (32) for an initial porosity
. The minimum cell wall thickness-governed porosity limit
is plotted via Eq. (34) for
and
= 3 (low
PMMA) and
= 4.2 (high
PMMA).
32
fftfTfTwMwMff3010f−=cf3010f−=c0/hawMc0/hawM Preprint of http://doi.org/10.1098/rspa.2019.0339
Appendix
Calibration of constitutive laws for the low
and high
PMMA
Constitutive laws are calibrated for the low
PMMA grade21 close to its glass transition
temperature. We follow the procedure of Van Loock and Fleck [19] who constructed
deformation and failure maps for the high
PMMA grade22 in uniaxial tension close to the
glass transition temperature. A series of uniaxial tensile tests were performed on the low
PMMA grade for a range of temperatures (T = 90 C to T = 170 C) and at a nominal strain rate
. The dogbone specimen geometry and the measurement procedures are
detailed in Van Loock and Fleck [19]. Note that the low
PMMA dogbone specimens are
machined from the foaming precursor sheets. The true stress versus true strain responses of the
low
PMMA dogbone specimens are plotted in Figure A1.a for
and in
Figure A.1b
. The true stress versus true strain curves of the high
PMMA grade are included in Figures A.1a and A.1b.
Loading-unloading uniaxial stress stress versus strain curves for the low
PMMA and high
low
PMM are shown in Figure A.2. At
= 0.93, the elastic unloading of the low
and the high
PMMA occurs in the manner of an elasto-viscoplastic solid, with
remnant a finite strain at zero load. The qualitative stress versus strain response of the low
and the high
PMMA is different when the temperature is increased to
= 1.06. The
elastic rubbery regime is entered for the high
PMMA and the unloading curve is almost
coincidental with the loading curve; there is negligible hysteresis and negligible remnant strain.
No rubbery regime is observed for the low
PMMA above the glass transition. At
=
1.06 and
= 1.12, the stress versus strain response of the low
PMMA in uniaxial
21 Altuglas V825T with
and
= 92 500 g mol-1.
22 Altuglas CN with
and
= 3 580 000 g mol-1.
33
wMwMwMwMwM2-15.91s0 e−=wMwMg0.94/1.01TTg1.04/1.14TTwMwMwMg/TTwMwMwMwMg/TTwMwMg/TTg/TTwMg114.5 CT=wMg116.5 CT=wM Preprint of http://doi.org/10.1098/rspa.2019.0339
tension is linear viscous. Unloading is accompanied by a finite remnant strain. The high
PMMA transitions from the rubbery regime to a viscous regime at
= 1.16.
First, consider the elasto-viscoplastic regime. The dependence of the measured flow strength
of the low
and high
PMMA grades upon
is shown in Figure A.3 for
. A single transition Ree-Eyring equation, Eq. (21), is fitted to the
versus
T/Tg response of the low
PMMA in the glassy and glass transition regime (corresponding
to 0.94
T/Tg
1.04). We assume that q equals
and
equals
for both the low
and the high
PMMA, as reported by Van Loock and Fleck [19]. The
activation volume v equals
for the low
PMMA, and
for the high
PMMA [19]. The resulting curve fits are included in Figure A.3. Second, consider the
viscous regime for the low
PMMA. We fit a linear, viscous constitutive law, Eqs. (23) and
(24), to the measured
versus
curves of the low
PMMA in the regime of 1.06
T/Tg
1.14 and
. The fitting values are
,
, and
. The resulting curve fit is adequate, see Figure A.3. Third, consider the rubbery
regime of the high
PMMA. The constitutive description, Eq. (22), is adequate upon
making use of previously measured values (
,
,
, and
[19]), as shown in Figure A.3.
Tensile ductility of the low
and high
PMMA
Van Loock and Fleck [19] measured the true tensile failure strain, that is ductility,
of the
high
PMMA grade by testing a dogbone geometry at
< 1 and an hourglass-shaped
specimen geometry at
≥ 1. The measured values for
of the high
PMMA grade
are plotted as a function of the normalized temperature
for a nominal strain rate
in Figure A.4. The
versus
failure envelope is adequately fitted by a
linear relation [19]:
34
wMg/TTywMwMg/TT-2-15.910 se=yσwM197.3110 J−056-11.510 s=wMwM-32.5 nmwM-31.8 nmv=wMwMyg/TTwM-2-15.910 se=602.810 Pas=13.2C=217.3 KC=wM0R65.8 MPaE=R= 0.80-1R= 1.58 s0.173n=wMwMfwMg/TTg/TTfwMg/TT215.91s0e−−=fg/TT Preprint of http://doi.org/10.1098/rspa.2019.0339
(A.1)
An additional series of uniaxial tensile tests have been conducted on the low
PMMA grade
by using the same measurement methods as that detailed in the work of Van Loock and Fleck
[19]. No failure was observed at T
145 C prior to the attainment of the maximum cross-head
extension. The measured
versus
curve is shown in Figure A.4. The failure envelope
of the low
PMMA grade close to the glass transition is also fitted by a linear relation:
(A.2)
35
fg37.36.TT−=wMfg/TTwMfg13.131.7TT−= Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure A.1: Measured true tensile stress
versus true tensile strain
curves for the low
and high
PMMA grades in uniaxial tension for a nominal strain rate
and for temperatures
ranging from (a) T = 90 C to T = 120 C and (b) T = 130 C to T = 170 C. A cross at the end of the
curve denotes specimen failure.
36
wMwM215.91s0−−= Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure A.2: Loading-unloading true stress versus true strain curves for the low
PMMA and
high
PMMA grades in uniaxial tension, at selected values of
, for a nominal strain
rate
.
37
wMwMg/TT415.91s0−−= Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure A.3: Deformation mechanism maps of the low
and high
PMMA grades. Flow
strength
(=
) versus
is plotted, with the curve fits of the constitutive models
included for a reference strain
.
38
wMwMyeg/TTref0.05= Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure A.4: The measured true tensile failure strain
as a function of normalized testing temperature
for the low
and high
PMMA grades, at a nominal strain rate
.
39
fg/TTwMwM21e5.90s1−−= Preprint of http://doi.org/10.1098/rspa.2019.0339
References
1.
2.
Costeux S. 2014 CO2-blown nanocellular foams. J. Appl. Polym. Sci. 131.
Notario B, Pinto J, Rodríguez-Pérez MÁ. 2015 Towards a new generation of
polymeric foams: PMMA nanocellular foams with enhanced physical properties.
Polymer (Guildf). 63, 116 -- 126. (doi:10.1016/j.polymer.2015.03.003)
3. Miller D, Kumar V. 2009 Microcellular and nanocellular solid-state polyetherimide
(PEI) foams using sub-critical carbon dioxide I. Processing and structure. Polymer
(Guildf). 50, 5576 -- 5584. (doi:10.1016/j.polymer.2011.04.049)
4. Martín-de León J, Bernardo V, Rodríguez-Pérez MÁ. 2017 Key production parameters
to obtain transparent nanocellular PMMA. Macromol. Mater. Eng. 302, 3 -- 7.
(doi:10.1002/mame.201700343)
5.
6.
Schmidt D, Raman VI, Egger C, du Fresne C, Schadler V. 2007 Templated cross-
linking reactions for designing nanoporous materials. Mater. Sci. Eng. C 27, 1487 --
1490. (doi:10.1016/j.msec.2006.06.028)
Notario B, Pinto J, Solorzano E, de Saja JA, Dumon M, Rodríguez-Pérez MÁ. 2014
Experimental validation of the Knudsen effect in nanocellular polymeric foams.
Polymer (Guildf). 56, 57 -- 67.
7. Wang G, Wang C, Zhao J, Wang G, Park CB, Zhao G. 2017 Modelling of thermal
transport through a nanocellular polymer foam: toward the generation of a new
superinsulating material. Nanoscale 9, 5996 -- 6009.
8. Martini JE. 1981 The production and analysis of microcellular foam (PhD thesis). MIT
(USA).
9. Martin-de León J, Bernardo V, Rodríguez-Pérez MÁ. 2016 Low density nanocellular
polymers based on PMMA produced by gas dissolution foaming : fabrication and
cellular structure characterization. Polymers (Basel). 8, 256.
10. Aher B, Olson NM, Kumar V. 2013 Production of bulk solid-state PEI nanofoams
using supercritical CO2. J. Mater. Res. 28, 2366 -- 2373. (doi:10.1557/jmr.2013.108)
11. Costeux S, Zhu L. 2013 Low density thermoplastic nanofoams nucleated by
nanoparticles. Polymer (Guildf). 54, 2785 -- 2795. (doi:10.1016/j.polymer.2013.03.052)
12. Bernardo V, Martin-de Leon J, Pinto J, Catelani T, Athanassiou A, Rodriguez-Perez
MA. 2019 Low-density PMMA/MAM nanocellular polymers using low MAM
contents: Production and characterization. Polymer (Guildf). 163, 115 -- 124.
(doi:10.1016/j.polymer.2018.12.057)
13. Costeux S, Khan I, Bunker SP, Jeon HK. 2014 Experimental study and modeling of
nanofoams formation from single phase acrylic copolymers. J. Cell. Plast. 51, 197 --
221. (doi:10.1177/0021955X14531972)
14.
Street JR. 1968 The Rheology of Phase Growth in Elastic Liquids. J. Rheol. (N. Y. N.
Y). 12, 103. (doi:10.1122/1.549101)
15. Amon M, Denson CD. 1984 A study of the dynamics of foam growth: Analysis of the
40
Preprint of http://doi.org/10.1098/rspa.2019.0339
growth of closely spaced spherical bubbles. Polym. Eng. Sci. 24, 1026 -- 1034.
(doi:10.1002/pen.760241306)
16. Venerus DC, Yala N, Bernstein B. 1998 Analysis of diffusion-induced bubble growth
in viscoelastic liquids. J. Nonnewton. Fluid Mech. 75, 55 -- 75. (doi:10.1016/S0377-
0257(97)00076-1)
17. Khan I, Adrian D, Costeux S. 2015 A model to predict the cell density and cell size
distribution in nano-cellular foams. Chem. Eng. Sci. 138, 634 -- 645.
(doi:10.1016/j.ces.2015.08.029)
18.
Shafi MA, Lee JG, Flumerfelt RW. 1996 Prediction of cellular structure in free
expansion polymer foam processing. Polym. Eng. Sci. 36, 1950 -- 1959.
(doi:10.1002/pen.10591)
19. Van Loock F, Fleck NA. 2018 Deformation and failure maps for PMMA in uniaxial
tension. Polymer (Guildf). 148, 259 -- 268. (doi:10.1016/j.polymer.2018.06.027)
20.
Pinto J, Solórzano E, Rodriguez-Perez MA, De Saja JA. 2013 Characterization of the
cellular structure based on user-interactive image analysis procedures. J. Cell. Plast.
49, 555 -- 575. (doi:10.1177/0021955X13503847)
21. Kumar V, Suh NP. 1990 A process for making microcellular thermoplastic parts.
Polym. Eng. Sci. 30, 1323 -- 1329. (doi:10.1002/pen.760302010)
22. Rides M, Morikawa J, Halldahl L, Hay B, Lobo H, Dawson A, Allen C. 2009
Intercomparison of thermal conductivity and thermal diffusivity methods for plastics.
Polym. Test. 28, 480 -- 489. (doi:10.1016/j.polymertesting.2009.03.002)
23. ASTM Int. 2015 Standard test method for open cell content of rigid cellular plastics.
24. Goel SK, Beckman EJ. 1995 Nucleation and Growth in Microcellular materials:
Supercritical CO2 as foaming agent. AIChe 41, 357 -- 367. (doi:10.1002/aic.690410217)
25. Timoshenko SP, Goodier JN. 2001 Theory of Elasticity. McGraw-Hill.
26. Hill R. 1950 The Mathematical Theory of Plasticity. Oxford University Press Ic., New
York.
27. Crank J. 1979 The Mathematics of Diffusion. Clarendon Press.
28. Guo H, Kumar V. 2015 Solid-state poly(methyl methacrylate) (PMMA) nanofoams.
Part I: Low-temperature CO2 sorption, diffusion, and the depression in PMMA glass
transition. Polymer (Guildf). 57, 157 -- 163.
29. Li R, Ye N, Shaayegan V, Fang T. 2018 Experimental measurement of CO2 diffusion
in PMMA and its effect on microcellular foaming. J. Supercrit. Fluids 135, 180 -- 187.
(doi:10.1016/j.supflu.2018.01.024)
30. Alessi P, Cortesi A, Kikic I, Vecchione F. 2003 Plasticization of polymers with
supercritical carbon dioxide: Experimental determination of glass-transition
temperatures. J. Appl. Polym. Sci. 88, 2189 -- 2193. (doi:10.1002/app.11881)
31. Verreck G, Decorte A, Li H, Tomasko D, Arien A, Peeters J, Rombaut P, Van den
41
Preprint of http://doi.org/10.1098/rspa.2019.0339
Mooter G, Brewster ME. 2006 The effect of pressurized carbon dioxide as a plasticizer
and foaming agent on the hot melt extrusion process and extrudate properties of
pharmaceutical polymers. J. Supercrit. Fluids 38, 383 -- 391.
(doi:10.1016/j.supflu.2005.11.022)
32. Chiou JS, Barlow JW, Paul DR. 1985 Plasticization of glassy polymers by CO2. J.
Appl. Polym. Sci. 30, 2633 -- 2642. (doi:10.1002/app.1985.070300626)
33. Wissinger RG, Paulaitis ME. 1991 Glass transitions in polymer/CO2 mixtures at
elevated pressures. J. Polym. Sci. Part B Polym. Phys. 29, 631 -- 633.
(doi:10.1002/polb.1991.090290513)
34. Chow S. 1980 Molecular interpretation of the glass transition temperature of polymer-
diluent systems. Macromolecules 364, 362 -- 364.
35. Li R, Li L, Zeng D, Liu Q, Fang T. 2016 Numerical Selection of the Parameters in
Producing Microcellular Polymethyl Methacrylate With Supercritical CO 2. 35, 309 --
328.
36. McLoughlin JR, Tobolsky A V. 1952 The viscoelastic behavior of polymethyl
methacrylate. J. Colloid Sci. 7, 555 -- 568.
37. Williams ML, Landel RF, Ferry JD. 1955 The temperature dependence of relaxation
mechanisms in amorphous polymers and other glass-forming liquids. J. Am. Chem.
Soc. 77, 3701 -- 3707.
38. Bin Ahmad Z, Ashby MF. 1988 Failure-mechanism maps for engineering polymers. J.
Mater. Sci. 23, 2037 -- 2050.
39. Rajendran A, Bonavoglia B, Forrer N, Storti G, Mazzotti M, Morbidelli M. 2005
Simultaneous measurement of swelling and sorption in a supercritical CO2-
poly(methyl methacrylate) system. Ind. Eng. Chem. Res. 44, 2549 -- 2560.
(doi:10.1021/ie049523w)
40.
41.
Pantoula M, Panayiotou C. 2006 Sorption and swelling in glassy polymer/carbon
dioxide systems. J. Supercrit. Fluids 37, 254 -- 262. (doi:10.1016/j.supflu.2005.11.001)
Pantoula M, von Schnitzler J, Eggers R, Panayiotou C. 2007 Sorption and swelling in
glassy polymer/carbon dioxide systems. Part II-Swelling. J. Supercrit. Fluids 39, 426 --
434. (doi:10.1016/j.supflu.2006.03.010)
42. Van Krevelen DW, Te Nijenhuis K. 2009 Properties of Polymers. (doi:10.1016/B978-
0-08-054819-7.X0001-5)
43. Rubinstein M, Colby HR. 2003 Polymer Physics. Oxford University Press.
44. Liu Y, Chen YC, Hutchens S, Lawrence J, Emrick T, Crosby AJ. 2015 Directly
measuring the complete stress-strain response of ultrathin polymer films.
Macromolecules 48, 6534 -- 6540. (doi:10.1021/acs.macromol.5b01473)
45. Bay RK, Shimomura S, Liu Y, Ilton M, Crosby AJ. 2018 Confinement effect on strain
localizations in glassy polymer films. Macromolecules 51, 3647 -- 3653.
(doi:10.1021/acs.macromol.8b00385)
42
Preprint of http://doi.org/10.1098/rspa.2019.0339
46. Cheng WM, Miller GA, Manson JA, Hertzberg RW, Sperling LH. 1990 Mechanical
behaviour of poly(methyl methacrylate). J. Mater. Sci. 25, 1931 -- 1938.
Electronic supplied material for Figure 1 in publication version
1
Martín-de León J, Bernardo V, Rodríguez-Pérez MÁ. 2017 Key production parameters to
obtain transparent nanocellular PMMA. Macromol. Mater. Eng. 302, 3 -- 7.
(doi:10.1002/mame.201700343)
2,11
Guo H, Nicolae A, Kumar V. 2015 Solid-state poly(methyl methacrylate) (PMMA)
nanofoams. Part II: Low-temperature solid-state process space using CO2 and the resulting
morphologies. Polymer (Guildf). 70, 231 -- 241. (doi:10.1016/j.polymer.2015.06.031)
3
Martín-de León J, Bernardo V, Rodríguez-Pérez MÁ. 2019 Nanocellular polymers: the
challenge of creating cells in the nanoscale. Materials (Basel). 12, 797.
(doi:10.3390/ma12050797)
4,7,8
Bernardo V, Martin-de Leon J, Pinto J, Catelani T, Athanassiou A, Rodriguez-Perez MA.
2019 Low-density PMMA/MAM nanocellular polymers using low MAM contents:
Production and characterization. Polymer (Guildf). 163, 115 -- 124.
(doi:10.1016/j.polymer.2018.12.057)
5,6
Martin-de León J, Bernardo V, Rodríguez-Pérez MÁ. 2016 Low density nanocellular
polymers based on PMMA produced by gas dissolution foaming : fabrication and cellular
structure characterization. Polymers (Basel). 8, 256.
9
43
Preprint of http://doi.org/10.1098/rspa.2019.0339
Bernardo V, Van Loock F, Martin-de Leon J, Fleck NA, Rodriguez-Perez MA. 2019
Mechanical Properties of PMMA-Sepiolite Nanocellular Materials with a Bimodal Cellular
Structure. Macromol. Mater. Eng. (doi:10.1002/mame.201900041)
10
Wang G, Zhao J, Mark LH, Wang G, Yu K, Wang C, Park CB, Zhao G. 2017 Ultra-tough
and super thermal-insulation nanocellular PMMA/TPU. Chem. Eng. J. 325, 632 -- 646.
(doi:10.1016/j.cej.2017.05.116)
12
Costeux S, Zhu L. 2013 Low density thermoplastic nanofoams nucleated by nanoparticles.
Polymer (Guildf). 54, 2785 -- 2795. (doi:10.1016/j.polymer.2013.03.052)
13
Costeux S, Khan I, Bunker SP, Jeon HK. 2014 Experimental study and modeling of
nanofoams formation from single phase acrylic copolymers. J. Cell. Plast. 51, 197 -- 221.
(doi:10.1177/0021955X14531972)
14
Reglero Ruiz JA, Dumon M, Pinto J, Rodriguez-Pérez MA. 2011 Low-density nanocellular
foams produced by high-pressure carbon dioxide. Macromol. Mater. Eng. 296, 752 -- 759.
(doi: https://doi.org/10.1002/mame.201000346)
44
Preprint of http://doi.org/10.1098/rspa.2019.0339
Figure 1 in publication version: Reported porosity f versus void size l data of high porosity
(P)MMA-based nanofoams produced via solid-state foaming. The '○' markers refer to results
obtained in this work. The'●' markers refer to data retrieved from references in the list above.
45
|
1807.08730 | 2 | 1807 | 2018-10-25T16:21:48 | Unidirectional zero sonic reflection in passive $\mathcal{PT}$ symmetric Willis media | [
"physics.app-ph"
] | In an effective medium description of acoustic metamaterials, the Willis coupling plays the same role as the bianisotropy in electromagnetism. Willis media can be described by a constitutive matrix composed of the classical effective bulk modulus and density and additional cross-coupling terms defining the acoustic bianisotropy. Based on an unifying theoretical model, we unite the properties of acoustic Willis coupling with $\mathcal{PT}$ symmetric systems under the same umbrella and show in either case that an exceptional point hosts a remarkably pronounced scattering asymmetry that is accompanied by one-way zero reflection for sound waves. The analytical treatment is backed up by experimental input in asymmetrically side-loaded wavesguides showing how gauge transformations and loss biasing can embrace both Willis materials and non-Hermitian physics to tailor unidirectional reflectionless acoustics, which is appealing for purposeful sound insulation and steering. | physics.app-ph | physics | Unidirectional zero sonic reflection in passive PT symmetric Willis media
Aur´elien Merkel,1 Vicent Romero-Garc´ıa,2 Jean-Philippe Groby,2 Jensen Li,3 and Johan Christensen1
1Department of Physics, Universidad Carlos III de Madrid,
Avenidad de la Universidad 30, 28911 Legan´es (Madrid), Spain
2Laboratoire de l'Universit´e du Mans, LAUM UMR-6613 CNRS,
Le Mans Universit´e, Avenue Olivier Messiaen, 72085 Le Mans cedex 9, France
3Department of Physics, The Hong Kong University of Science and Technology, Hong Kong, China
In an effective medium description of acoustic metamaterials, the Willis coupling plays the same
role as the bianisotropy in electromagnetism. Willis media can be described by a constitutive matrix
composed of the classical effective bulk modulus and density and additional cross-coupling terms
defining the acoustic bianisotropy. Based on an unifying theoretical model, we unite the properties
of acoustic Willis coupling with PT symmetric systems under the same umbrella and show in
either case that an exceptional point hosts a remarkably pronounced scattering asymmetry that is
accompanied by one-way zero reflection for sound waves. The analytical treatment is backed up by
experimental input in asymmetrically side-loaded wavesguides showing how gauge transformations
and loss biasing can embrace both Willis materials and non-Hermitian physics to tailor unidirectional
reflectionless acoustics, which is appealing for purposeful sound insulation and steering.
Metamaterials and metasurfaces have emerged as
artificial structures that enable one to tailor the wave
propagation at will [1]. Whether the field of interest
is optics, acoustics, or elasticity, recent efforts have
enabled exiting wave phenomena ranging from cloaks
of invisibility and unhearability, zero-index behaviour
and negative refraction, just to name a few of many
counterintuitive effects [2 -- 6]. Wave-based diodes and
rectifiers enabling one-way sound or light propagation
have recently also been put on the map of contemporary
metamaterial research [7 -- 12].
Waves irradiating onto a passive reciprocal medium
usually display symmetric transparency in that the
transmittance does not depend on the side at which
waves are launched. In contrast, if the elementary build-
ing units of the medium involved lack intrinsic inversion
symmetry, then the reflectance depends on from which
side of the slab waves are irradiated. Commonly, for
electromagnetic (EM) waves, such asymmetric response
are known to occur in bianisotropic media, which are
in applications comprising unidirectional
important
single-sided light detection and emission
radiation,
[13 -- 17].
The intrinsic asymmetry of the involved
meta-atoms leads to subwavelength cross-coupling where
magnetic states can be excited also by electric fields,
and likewise, electric states that can be excited through
magnetoelectric coupling. An analogue picture for the
case of sound is provided through the cross-coupling
between strain and velocity in so-called Willis media
[18, 19]. Local and non-local coupling in the form of
microstructural asymmetry and finite phase changes
across inhomogeneities, respectively, define the very
nature of Willis coupling in bianisotropic artificial
acoustic or mechanical media [20 -- 25]. To this extent,
Willis media have not only emerged to obtain physically
meaningful effective parameters, but also for the design
of passive acoustic one-way reflectionless systems with
the potential to engineer asymmetric steering of sound
and to create mechanical structures that suppress echoes
from one side, but act reflective from the other.
Asymmetries related to wave propagation have been
made also possible in systems disobeying parity P and
time-reversal T symmetry, but that are nevertheless
symmetric under simultaneous operation. PT symmetry
or non-Hermiticity in optics or acoustics, embraces
the introduction of loss and gain constituents in new
synthetic material such as metamaterials.
In optics,
gain is an essential ingredient to overcome metamaterial
losses but has found increasing importance in one-way
waveguides, single-sided scattering, and enhanced sens-
ing at the exceptional point (EP), which marks the
transition between an exact and a broken PT -symmetry
phase [26]. Specifically, among the phenomena occurring
at the EP, unidirectional reflectionless wave propagation,
which is also called anisotropic transmission resonance,
originates from a non-Hermitian degeneracy of the
scattering matrix [27 -- 31].
Interestingly, unidirectional
zero reflection has also been found in entirely passive
non-Hermitian systems without involving gain,
thus
posing significant ease for experimental realizations
[32 -- 40].
In this letter we present a theoretical unification
framework that merges the properties of the acoustic
constitutive relations containing local and nonlocal
Willis coupling with a conventional PT symmetric
system representation at the EP in asymmetrically
side-loaded waveguides.
In details, based on experi-
mental data, we map the one-way reflectionless sound
propagation in Willis media onto an ideal, that is, a
gain and loss balanced PT symmetric system through
a gauge transformation comprising a shift in the P
8
1
0
2
t
c
O
5
2
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
2
v
0
3
7
8
0
.
7
0
8
1
:
v
i
X
r
a
2
operation associated with an average loss bias. This
unconventional strategy to tailor the asymmetric flow of
sound further extends the family of non-Hermitian wave
physics to synthesized PT matter.
We begin by considering an effective medium contain-
ing Willis coupling, hence, the conservation of momen-
tum in an acoustic waveguide is
∇P = − ρeff
ρ0c0S
∂U
∂t
− χ1
∂P
∂t
,
(1)
and the conservation of mass reads
∂P
∂t
− χ2 · ∂U
∂t
∇ · U = − ρ0c0S
Keff
,
(2)
where P ≡ P/(ρ0c0) is the acoustic pressure, ρ0 is the
mass density and c0 is the speed of sound in the back-
ground medium, U ≡ SU is the acoustic volume flow and
S is the cross-section of the cylindrical waveguide, χ1,2
are the Willis coupling parameters, and ρeff, Keff repre-
sent the effective mass density and bulk modulus, respec-
tively. Because of reciprocity, the relation −χ1 = χ2 ≡ χ
is always satisfied. The constitutive matrix M in one di-
mension along the z axis with the time convention e−iωt,
has the following form
(cid:19)
(cid:18) P
U
∂
∂z
Π
= iω
with the operator
(cid:20) Sρ0c0/Keff
(cid:18) 0 1
−χ
Π =
1 0
(cid:21)(cid:18) P
(cid:19)
U
,(3)
(4)
χ
ρeff/(Sρ0c0)
(cid:19)
.
We can find the elements of the constitutive matrix
from the scattering coefficients, which can be derived
from the transfer matrix method [41, 42]. The acous-
tic pressure and velocity located before and after the cell
of length L can be related via the constitutive matrix M
as follows
= eiωΠML
(cid:21)
Uz=L
(cid:20) Pz=L
(cid:20) 1/(ρ0c0)
(cid:20) 1/(ρ0c0)
×
1/(ρ0c0)
1/(SZw) −1/(SZw)
1/(ρ0c0)
1/(SZw) −1/(SZw)
,
(cid:21)
Uz=0
(cid:20) Pz=0
(cid:21)(cid:18) t − rLrR/t rR/t
(cid:21)−1
−rL/t
1/t
(5)
(cid:19)
,
(6)
leading to
eiωΠML =
where t is the complex and reciprocal transmission coef-
ficient, rL,R are the complex reflection coefficients from
left L and right R incidence, respectively, and Zw is the
acoustic impedance of the waveguide. In Eq. (3) we can
refrain from normalizing the acoustic parameters to ob-
tain an effective constitutive matrix
Meff =
,
(7)
(cid:18) K−1
eff
χ
−χ ρeff
(cid:19)
FIG. 1. (color online) (a) Picture of the experimental setup
with the acoustic waveguide side-loaded by two quarter-
wavelength resonators of different resonance frequencies. (b)
Experimentally derived effective parameters Keff, ρeff and χ.
(c) Theoretical (continuous curves) and experimental (dots)
eigenvalues ζ1,2 of the constitutive matrix Meff. The EP
where the eigenvalues coalesce is marked with an orange ver-
tical line.
which has the eigenvalues ζ1,2. In order to realize such
effective metamaterial system containing an asymmetric
subwavelength cell, we fabricated an acoustic waveguide
with two side-loaded resonators tuned at two distinct res-
onance frequencies as depicted in Fig. 1(a). We take
the case of two quarter-wavelength resonators of radius
R1,2 = 5.35 mm having their resonance frequencies at
f1 = 895 Hz and f2 = 862 Hz, corresponding to the
lengths L1 = 9.141 cm and L2 = 9.51 cm, respectively
and separated by a distance L = 3 cm in the waveguide of
radius Rw = 1.5 cm. The scattering coefficients are mea-
sured in an impedance tube with the four microphone
method. The effective parameters, which are derived
from the matrix logarithm of Eq. (6), exhibit nonzero
Willis coupling as can be seen in Fig. 1(b). Whereas pe-
culiar features do not appear obvious in ρeff and Keff,
the eigenvalues ζ1,2 coalesce at approximately 860 Hz as
can be seen in Fig. 1(c) demonstrating the formation of
an EP. At this point, we show that the constitutive ma-
trix Meff can be mapped onto a conventional 2×2 PT
symmetric system matrix of the form [34, 43 -- 45]
(cid:18) a + i∆γ
−iκ
(cid:18) a + ∆γ
(cid:19)
(cid:19)
κ
ΥP T =
iκ
a − i∆γ
,
(8)
where a, ∆γ and κ are real numbers. Similarly, Meff can
also be mapped onto an entirely real 2×2 matrix that has
the form
ΥR =
(9)
These two PT -symmetric two-level matrices have an EP
occurring when ∆γ = ±κ. At the EP of our system,
a − ∆γ
−κ
.
FIG. 2. (color online) Gauge transformation of Eq. (10) com-
prising the elements of the constitutive matrix Meff at the
EP. The gauge transformation maps the constitutive matrix
onto a PT -symmetric matrix ΥP T + iγ0I, where Re(χ) = 0,
marked with vertical gray dotted lines in the gray areas or
onto a real matrix ΥR + iγ(cid:48)
0I, where Im(χ) = 0, marked with
vertical gray dashed lines in the blue areas.
eff ) (cid:54)= −Im(ρeff) and χ is com-
Re(K−1
plex as can be seen in Fig. 1(b), hence, at first sight, no
form similarity seems to exist between Meff, ΥP T or ΥR.
However, the constitutive matrix Meff can be gauged
through an unitary transformation [46]
MG = UuMeffU−1
u ,
eff ) (cid:54)= Re(ρeff), Im(K−1
(10)
where Uu is the propagator matrix
(cid:18) cos(kδ)
(cid:19)
Uu =
i sin(kδ)
i sin(kδ) cos(kδ)
,
(11)
in which k is the wavenumber and δ is real. The matrix
Uu is unitary, i.e., U−1
u, ensuring the eigenvalues
u = U†
3
of any transformed matrix MG remain invariant. This
unitary gauge transformation maps ideal PT symmetric
systems such as ΥP T and ΥR of symmetric P operation,
that is, z → −z, which is a mirror operation at the center
of the cell onto shifted PT symmetric systems accommo-
dating the generalized P operation z → 2δ − z, which is
a mirror operation at the position δ from the center of
the cell [47]. We begin by gauging the constitutive ma-
trix Meff through the dimensionless factor kδ in order
to obtain fully imaginary Willis coupling, Re(χ) = 0, as
highlighted with vertical dotted lines in the gray areas
in Fig. 2, which is accompanied by two exact analyt-
ical constitutive relationships Re(K−1
eff ) = Re(ρeff) and
eff − ρef f ) = ±2Im(χ). The resulting gauged con-
Im(K−1
stitutive matrix then obtains ±Im(χ) in the off-diagonal
terms whereas the real parts of the diagonal terms be-
come Re(K−1
ef f ). Since ΥP T represents an ideal balance
of loss and gain, Meff cannot exactly assume the same
form. However, in order to accomplish this, we intro-
duce an additional gauge transformation that is biasing
the system with an average level of loss γ0 [32 -- 40]. In do-
ing this, the ideal PT -symmetric Hamiltonian is mapped
onto an effective but passive Hamiltonian that yields a
loss-biased constitutive matrix of the following form
(cid:18) a + i∆γ + iγ0
−iκ
= Uu(ΥP T + iγ0I)]U−1
u ,
Meff = Uu
iκ
a − i∆γ + iγ0
U−1
u ,
(12)
where I is the identity matrix and the condition for the
EP remains as earlier, ∆γ = ±κ. Apart from the above,
we can also choose to gauge the system towards fully real
Willis coupling such that Im(χ) = 0. In this case, one
can see in Fig. 2 marked with vertical gray dashed lines
in the blue areas that Im(K−1
0 and
eff − ρeff) = ±2Re(χ), which correspond to
that Re(K−1
the EP condition for the physical effective system when
transformed from ΥR
eff ) = Im(ρef f ) = −iγ(cid:48)
(cid:18) a + ∆γ + iγ(cid:48)
κ
a − ∆γ + iγ(cid:48)
0
U−1
u ,
Meff = Uu
−κ
0I)]U−1
= Uu(ΥR + iγ(cid:48)
u .
0
(13)
Thanks to the fact that the signature of PT symmetry
breaking continues to exist even when an average loss
bias is applied to ideal PT Hamiltonians, the EP em-
bedded in the passive acoustic system will not cease to
exist. Hence, our passive system that is described by
effective constitutive relations containing the Willis cou-
pling can be mapped onto two different forms of the ideal
PT Hamiltonians using an unitary gauge transformation,
which conclusively unifies the resulting asymmetry in the
acoustic reflections.
In what follows, we show how the EP embedded in the
Meff matrix translates into the scattering properties of
the sample. The reciprocal scattering matrix that relates
the amplitude of the outcoming waves to the amplitude
(cid:19)
(cid:19)
4
tance. For right incidence on the other hand, the wave
is mostly reflected with lower level of absorption, denot-
ing a high level of asymmetry in both the absorptions
and reflections. Interestingly, the PT conditions for the
one-dimensional scattering matrix using the generalized
parity operator without considering the loss bias assume
now the following form
Re−4ikδ = r∗
LrRe4ikδ = 1 − t2,
(15)
Lte4ikδ = rRt∗ + r∗
Rte−4ikδ = 0.
(16)
rLr∗
rLt∗ + r∗
In summary, we have demonstrated that an EP can
appear in the constitutive relations of a passive acoustic
Willis media, which can be mapped onto ideal PT
Hamiltonians through a gauge transformation and an
average loss bias. This EP translates into an unidirec-
tional reflectionless propagation, which is, of primary
importance for sound absorption because, if combined
with coherent perfect absorption, it results in an uni-
directional perfect absorber [48]. Further, our present
findings can lead to a deeper insight into the unidirec-
tional invisibility phenomena [27, 29], altogether showing
how the Willis coupling broadens the possibilities of
embracing both worlds of acoustic metamaterials and
PT symmetry physics at once to achieve unprecedented
control of sound and vibrations.
J.C. acknowledges the support from the European Re-
search Council (ERC) through the Starting Grant No.
714577 PHONOMETA and from the MINECO through
a Ram´on y Cajal grant (Grant No. RYC-2015-17156). J.-
P. Groby and V. Romero Garc´ıa gratefully acknowledge
the support from the ANR Project METAUDIBLE No.
ANR-13-BS09-0003 co-founded by ANR and FRAE and
the support from the RFI Le Mans Acoustique (R´egion
Pays de la Loire) PavNat project. This article is based
upon work from COST Action DENORMS CA15125,
supported by COST (European Cooperation in Science
and Technology).
[1] M. Wegener, Science 342, 939 (2013).
[2] G. W. Milton, M. Briane, and J. R. Willis, New J. Phys.
8, 248 (2006).
[3] A. N. Norris, Proc. R. Soc. A 464, 2411 (2008).
[4] D. Torrent and J. S´anchez-Dehesa, New J. Phys. 10,
063015 (2008).
[5] V. M. Garc´ıa-Chocano, J. Christensen, and J. S´anchez-
Dehesa, Phys. Rev. Lett. 112, 144301 (2014).
[6] M. R. Haberman and A. N. Norris, Acoust. Today 12,
31 (2016).
[7] B. Liang, X. S. Guo, J. Tu, D. Zhang, and J. C. Cheng,
Nat. Mater. 9, 989 (2010).
[8] N. Boechler, G. Theocharis, and C. Daraio, Nat. Mater.
10, 665 (2011).
FIG. 3. (color online) (a) Eigenvalues spectrum of the scat-
tering matrix. The EP that originates from the coalescence of
the eigenvalues of the scattering matrix, is marked with the
crossing of orange lines. (b) Theoretical (continuous curves)
and experimental (dots) scattering coefficients of the sample.
The EP where rL → 0 is marked with an orange vertical line.
of the incoming waves is here defined as
(cid:18) t
(cid:19)
S =
rR
rL t
.
(14)
√
rR,
rR,±√
Its eigenvalues are ξ1,2 = t ± (rLrR)1/2 and its
√
(±√
rL) or
eigenvectors can be written either as (
rL). The coalescence of the eigenstates of this
scattering matrix is characterized by either rL = 0 or
rR = 0 and rL (cid:54)= rR.
In Fig. 1(c) we demonstrated
that the eigenvalues ζ1,2 of the constitutive matrix coa-
lesce at 860 Hz and agrees perfectly well with the spec-
tral location of the EP of the associated scattering ma-
trix as it can be seen in Fig. 3(a). The coalescence of
the scattering eigenvalues at that point yields reflection-
less propagation for left incidence rL → 0 (cid:54)= rR as seen
in Fig. 3(b), where experimental data agree very well
to theoretical predictions in asymmetrically side-loaded
waveguides. We further emphasize that the size of the
cell is substantially smaller than the incident wavelength
λ, i.e., L < λ/13 and that the unidirectional reflection-
less behavior is associated with high levels of absorp-
tion within the sample and a consequential low transmit-
5
[9] R. Fleury, D. L. Sounas, C. F. Sieck, M. R. Haberman,
and A. Al`u, Science 343, 516 (2014).
[10] T. Devaux, V. Tournat, O. Richoux, and V. Pagneux,
[30] C. Shi, M. D. ans Yun Chen, L. Cheng, H. Ramenazi,
Y. Wang, and X. Zhang, Nat. Commun. 7, 11110 (2016).
[31] J. Christensen, M. Willatzen, V. R. Velasco, and M.-H.
Phys. Rev. Lett. 115, 234301 (2015).
Lu, Phys. Rev. Lett. 116, 207601 (2016).
[11] G. Trainiti and M. Ruzzene, New J. Phys. 18, 083047
(2016).
[12] A. Merkel, M. Willatzen, and J. Christensen, Phys. Rev.
Applied 9, 034033 (2018).
[13] D. K. Cheng and J.-A. Kong, Proc. IEEE 56, 248 (1968).
[14] J.-A. Kong, Proc. IEEE 60, 1036 (1972).
[15] C. E. Kriegler, M. S. Linden, and M. Wegener, IEEE J.
[32] A. Guo, J. Salamo, D. Duchesne, R. Morandotti,
M. Volatier-Ravat, V. Aimez, G. A. Siviloglou, and D. N.
Christodoulides, Phys. Rev. Lett. 103, 093902 (2009).
[33] C. E. Ruter, K. G. Makris, R. El-Ganainy, D. N.
Christodoulides, M. Segev, and D. Kip, Nat. Phys. 6,
192 (2010).
[34] M. Kang, F. Liu, and J. Li, Phys. Rev. A 87, 053824
Sel. Top. Quant 16, 367 (2010).
(2013).
[16] R. Marqu´es, F. Medina, and R. Rafii-El-Idrissi, Phys.
Rev. B 65, 144440 (2002).
[17] L. Peng, K. Wang, Y. Yang, Y. Chen, G. Wang,
[35] L. Feng, Y.-L. Xu, W. S. Fegadolli, M.-H. Lu, J. E. B.
Oliveira, V. R. Almeida, Y.-F. Chen, and A. Scherer,
Nat. Mater. 12, 108 (2013).
B. Zhang, and H. Chen, Adv. Sci. , 1700922 (2018).
[36] Y. Sun, W. Tan, H.-Q. Li, J. Li, and H. Chen, Phys.
[18] J. R. Willis, Wave Motion 3, 1 (1981).
[19] G. W. Milton and J. R. Willis, Proc. R. Soc. A 463, 855
(2007).
Rev. Lett. 112, 143903 (2014).
[37] M. Lawrence, N. Xu, X. Zhang, L. Cong, and J. H. W.
Z. S. Zhang, Phys. Rev. Lett. 113, 093901 (2014).
[20] D. Torrent, Y. Pennec, and B. Djafari-Rouhani, Phys.
[38] J.-H. Wu, M. Artoni, and G. C. L. Rocca, Phys. Rev.
Rev. B 92, 174110 (2015).
Lett. 113, 123004 (2014).
[21] S. Koo, C. Cho, J.-H. Jeong, and N. Park, Nat. Commun.
[39] Y. Yan and N. C. Giebink, Adv. Optical Mat. 2, 423
7, 13012 (2016).
(2014).
[22] M. B. Muhlestein, C. F. Sieck, A. Al`u, and M. R. Haber-
man, Proc. R. Soc. A 472, 20160604 (2016).
[23] M. B. Muhlestein, C. F. Sieck, P. S. Wilson, and M. R.
Haberman, Nat. Commun. 8, 15625 (2017).
[24] C. F. Sieck, A. Al`u, and M. R. Haberman, Phys. Rev.
B 96, 104303 (2017).
[25] L. Quan, Y. Ra'di, D. Sounas, and A. Al`u, Phys. Rev.
Lett. 120, 254301 (2018).
[26] R. El-Ganainy, K. G. Makris, M. Khajavikhan, Z. H.
Musslimani, S. Rotter, and D. N. Christodoulides, Nat.
Phys. 14, 11 (2018).
[27] Z. Lin, H. Ramenazi, T. Eichelkraut, T. Kottos, H. Cao,
and D. N. Christodoulides, Phys. Rev. Lett. 106, 213901
(2011).
[28] L. Ge, Y. D. Chong, and A. D. Stone, Phys. Rev. A 85,
023802 (2012).
[29] R. Fleury, D. Sounas, and A. Al`u, Nat. Commun. 6,
5905 (2015).
[40] T. Liu, X. Zhu, F. C. S. Liang, and J. Zhu, Phys. Rev.
Lett. 120, 124502 (2018).
[41] A. Merkel, G. Theocharis, O. Richoux, V. Romero-
Garc`ıa, and V. Pagneux, Appl. Phys. Lett. 107, 244102
(2015).
[42] N. Jim´enez, V. Romero-Grac´ıa, V. Pagneux, and J.-P.
Groby, Sci. Rep. 7, 13595 (2017).
[43] C. M. Bender, D. C. Brody, and H. F. Jones, Phys. Rev.
Lett. 89, 270401 (2002).
[44] A. Mostafazadeh, J. Phys. A: Math. Gen. 36, 7081
(2003).
[45] K. Ding, G. Ma, M. Xiao, Z. Q. Zhang, and C. T. Chan,
Phys. Rev. X 6, 021007 (2016).
[46] J. Gear, Y. Sun, S. Xiao, L. Zhang, R. Fitzgerald, S. Rot-
ter, H. Chen, , and J. Li, New J. Phys. 19, 123041 (2017).
[47] F. Cannata, J.-P. Dedonder, and A. Ventura, Annals of
Physics 322, 397 (2007).
[48] H. Ramenazi, Y. Wang, E. Yablonovitch, and X. Zhang,
IEEE J. Sel. Topics Quant. Mech. 22, 5000706 (2016).
|
1803.04707 | 1 | 1803 | 2018-03-13T10:00:45 | Particle cooling in Vaporizing Coolant | [
"physics.app-ph"
] | The approximate mathematical model for a cooling of the particle in a volatile liquid is developed and analyzed. Despite the precise model is complex and requires the solution of the nonstationary two-phase flow equations with the conjugated heat transfer boundary problem for the particle, vapor, and liquid cooling pool, the considered simple model may be of interest. Vapor is permanently produced and removed from the coolant pool. Analysis of the model obtained resulted in some correlations for the three main parameters of the cooling process, which may be used for estimation of the particle cooling. | physics.app-ph | physics | Particle cooling in Vaporizing Coolant
Ivan V. Kazachkov1,2
Dept of Information Technologies and Data Analysis, Nizhyn Gogol state university, Ukraine
Dept of Energy Technology, Royal Institute of Technology (KTH), Stockholm, Sweden)
[email protected]
Abstract: The approximate mathematical model for a cooling of the particle in a volatile liquid is
developed and analyzed. Despite the precise model is complex and requires the solution of the
nonstationary two-phase flow equations with the conjugated heat transfer boundary problem for the
particle, vapor, and liquid cooling pool, the considered simple model may be of interest. Vapor is
permanently produced and removed from the coolant's pool. Analysis of the model obtained resulted
in some correlations for the three main parameters of the cooling process, which may be used for
estimation of the particle's cooling.
Keywords: cooling, drops, particles, mathematical model, cooling rate, vapor
1. Introduction and statement of the problem
In many industrial and technical applications (granulation of metal and allows, development of
the amorphous materials, etc.) [1-18] the problem of cooling of the high-temperature drops and
particles is of great importance. For example, by the development of the passive protection systems
against severe accidents at the nuclear power plant, one of the key problems is cooling the corium
melt, which is speaded in a containment [19]. By cooling of the high-temperature particles and drops
in a volatile coolant, an intensive vaporization around the particle (crisis of heat transfer) is
dramatically decreasing the intensity of cooling. Vapor around cooling particle decrease heat transfer
by 2-3 decimal orders, which is a so-called crisis of heat transferThe theory and methods of parametric
excitation and suppression of oscillations on the film flow surfaces under an action of the
electromagnetic fields and vibrations.
Experimental facilities for rapid cooling of the drops, with up to ten thousand degrees per second
[3, 13-15], which allow producing amorphous nanostructured granules for the new materials'
production with unique properties are of serious interest for application and as such must be developed
further. Therefore, the presented paper is devoted to studying of the particle cooling in a vaporizing
coolant. The theoretical base and engineering applications, including the new methods and devices for
granulation of the liquid metals, have been developed at the Institute of Electrodynamics of the
Ukrainian National Academy of Sciences [1-15]. The principle of operation of the installations for
electromagnetic batching of metal alloys with a melting temperature up to 103 Celsius degrees and
slightly over, which intended for production of the metallic granules, is based on the jet disintegration
on the first mode of Eigen oscillations. Applying the crossed electric and magnetic fields at the nozzle
leads to a resonance regime with a frequency of the alternating ponderomotive electromagnetic force
equal to a frequency of the jet's Eigen oscillations). The magnetic field is created by a permanent
magnet. A current in a melt at the nozzle has the industrial frequency 50 Hz.
The installation includes such elements as an induction furnace, a liquid-metal pump for
delivering a melt through a channel to the electromagnetic batcher, and a crystallizer of the drops (for
obtaining the granules). Electromagnetic batcher implements the controlling capillary disintegration
(fragmentation) of a cylindrical jet of liquid metal in the range of 100...400 Hz (electromagnetic force
has 100 Hz frequency, it is created applying the electric current of 50 Hz). The diameter of metal
particles (granules) is determined by the resonant frequency of a jet at the first harmonics, which is
approximately ω=0.23u/d for the low viscous melts [2, 10-12]. The velocity of liquid metal flow u and
1
the nozzle diameter d determine the size of particles due to the disintegration of a jet and productivity
of the granulation machine. A diameter of the spherical particles, which are produced due to a jet's
decay, is roughly estimated as dp=1.88d. Because the capillary forces are growing with a decrease of
the nozzle's diameter (inversely proportional), the controlled disintegration of a jet is available up to
the nozzle's diameter of about 1 mm. Afterward the Coanda effect makes the jet chaotically directed
flow from the nozzle. It is breaking the jet's strictly vertical flow. In addition, even an accidental slight
clogging of the melt with some impurities can also destroy the jet's flow regime and cause the
granulator to stop. Therefore, the jet's type granulators have a restriction in a size of the resulting
particles about 2 mm in diameter.
Because the jet granulators, despite having big advantage in getting nearly ideal spherical granules
of the same given size, revealed serious impediment for the size limitation about 2 mm from below,
the theory of parametrically controlled disintegration of the liquid metal film flow has been developed
[1, 3, 13-15], and the methods and devices for film granulation were invented [5-9]. The last ones
produce granules in the range of 0.5-1.5 from the average size, but they practically do not have any
limitations on a size of the producing granules up to the micrometer size. Due to this, the high cooling
rate was obtained in a liquid nitrogen (up to 104 Celsius degrees per second). For this, except the small
size of the drops (granules after their solidification), the avoiding of heat transfer crisis was invented.
Drops and particles during their cooling were passing through a series of the liquid metal films to
permanently destroy the vapor film on the hot particle in a liquid coolant.
Examples of the processes for obtaining the liquid drops of given size with their further
solidification into granules are presented in Fig. 1 and Fig. 2 (the drops and particles were cooled in
the film granulation machine with high cooling rate up to 104 oC):
Fig. 1 Vibration controlled soliton-like regime (to the left) and electromagnetically controlled
the resonant disintegration of the film flow (to the right)
Fig. 2 Granules of the zink (left) and aluminium (right) cooled in a liquid nitrogen
2. Physical statement for the model of the particle cooling
Let us consider physical situation according to the pictures presented in Fig. 3. The hot particle is
surrounded by a vapor, which is intensively flowing around a particle. Vapor is permanently produced
due to intensive heat release from the hot particle. In a granulation machine, it is around thousand
Celsius degrees or so, while during severe accidents at the nuclear power plants (NPP) the drops and
particles of the corium may be of temperature even higher (2-3 thousand Celsius degrees) [19]. In case
of granulation the phenomenon of the crisis of heat transfer substantially worsens the coolability
conditions and the cooling rate. In case of cooling of the corium melt during postulated severe
accidents at NPP this phenomenon is crucial as far as permanently generating heat by the radioactive
nuclear fuel must be removed with the 100% guaranty to avoid the catastrophe.
The described important heat transfer problem is considered in this paper through development
and analysis of the mathematical model for particle cooling.
Fig. 3 Schematic of the hot particle in a volatile coolant with a vapor flow around the particle
3. A mathematical model for the cooling process of a particle in a volatile liquid
Mechanical equilibrium of a particle on a surface of liquid coolant
Development of the mathematical model for physical processes during the cooling of a drop or
particle is done starting from the mechanical equilibrium of particle on a surface of liquid coolant (e.g.
in case of the drop falling down into a pool of coolant):
3
where
, (1)
, (2)
here
is the radius of the equivalent circle of the capillary meniscus,
are the density and
radius of particle and acceleration due to gravity, correspondingly,
- density of liquid coolant,
-
the coefficient of a surface tension;
are the angles in the considered spherical coordinate system,
are the pressure of vapor and the shear stress due to a vapor flow around the particle,
respectively,
is the angle corresponding to a length h of a particle penetration into a coolant. The
parameter
in Fig. 3 is a width of a layer of vapor flow. The function
and a length h of
penetration into a coolant depend on particle and coolant density ratio, the temperature of the particle
and other physical properties, which are revealed later on. Those determine the intensity of the
coolant's vaporization, which, in turn, determines the action on a particle hydrodynamic force.
The equation (1) represents the condition of a particle levitation on the surface of liquid coolant.
The weight of the particle is in the equilibrium with two hydrodynamic forces acting on the particle:
pressure and shear force. Therefore, the length h of particle's penetration into a coolant may be
different depending on the above-mentioned physical conditions.
The radius
of the equivalent circle of the capillary meniscus is connected to the width of a
vapor layer
as follows:
where
, а is the distance between the center of the particle and the equivalent circle
of the capillary meniscus, to be computed further. Here from yields
,
.
If
is small, then
, and by small values of
, which are even smaller ones, is got
and then
,
. (3)
In the expression (3) thus obtained the second term in the brackets is small comparing to 1, therefore
.
3042sin(cossin)3hpvwgRRpRd0112(cos)vlppgRRRR1R,,pRgl,,vwph()()1R2220221()2()cos(180)()2()cosRaRaRaRaR10RRa2221cossinRaRa221Ra2211cossin2aRaRR2212sin1cos22oaaR202sin2a
Now the equation (2) can be represented in a form
. (2')
Then compute the shear stress on a surface of a particle:
, or
. (4)
and substituted the (
), (3), (4) into the equation (1), we get the following
, (5)
with account
. Here
are dynamic viscosity coefficient and velocity of vapor,
- derivative by time from a mass of vapor (vapor mass rate).
Thermodynamic equilibrium of a particle and a coolant
According to the above-mentioned and the schematic of the model physical situation (Fig. 3), the
heat balance for the vapor flow in a layer between the particle and liquid coolant of consideration is as
follows (difference of enthalpies between vapor and liquid spent in a vapor flow)
where from, because of (6) we obtain
, (6)
. (7)
Here
are enthalpies of the vapor and liquid coolant, respectively,
- temperatures of the
vapor and liquid,
- the coefficient of the heat conductivity in a vapor flow. The local temperature of
a vapor and particle are considered equal due to the small size of the particle and comparably fast local
temperature equilibrium process between vapor and particle.
Momentum conservation for the elementary volume
Now the equation of the momentum conservation is considered for the elementary volume of a
vapor in a layer between the cooling particle and liquid coolant pool as shown in the Fig. 4:
, (8)
5
20122sin2vlppghRR3vvwV.232sinvvwvmR'2.220210222(sinsin)cos3sin3222RvvRplvmgRpgRdRR22sin2hhR,vvV.vm2.sin()2()vvvlvlkRhhdmTTd2.()sin2()vvlvvlkTTRdhhdm,vlhh,vlTTvk222()sinsinsin0wlvvRdgRddpR
. (10)
, (9)
is a shear stress of a vapor layer with a liquid coolant,
where
is a coefficient (unknown for the
moment).
Fig. 4 To the mathematical model of a particle's cooling
With an account of the equations (8) - (10), yields:
,
where from:
,
;
which accounting the (6) further results in the following:
or
,
122sincossin222vlvdpgRdgRd.24()sin24sin2vvvvwlvVmRal.222222sinsinsin2sinvvvlvmRdgRdgRdR.222sinsin2vvvlvmdgdgdR.1113331233[2()]sinvvvlvmRg.1112.2133333sin2()2()sinvvlvvvvlvvlkRTTmdmRgdhh
. (11)
4. Analysis of the mathematical model for a particle in a volatile coolant
A solution of the equation
Now we can perform the integration of the equation (11) thus obtained with the initial condition
(zero mass flow rate at the start of the particle's cooling process):
. (12)
Substituting the expression (8) into (5) yields the following equation for the function
:
. (13)
We have two equations for function
- (10) and (12), and the equation (13) for
. The unknowns
here are: the length h of a particle penetration into a coolant (the corresponding angle is
), the initial
width of a vapor film on a particle
, and the radius of an equivalent circle
(or
).
To close the equation array (8), (12), (13), we need to compute
, (14)
Based on the given values of the parameters
, where
are the temperature
of the particle and heat transfer coefficient from particle to a coolant. Thus, substituting (14) into (10),
(12) results in
, (15)
.
7
.11373.53333()()2sin()vvlvvlvvvvlkRgTTmdmdhh.,00vm123334.31340()()23sin4()vvlvvlvvvlkRgTTmdRhhh22222012112sinsinsin4sincos3262hplhhlhhgRgRgRR2020021ln(1sin3()212()1sin2hvvvlhvvlahRTTahha.vmhh01R10aRR22.,2sin()2()hpwlvhvlRTTmhh,,,,,,wlvlhpTThhR,wpT2202expsin1sin222hhpvaak123343434530sin()()2()3sin4hhvvlvvlvpwlkghhRTTd With estimation made for a small length of a particle penetration into the cooling pool,
Analysis of the solution obtained
,
, therefore from (15), (13) follows:
,
, (16)
.
Thus, (16) determines three main parameters of the process under investigation:
for the
given physical data. Because of
, and
, we can consider two
cases:
and
. (17)
The last case in (17) better corresponds to our conditions, especially for heavy particles, when contact
area of the particle with coolant is larger. Therefore:
1)
2)
,
,
,
,
,
,
;
,
: а)
; б)
;
.
Further according to the expressions obtained we can analyze the main parameters of the process for
each specific case.
5. Conclusion
The developed mathematical model fits for a rough approximation of the particle's cooling in a
volatile liquid. The precise model is complex and requires the solution of the nonstationary two-phase
22sin24hh2233sin143333()()40,45()vlvvlhvpvwlghhkRTT20exp212phvaak220020001ln(1)3()222132()12hhvvwlphvvlahRTTgRpaRahha0,,ha23/1010pvNuak0,20,5h2/21hNu2/21hNu2hpvak1433()()40,45()vlvvlhvpvwlghhkRTT024vphk0a2000()223()23vvlhpvvwlhhagRhTTR2hpvak1433()()40,45()vlvvlhvpvwlghhkRTT202exp2phvaak0a22exp142phhvak22exp142phhvak2222200003()3()2203232()()hhvvwlvvwlpphvvlvvlRhTTRhTTgRpagRpRaRhhhh
flow equations with the conjugate heat transfer problem for the particle, vapor, and liquid cooling
pool. Vapor is permanently produced and removed from the coolant's pool. The process is going until
the moment when a temperature of the particle during its cooling falls down below the vaporization
temperature. Then particle is further cooled by direct contact with the cooling liquid.
References
I.V. Kazachkov, Parametric wave excitation and suppression on the interfaces of continua, D.Sc. thesis,
I.V. Kazachkov, A.F. Kolesnichenko, Pis'ma v Zh. Tekhn. Fiziki (Letters to J. Techn. Phys.) 1991, 17, 3,
I.V. Kazachkov, A.F. Kolesnichenko, V.P. Kuchinsky, et al. USSR Patent 1814254 1992.
I.V. Kazachkov, A.F. Kolesnichenko, V.P. Kuchinsky, et al. USSR Patent 1570154 1991.
[1] A.F. Kolesnichenko, I.V. Kazachkov, V.O. Vodyanuk and N.V. Lysak Capillary MHD Flow with Free
Surfaces, Naukova Dumka, Kiev 1988 (in Russian).
[2] A.F. Kolesnichenko, I.V. Kazachkov, Yu.M. Gorislavets, et al., Liquid metal MHD, Kluwer Acad. Publ.
Holland 1989, 293.
[3]
Kiev, Institute of Electrodynamics 1989 (in Russian).
[4]
40.
[5]
[6]
[7] V.O. Vodyanuk, I.V. Kazachkov, A.F. Kolesnichenko, et al. USSR Patent 1476736 1990.
[8]
I.V. Kazachkov, A.F. Kolesnichenko, V.P. Kuchinsky, et al. USSR Patent 1412135 1989.
[9]
I.V. Kazachkov, A.F. Kolesnichenko, V.P. Kuchinsky, et al. USSR Patent 1184153 1988.
[10] A.F. Kolesnichenko, Technological MHD facilities and processes, Naukova Dumka, Kiev 1980 (in
Russian).
[11] Yu.M. Gorislavets, V.V. Malakhov, A.I. Glukhenkii, Magnetohydrodynamic installation for the production
of lead shot// J. Tekhnichna elektrodynamika.- 1997.- 6.- P. 68-69 (in Russian).
[12] Yu.M. Gorislavets, A.I. Glukhenkii, V.M. Mykhalskyi, A.V. Tokarevskyi, Installation of electromagnetic
batching of liquid metal with high productivity// J. Tekhnichna elektrodynamika.– 2012.- 5.- P. 74-80 (in
Russian).
[13] А.Ya. Voloshyn, I.N. Ivanochok, I.V. Kazachkov, A.F. Kolesnichenko, V.P. Kuchinsky, V.V. Malakhov,
Electrodynamic excitation of the liquid film decay and development of the perspective MHD granulators- Kiev:
In-t Electrodynamics.- Preprint N587.- 1988.- 38 pp. (In Russian).
[14] А.I. Glukhenkii, I.V. Kazachkov, A.F. Kolesnichenko, V.P. Kuchinsky, V.V. Malakhov, Electromechanical
control of the jets and films of viscous fluid and development of perspective granulators.- Kiev: In-t
Electrodynamics.- Preprint №527.- 1988.- 40 pp. (In Russian).
[15] Yu.M. Gorislavets, I.V. Kazachkov, A.F. Kolesnichenko, et al. Development of the methods for MHD
granulation of Aluminum alloys.- Kiev: In-t of Electrodynamics.- Report.- 1987.- 132 pp. (In Russian).
[16] P. Vesterberg, K. Beskow, C-J. Rick. Granulation of ferroalloys - results from industrial operations and
comparative study on fines generation. Efficient technologies in ferroalloy industry/ The thirteenth International
Ferroalloys Congress. June 9-13, 2013.- Almaty, Kazakhstan./ P. 64-72.
[17] David Norval. Metals and alloys water solidification a bridge from pyro- to hydro-metallurgy/ The South
African Institute of Mining and Metallurgy The Third Southern African Conference on Base Metals.- P.97-106.
[18] Per-Åke Lundström, Conroy van der Westhuizen, Roelof Hattingh, and Mårten Görnerup Pig Iron
Granulation at Iscor Saldanha Steel/ AISTech 2004 Proceedings.- Vol. 1.- P. 517-524.
[19] Kazachkov I.V. and Ali Hasan Moghaddam. Modeling the thermal hydraulic processes by severe accidents
at the NPPs.- Monograph.- Kyiv, NTUU "KPI".- Zhytomyr.- Polissya.- 2008.- 170pp. (In Russian).
9
|
1805.01570 | 2 | 1805 | 2018-06-28T20:17:03 | Aqueous ion trapping and transport in graphene-embedded 18-crown-6 ether pores | [
"physics.app-ph",
"cond-mat.mes-hall"
] | Using extensive room-temperature molecular dynamics simulations, we investigate selective aqueous cation trapping and permeation in graphene-embedded 18-crown-6 ether pores. We show that in the presence of suspended water-immersed crown-porous graphene, K+ ions rapidly organize and trap stably within the pores, in contrast with Na+ ions. As a result, significant qualitative differences in permeation between ionic species arise. The trapped ion occupancy and permeation behaviors are shown to be highly voltage-tunable. Interestingly, we demonstrate the possibility of performing conceptually straightforward ion-based logical operations resulting from controllable membrane charging by the trapped ions. In addition, we show that ionic transistors based on crown-porous graphene are possible, suggesting utility in cascaded ion-based logic circuitry. Our results indicate that in addition to numerous possible applications of graphene-embedded crown ether nanopores, including deionization, ion sensing/sieving, and energy storage, simple ion-based logical elements may prove promising as building blocks for reliable nanofluidic computational devices. | physics.app-ph | physics | Aqueous Ion Trapping and Transport in Graphene-embedded 18-crown-6 Ether Pores
Alex Smolyanitsky*, Eugene Paulechka, Kenneth Kroenlein
Applied Chemicals and Materials Division,
National Institute of Standards and Technology
Boulder, CO 80305
*To whom correspondence should be addressed: [email protected]
Abstract
Using extensive room-temperature molecular dynamics simulations, we investigate
selective aqueous cation trapping and permeation in graphene-embedded 18-crown-6 ether pores.
We show that in the presence of suspended water-immersed crown-porous graphene, K+ ions
rapidly organize and trap stably within the pores, in contrast with Na+ ions. As a result,
significant qualitative differences in permeation between ionic species arise. The trapped ion
occupancy and permeation behaviors are shown to be highly voltage-tunable. Interestingly, we
demonstrate the possibility of performing conceptually straightforward ion-based logical
operations resulting from controllable membrane charging by the trapped ions. In addition, we
show that ionic transistors based on crown-porous graphene are possible, suggesting utility in
cascaded ion-based logic circuitry. Our results indicate that in addition to numerous possible
applications of graphene-embedded crown ether nanopores, including deionization, ion
sensing/sieving, and energy storage, simple ion-based logical elements may prove promising as
building blocks for reliable nanofluidic computational devices.
Keywords: graphene, crown ether, ion transport, ionic transistor, nanofluidics
1
Crown ethers1 are a family of electrically neutral cyclic ethylene oxide molecules, which
exhibit marked selective affinity for metal cations. A representative example of a crown ether is
shown in Fig. 1 (a). The underlying nearly circular van der Waals coordination allows for a rich
variety of "host-guest" binding preferences, depending on the crown size and composition.
Unsurprisingly, after the initial discovery of crown ethers more than half a century ago,
numerous intriguing applications were proposed, ranging from nanoscale self-assembly 2 to
cation sensing and separation.3-5
Figure 1. A stand-alone 18-crown-6 ether with an embedded potassium ion (a) and its graphene-
embedded analog (b).
The hexagonal symmetry of graphene naturally lends itself to the possibility of
embedding various types of crown-like pores in a graphene layer. Fabrication of such pores via
vacancy oxidation was recently reported.6 In particular, a hexagonally symmetric 18-crown-6
pore in graphene is shown in Fig. 1 (b). It can be viewed geometrically as a result of removing an
entire carbon hexagon, followed by replacing the remaining edge carbons with oxygen atoms.
From the chemical standpoint, introduction of an 18-crown-6 pore results in formation of
additional defects (or healing of existing defects) in the graphene sample – see section S4 of the
Supporting Information (SI). Such a pore is expected to exhibit significant binding preference for
aqueous K+ ions, as shown in the aqueous titration studies of stand-alone crown ethers7 and from
density functional theory (DFT) calculations performed in vacuum for the graphene-embedded
18-crown-6 structure analog.6 Given the potential utility of ultra-narrow nanopores in atomically
2
thin membranes for applications ranging from water filtration and sensing8-9 to energy storage,10-
11 these membrane-embedded pores are expected to open various additional possibilities.
All liquid-phase studies to date have been focused on the cation-binding properties of
stand-alone crown ethers. Currently, given a realistic possibility of fabricating atomically thin
membranes with embedded crown ethers6 and similar structures,12-13 ion-binding and permeation
properties of such membranes may prove promising for a wide range of applications. Here, with
the use of molecular dynamics (MD) simulations, we investigate aqueous binding and transport
properties of the 18-crown-6 pores embedded in monolayer graphene, immersed in aqueous
solutions of NaCl and KCl of varying concentrations. Our results confirm zero anion
permeability of the ether-porous membranes, as well as demonstrate semiconductor-like
transport of K+ ions, while transport of Na+ ions follows the expected Michaelis-Menten
saturation.14 Our findings show potential promise of membrane-embedded crown ether pores for
numerous applications, including water filtration and fundamental studies of ionic hydration at
the deep nanoscale. Importantly, we demonstrate that simple ion-based logical operations can be
performed using graphene-embedded crown pores.
Figure 2. An example of the simulated system: graphene membrane with an embedded array of
20 18-crown-6 ethers, in an aqueous ionic bath. Dissolved ions are shown as solid spheres and
water molecules are transparent for clarity. The system dimensions are 5.5 nm × 6.4 nm × 5.0
nm.
3
Results and discussion
The model systems investigated in this work are based on a 5.5 nm × 6.4 nm suspended
graphene sheet with one or multiple embedded 18-crown-6 pores. The sheet was restrained along
its perimeter and immersed in a rectangular periodic aqueous ionic bath (Fig. 2).
We begin by investigating aqueous ion dynamics in the presence of a room-temperature
graphene membrane with nine embedded 18-crown-6 ether pores. Separate simulations of 0.15
M KCl and NaCl were performed for 100 ns. A representative initial state of the system is shown
in Fig. 3 (a). In the case of KCl, all pores become populated by K+ ions (Fig. 3 (b)) within the
first 50 ns of the simulated time and such a configuration remains stable throughout the
remainder of the simulation. In contrast, this behavior is not observed for NaCl, as expected from
the 18-crown-6's binding preference for potassium cations in water.7 We consider the energetics
of the observed ion assembly and trapping further by combining umbrella sampling simulations
and the weighted histogram analysis method (WHAM)15 to extract the graphene-embedded
crown ether's binding Gibbs free energy to aqueous K+ and Na+ ions.
Figure 3. Initial (a) and final (b) state of a graphene membrane featuring an array of nine ether
pores, as obtained in a room-temperature molecular dynamics simulation in aqueous 0.15 M KCl
solution. No stable ion trapping was observed in an identical system in NaCl solution.
The results for the binding Gibbs free energy ΔG as a function of "reaction coordinate"
(here, distance along the Z-axis from the membrane) are shown in Fig. 4 (a). For both cation
4
species, we observe ΔG minima and maxima corresponding to ions' interaction with the pore, as
well as the hydration-induced transfer barrier, essentially delineating the transition of the ion
from bulk solution to the pore confinement. The Gibbs free energy minima corresponding to the
equilibrium binding site are -14.4 kJ/mol and -3.8 kJ/mol for K+ and Na+, respectively – in
reasonable agreement with the corresponding calorimetric data of -11.72 kJ/mol and -4.79
kJ/mol obtained for a stand-alone 18-crown-6 ether.7 As expected, a significantly larger affinity
for K+ ions is observed, so that the probability of trapping K+ ions is exp (
ΔG𝐾+−ΔG𝑁𝑎+
𝑘𝑏𝑇
) ≈ 25.5
times higher than that of Na+ (𝑘𝑏 and T = 300 K are the Boltzmann constant and the system
temperature, respectively). Above, ΔG𝑁𝑎+ = 12.4 kJ/mol and ΔG𝐾+ = 20.5 kJ/mol are the
"peak-to-peak" Gibbs energies, including the transfer barriers in Fig. 4 (a). The significant
planarity-induced binding enhancement in vacuum relative to stand-alone ethers6 is not observed
for a membrane suspended in room-temperature solvent. Solvent screening is expected to
account for most of the binding energy reduction in our simulations, as well as in calorimetric
measurements of free aqueous crowns. In addition, the static density functional theory
calculations6 considered a planar graphene structure, and thus flexural relaxation of the pore
region necessary for suspended graphene at finite temperature was excluded. At the same time,
immediate trapping of K+ ions and a low "bulk-to-pore" transfer barrier shown in Fig. 4 (a) are
consistent with predictions in,6 as graphene-embedded crowns indeed do not require significant
radial stretching to trap an ion.
It is worth noting that the ΔG minimum for Na+ in Fig. 4 (a) is located away from the
plane of the ether-containing membrane. To investigate this observation further, we plotted the
cation-water radial distribution functions g(r) for each ionic species inside and outside of the
crown ether pore. The results are shown in Fig. 4 (b), where the distributions for Na+ inside the
5
pore and in bulk solution are quite similar, whereas for K+ a considerable reduction in hydration
is observed inside the pore. Given a generally more stable hydration shell around Na+,16-17 non-
preferential binding and thus a "wobbly" host-guest interaction for the Na+ ion results in slight
deformation of its hydration shell while the ion resides in the ether pore. The presence of the
deformed hydration shell around Na+ ion then results in the corresponding ΔG minimum position
away from the membrane plane. The observed contrast with the "tight fit" observed between K+
and the 18-crown-6 ether is consistent with earlier results18 and, together with the considerably
larger absolute value of ΔG minimum for K+, can be viewed as a clear manifestation of the host-
guest ion recognition in the case of an ether embedded in a membrane.
Figure 4. MD-simulated binding Gibbs free energy as a function of ion's distance from the center
of the crown-like nanopore (a) and cation-water RDF curves for Na+ and K+ (b).
It is important to note that the highly selective K+ trapping behavior observed here is beyond
merely capturing ions from the solution. After effectively becoming a stable part of the
membrane, potassium ions not only block the membrane for permeation, but also result in a
"macroscopic" solution-exposed sheet charge, leading to a long-range electrostatic barrier in
addition to those discussed above (Fig. 4). The effect of this sheet charge on the system
energetics can be estimated. The magnitude of the effective electric field induced by a sheet
6
charge density 𝜎 is 𝐸 = 𝜎/2𝜀𝜀0, where 𝜀 and 𝜀0 are the effective dielectric constant of water
near the membrane and the vacuum permittivity, respectively. With a total of nine K+ ions
embedded in a 5.5 nm × 6.4 nm membrane, the sheet charge density is 0.04 C/m2 and thus with
bulk 𝜀 = 80 we find 𝐸 ≈ 30 MV/m. For example, at a distance of 1 nm away from the membrane,
such a field introduces a non-negligible potential difference of 30 mV relative to the membrane
plane. In simulations of a membrane containing a total of 20 embedded crown ether pores
(system shown in Fig. 2), similar K+ trapping was observed, increasing the sheet-induced field
estimate above by a factor of greater than two. Note that the presented calculations likely
underestimate the effective local field, because 𝜀 of a polar liquid is known to be spatially
distributed near a charged solid surface and reduced by as much as an order of magnitude in the
0.2-0.3 nm wide interface region, compared to the bulk value of 80.19 The implications of the
sheet charge induced by the trapped ions also result in interesting possible applications, as
discussed later.
Resulting from the field induced by the trapped charge, an electric double layer (EDL) is
expected to form in an aqueous salt solution between the ion-containing membrane and mobile
counterions. EDL formation in the case of KCl is confirmed in Fig. 5, where we plot the time-
averaged ion densities (averaged per XY-slice, each parallel to the membrane) as functions of
distance along the Z-axis for different ion species. In contrast, because the membrane occupancy
by Na+ ions is significantly lower, virtually no EDL formation is observed for NaCl, and overall
considerably weaker charge polarization is observed throughout the interfacial region.
7
Figure 5. Ion density distributions in 0.15 M simulations of NaCl and KCl. The membrane is
positioned at Z = 0.
EDL formation essentially replaces the short-range van der Waals interactions between
ions and the individual unoccupied pores with longer-range interactions. Interestingly, the
system presented here can be viewed as somewhat of an opposite of the porous graphene
considered earlier, where unoccupied pores with carboxyl interiors exhibited no ion trapping
behavior and carried pH-dependent electric charge.20 It is also different from the protein ion
channels embedded in relatively thick lipid bilayers and featuring buried charges responsible for
their selectivity. We see that the effective ether pore occupancy (the average time an ion is
trapped by the membrane) is responsible for the presence of the EDL at steady-state, and thus for
the interactions in the system, as well as possible effects thereof on the ionic transport. As a
result, one can expect significant differences in ionic permeation and conduction selectivity,
depending on the ion species. Ion transport induced by externally applied transmembrane fields
is considered next.
We performed a series of MD simulations of ionic transport via crown ether pores
embedded in graphene membranes. Different NaCl and KCl concentrations (0.15 M, 0.5 M, and
1 M) were simulated. Each simulated current value was obtained from a 200 ns-long simulation
8
and an external bias in the form of a constant electric field was applied perpendicularly to the
membrane (along the Z-axis). The external field magnitude 𝐸𝑒𝑥𝑡 varied in the range 0-100 MV/m
in 10 MV/m increments. For simplicity of presentation, we report all currents as functions of the
"transmembrane" potential drop across the simulated system height (along the Z axis), which in
this case is 𝐸𝑒𝑥𝑡 × h, where h = 5 nm. The per-channel current at each value of transmembrane
potential was calculated from the slope of a straight line fitted to the cumulative ion flux via
membrane, rescaled by the number of channels (also, see section S1 of the SI). The results are
shown in Fig. 6.
Figure 6. Simulated current-voltage curves for K+ (a) and Na+ (b); no anion transport was
observed regardless of concentration, or the external bias magnitude, suggesting that 18-crown-6
pores are infinitely cation-selective. The insets show corresponding current estimates obtained
from the theoretical model described in the text.
For KCl, virtually no current is observed in Fig. 6 (a) at voltages ≤ 150 mV and the
membrane is in the "off" permeation mode regardless of the concentration, followed by a
transition to the "on" mode at voltages ≥ 300 mV. Essentially, at low voltages the membrane
remains blocked by the trapped ions, while the process of knocking out the trapped ions by the
incoming mobile ions is likely suppressed by the accompanying presence of the EDL. As the
9
transmembrane voltage increases, the probability of thermally induced dissociation of the
trapped ions increases, while knock-on events also become more prominent, once again
compounded by the weakening EDL (the effective occupancy decreases, thus reducing 𝜎). The
results for NaCl in Fig 6 (b) are different qualitatively, because no ion trapping occurs and the
additional effects of the EDL are less pronounced – we revisit this point in greater detail later
when discussing the data shown in the Fig. 6 insets. Given the simulated data in Fig. 6, 18-
crown-6 pores are sodium-selective in terms of permeation, as shown in Fig. 7. For low voltages,
the selectivity ratio varies from 20 to 35, eventually approaching unity at high bias values, as the
membrane occupancy by K+ decreases (see inset of Fig. 7) and the conduction regimes become
identical for both ionic species.
Figure 7. Na+/K+ current selectivity for various salt concentrations, as obtained from the data
presented in Fig. 6. The inset shows K+ ion distributions at different transmembrane voltages for
a 0.15 M KCl simulation.
Our ionic permeation findings are in contrast with those obtained for graphene-embedded
carboxyl pores,20 where qualitatively similar trends were observed for Na+ and K+. The current-
voltage trends observed here, however, can be qualitatively explained with the use of existing
theoretical treatment of reaction kinetics. Although most analytical models lack a detailed
10
treatment of the interatomic interactions, hydration effects, mechanisms of mobile ions knocking
out trapped ions, or the effects of the EDL, qualitative insight can still be obtained for
illustration. Consider the association-dissociation model described in detail elsewhere.21 For the
18-crown-6 pores, strictly one ion permeates at a time, which allows for defining distinct binary
occupancy states of the channel. Assuming ionic concentration symmetry above and below the
membrane, the resulting current is
𝐼 =
𝑞𝑘𝑑c sinh (
𝑞𝑉
2𝑘𝑏𝑇
)
𝐾𝑑 cosh(
𝑞𝑉
2𝑘𝑏𝑇
)+𝑐
,
(1)
where q is the cation charge, 𝑘𝑑 is the ion-pore dissociation rate, 𝐾𝑑 = 𝑘𝑑/𝑘𝑎 is an effective
constant determining the association rate 𝑘𝑎 for each ionic species, and c and V are the salt
concentration and transmembrane voltage, respectively. Mathematically, the expression above is
of the Michaelis-Menten type,14 with saturation occurring when the association term in the
denominator begins to approach the voltage-dependent dissociation term in the numerator at a
given concentration and voltage.
With 𝑘𝑑 = 1.2 × 107 s-1, 𝐾𝑑 = 0.011 M and 𝑘𝑑 = 1.8 × 108 s-1, 𝐾𝑑 = 0.37 M selected as
rough fitting parameters for K+ and Na+, respectively, theoretical estimates for each ion species
are provided in the corresponding insets of Fig. 6 (a) and (b). Because our aim here is basic
illustration, we deliberately avoid plotting the theoretical estimates alongside the simulated data,
or fine-tuning the model parameters to quantitatively fit the simulated current-voltage curves.
Theoretical curves reveal that the qualitative difference between K+ and Na+ permeation trends is
due to the large difference in the ion-pore dissociation rates. For a channel that allows
permeation of only one ion at a time, the current is essentially 𝑞/𝜏, where 𝜏 is the effective
average ion transition time. The latter can be defined as 𝜏 = 𝜏𝑑 + 𝜏𝑎, where 𝜏𝑑 is the average
11
time an ion spends trapped in the pore (dissociation time) and 𝜏𝑎 is the salt concentration
dependent effective time it takes for an ion to enter the pore from bulk solution (association
time). For K+ at low voltages, 𝜏𝑑 ≫ 𝜏𝑎 and thus, regardless of 𝜏𝑎 (and of the ionic concentration)
𝜏 ≈ 𝜏𝑑, confirmed in Fig. 6 (a). For Na+, 𝜏𝑑 is considerably shorter (at zero bias, 𝜏𝑑,𝐾+/𝜏𝑑,𝑁𝑎+ =
𝑘𝑑,𝑁𝑎+/𝑘𝑑,𝐾+= 15, of the same order as ~25.5 calculated earlier from the Gibbs energies),
allowing considerable permeation at small voltages and further decreasing upon increasing bias.
At high bias, 𝜏𝑑 eventually becomes smaller than 𝜏𝑎, which leads to association-limited
saturation.
The distinguishable K+ conduction regimes discussed above and especially the
dependence of membrane effective ion occupancy on the transmembrane voltage (see inset of
Fig. 7) immediately suggest an intriguing possibility of performing some elementary ion-based
logical operations. Direct electrical measurement of the membrane potential is then effectively a
read operation. As follows from earlier discussions, the time-averaged trapping occupancy (or
charge) is estimated to be
𝑞0
1+
𝜏𝑎
𝜏𝑑
, where 𝑞0 = +𝑁 × 𝑒 is the maximum trapped charge in a
membrane with N crown pores. At sufficiently low KCl concentration, such that 𝜏𝑎 ≫ 𝜏𝑑 at large
transmembrane bias, the highly conductive ("on") regime coincides with low occupancy, and
vice versa. The effect of trapped charge variation on the electrical potential in the region
occupied by the membrane was simulated directly using MD at various transmembrane
potentials. The time-averaged results are shown in Fig. 8 (a). As simulated, the membrane
potential indeed decreases by more than 100 mV with increasing transmembrane bias; 𝑑𝑉𝑜𝑢𝑡/
𝑑𝑉𝑖𝑛 is shown to be lower at the higher salt concentration due to overall shorter association
times 𝜏𝑎, as expected. Note that the trapping-induced maximum potential at low transmembrane
12
bias (proportional to the number of pores) can be increased with a higher pore count (see section
S2 of the SI). In the context of Figs. 6 (a) and 8 (a), consider a low transmembrane voltage Vin (<
200 mV), denoted "0," applied to a cell with appropriately selected KCl concentration, as shown
in the top pane of Fig. 8 (b). In this case, the cell is nearly non-conductive ("off") and the
membrane is fully occupied by the trapped ions. Therefore, the potential Vout measured directly
at the membrane is relatively high, denoted as "1." Conversely, when high Vin (> 300 mV),
denoted "1," is applied, the cell is highly conductive ("on"), fewer ions are trapped in the
membrane, and thus a low ("0") state of Vout is measured. With appropriately set "0/1"
thresholds, the input-output relationship here can be viewed as the Boolean NOT operation. As a
straightforward extension, exclusive OR (XOR) operation is possible with two independently
driven cells shown in the bottom pane of Fig. 8 (b). When measuring the absolute value of
differential Vout between the two membrane states, one indeed observes state "1" only when
either of the two cells is highly conductive, corresponding to a XOR operation.
The effective state switching speeds can be estimated. During the "on-off" transition, K+
ions begin to populate the pores, limited by the ionic diffusivity and thus the low-bias association
time 𝜏𝑎,𝑙𝑏. Switching in the opposite direction is limited by the high-bias dissociation time 𝜏𝑑,ℎ𝑏.
The order of magnitude for 𝜏𝑎,𝑙𝑏 and 𝜏𝑑,ℎ𝑏 can be estimated from the permeation data in Fig. 6
as (𝑞/𝐼) at corresponding bias strengths, where q is the cation charge and I is the single-channel
current (also, see supplementary movie of K+ ions trapping in an initially empty membrane). For
the pore spacing presented here, at 0.5 M KCl the characteristic time limit 𝑚𝑎𝑥(𝜏𝑎,𝑙𝑏, 𝜏𝑑,ℎ𝑏) is of
order of a few nanoseconds, corresponding to hundreds of MHz.
13
Figure 8. Membrane potential as a function of transmembrane voltage, simulated at 0.15 M and
0.5 M KCl (a), and simple NOT and exclusive OR (XOR) logical operations obtained from
measuring the effective membrane potential in the "on" and "off" permeation modes (b). For
clarity, a baseline potential of ≈ 550 mV, arising from water ordering at the graphene-water
interface 22, was subtracted from all simulated Y-axis values in (a) to yield potential close to zero
at the highest transmembrane voltage, corresponding to the lowest ion occupancy.
The ion-based logic examples presented above are attractive due to their conceptual
simplicity, compared with the devices proposed earlier,23-24 especially given that the distinction
between the "on" and "off" permeation regimes is achievable with voltage magnitudes induced
by salt concentration gradients often found in biological systems. More sophisticated cascaded
logic with ionic diodes and transistors may then be high interest. For instance, in multilayer
systems featuring membranes with different numbers of embedded crown ether pores, the
resulting asymmetry (including electrostatic asymmetry) suggests the possibility of diode-like
current rectification, similar to nanofluidic diodes described earlier.25-28 Here, we explore the
possibility of electrostatically gating the cell shown in Fig. 8, essentially resulting in a simple
ionic transistor.
In the case of ultra-narrow pores in graphene (as well as in other conductive atomically
thin materials), immediate exposure of the membrane to the surrounding ionic solution suggests
14
an opportunity for control of the EDL and thus for gated ionic flow. The control voltage is
applied directly to the membrane, as described schematically in Fig. 9 (a). In contrast with
buried-gate ionic transistors featuring field-induced narrowing of the relatively wide conducting
channels 29-30, the effective "supply" of ions to the pores (i.e. the effective association rate in Eq.
(1)) is subject to control here.
To assess tuneability of the ionic current, we mimicked a gate voltage applied directly to
the graphene membrane. Because it is impossible to set specific voltages in simulations
involving periodic boundaries and Ewald summation for solving Poisson's equation, we did so
by simulating delocalized excess charge externally injected into the membrane in the form of a
small nonzero charge of every "bulk" carbon atom. However, we show below that it is possible
to analytically estimate the effective gate voltage associated with this excess charge. The
resulting system was neutralized by modifying the number of Cl- ions accordingly. The number
of dissolved cations was left unchanged for direct comparison of the current with the floating
gate case (corresponding data from Fig. 6). The effective "excess" gate voltage variation (in
addition to that arising from the charge induced by trapped ions, which, as discussed earlier,
depends on the transmembrane voltage) relative to "drain," can be estimated analytically: 𝑉𝑔 =
𝛥𝑄
2𝐴𝜀𝜀0
×
h
, where 𝛥𝑄 and 𝐴 are the total externally induced charge and the area of the membrane,
2
respectively. For a relatively low 𝛥𝑄 = 10e and 1136 "bulk" carbons in the simulated
membrane, the per-carbon charge is ~8.803 × 10-3e. The corresponding "excess" gate voltage 𝑉𝑔
is then estimated to be 80 mV ≈ 3kbT. The current-voltage curves for 0.5 M KCl and 𝛥𝑄 = ±10e
were simulated in two series of independent 200 ns long simulations. The results, along with the
𝛥𝑄 = 0 case (0.5 M KCl data from Fig. 6 (a)), are shown in Fig. 9 (b).
15
Figure 9. Ionic transistor schematic (a) and current-voltage curves (0.5 M KCl) for various
values of 𝛥𝑄 (b). The inset in (b) shows K+ density distributions at various 𝛥𝑄 and zero
transmembrane voltage.
As shown, current modulation is significant even with a gate voltage variation of the order of a
few kbT. For example, at the transmembrane voltage of 300 mV, the estimated effective per-
channel transconductance is >70 pS, suggesting that with an appropriately selected transistor
load and an accurate "on/off" output voltage threshold, sensitive switching may be possible. As
shown in the inset of Fig. 9 (b), the mobile K+ population adjacent to the membrane, central to
the gating effect on current in this system, is indeed modulated. Qualitatively, it may be viewed
as replacing the value of c in Eq. (1) with 𝑐 × exp (−
𝑞𝑉𝑔
𝑘𝑏𝑇
). It can be shown that, given Eq. (1),
the transconductance 𝑔 =
𝑑𝐼
𝑑𝑉𝑔
exhibits resonant behavior with respect to c and thus can be further
optimized in terms of salt concentration for a selected transmembrane voltage (see section S3 of
the SI). Overall, the observed current sensitivity to external gating is likely sufficient for
implementing cascaded ion-based logic in the form of ionic circuit elements supplying control
signals, as well as those subject to control.
16
Conclusions
We performed extensive MD simulations to investigate ion trapping and permeation in
graphene-embedded 18-crown-6 pore arrays immersed in aqueous KCl and NaCl. In agreement
with the experimental data for stand-alone 18-crown-6 ethers in water, we estimated the binding
Gibbs energies at -14.4 kJ/mol and -3.8 kJ/mol for K+ and Na+, respectively. Spontaneous stable
trapping of K+ ions in the crown pores was shown at room temperature. Externally induced
permeation across membranes featuring embedded crown ether pores reveals qualitatively
different current-voltage characteristics for KCl and NaCl. For KCl, "on" and "off" conduction
regimes exist due to the presence of an explicit permeation "gap" (in the form of the ion-pore
binding energy), accompanied by the presence of EDL, arising from the trapped ionic charge,
especially at low (≤ 150 mV) transmembrane voltages. For NaCl, no cation trapping occurs and
Michaelis-Menten-type current-voltage behavior is observed. Given the K+-selective trapping,
permeation is shown to be highly Na+-selective, with Na+/ K+ selectivity ratio decreasing as the
transmembrane voltage increases. For both NaCl and KCl, permeation is demonstrated to be
infinitely cation-selective, regardless of salt concentration.
Interestingly, we demonstrated utilization of membrane charging by trapped K+ ions for
ion-based logic, and simple examples of NOT and XOR gates were presented. Together with a
sensitive ionic transistor, also shown in this work, cascaded ion-based logic may be possible. Our
results therefore suggest a host of potential applications. Stable ion trapping, chemically tunable
to target various ion types, may be utilized for energy storage, highly selective gas- and liquid-
phase ion sensing, as well as for deionization against specific ion species. Ether pores embedded
in suspended graphene may also be used to study ionic solvation at a single ion resolution.
Voltage-tunable trapping of ions in membrane-embedded crowns may hold promise for storing
17
information, while sophisticated ion-based logic from combining conceptually simple state-
switching ionic cells and ionic transistors may be achieved. Potential applications are of course
not limited to graphene, and membrane-embedded crown-like pores in other atomically thin
materials may be of even greater interest in terms of the structure-function relationship.
Transition metal dichalcogenides and boron nitride are among the candidates for applications in
liquid and gas phase, as a wide range of membrane pore structures and hydrophobicity levels is
expected in these materials.
Methods
The intramolecular model for graphene was based on a harmonic potential informed by
the optimized bond-order potential,31 as utilized earlier.32-34 All intermolecular interactions in the
system were based on the well-established OPLS-AA forcefield,35-36 including the partial atomic
charges (see section S4 of the SI for details) and the Lennard-Jones parameters, responsible for
the interactions between ions and the crown ether pores. The TIP4P model37-38 was used to
represent water molecules. All MD simulations were performed using GROMACS v. 2016.4
software39-40 with GPU acceleration. Prior to production runs, all systems were pre-relaxed in
NPT simulations at T = 300 K and p = 0.1 MPa with a time step of 1 fs. All production
simulations were performed in an NVT ensemble with a time step of 2 fs for hundreds of
nanoseconds, as specified in the text.
Supporting Information
Supplementary details of the simulated system, methods, and additional results and discussion
(PDF).
Animation of MD-simulated atomic trajectories depicting 50 nanoseconds of room-temperature
aqueous ion dynamics (0.15 M KCl; water molecules omitted from visualization for clarity) in
18
Grote, Z.; Scopelliti, R.; Severin, K., Ph-Triggered Assembly of Organometallic Receptors for
Grote, Z.; Wizemann, H.-D.; Scopelliti, R.; Severin, K., Lithium Isotope Separation by 12-
Wei, W.; Xu, C.; Ren, J.; Xu, B.; Qu, X., Sensing Metal Ions with Ion Selectivity of a Crown
the presence of a locally suspended graphene membrane with a total of nine 18-crown-6 ethers
embedded. K+ and Cl- ions colored white and blue, respectively (AVI).
Acknowledgment
We thank A. Fang, J. A. Lemkul, B. I. Yakobson, and A. N. Chiaramonti for useful discussions.
Authors gratefully acknowledge support from the Materials Genome Initiative. This work is a
contribution of the National Institute of Standards and Technology, an agency of the US
government. Not subject to copyright in the USA. Trade names are provided only to specify
procedures adequately and do not imply endorsement by the National Institute of Standards and
Technology. Similar products by other manufacturers may be found to work as well or better.
References
Pedersen, C. J., Cyclic Polyethers and Their Complexes with Metal Salts. J. Am. Chem. Soc.
Ballesteros-Garrido, R.; de Miguel, M.; Domenech-Carbo, A.; Alvaro, M.; Garcia, H., Tunability
1.
1967, 89, 2495-2496.
2.
Lithium Ions. J. Am. Chem. Soc. 2004, 126, 16959-16972.
3.
Metallacrown-3 Complexes. Z. Anorg. Allg. Chem. 2007, 633, 858-864.
4.
Ether and Fluorescence Resonance Energy Transfer between Carbon Dots and Graphene. Chem.
Commun. 2012, 48, 1284-1286.
5.
by Alkali Metal Cations of Photoinduced Charge Separation in Azacrown Functionalized Graphene.
Chem. Commun. 2013, 49, 3236-3238.
6.
Chisholm, M. F., Crown Ethers in Graphene. Nat. Commun. 2014, 5, 5389.
7.
Christensen, J. J., Calorimetric Titration Study of the Interaction of Several Uni- and Bivalent Cations
with 15-Crown-5, 18-Crown-6, and Two Isomers of Dicyclohexo-18-Crown-6 in Aqueous Solution at
25.Degree.C And .Mu. = 0.1. J. Am. Chem. Soc. 1976, 98, 7620-7626.
8.
Subnanometre Trans-Electrode Membrane. Nature 2010, 467, 190-193.
9.
Commun. 2016, 7, 11408.
10.
Aluru, N.; Bashir, R., Electrochemistry at the Edge of a Single Graphene Layer in a Nanopore. ACS Nano
2013, 7, 834-843.
11.
Energy Storage. Adv. Energy Mater. 2016, 6, 1502159.
12.
Feng, X.; Müllen, K., A C216-Nanographene Molecule with Defined Cavity as Extended Coronoid. J.
Am. Chem. Soc. 2016, 138, 4322-4325.
13. Moreno, C.; Vilas-Varela, M.; Kretz, B.; Garcia-Lekue, A.; Costache, M. V.; Paradinas, M.;
Panighel, M.; Ceballos, G.; Valenzuela, S. O.; Peña, D.; Mugarza, A., Bottom-up Synthesis of
Multifunctional Nanoporous Graphene. Science 2018, 360, 199.
Beser, U.; Kastler, M.; Maghsoumi, A.; Wagner, M.; Castiglioni, C.; Tommasini, M.; Narita, A.;
Rollings, R. C.; Kuan, A. T.; Golovchenko, J. A., Ion Selectivity of Graphene Nanopores. Nat.
Banerjee, S.; Shim, J.; Rivera, J.; Jin, X.; Estrada, D.; Solovyeva, V.; You, X.; Pak, J.; Pop, E.;
Ji, L.; Meduri, P.; Agubra, V.; Xiao, X.; Alcoutlabi, M., Graphene-Based Nanocomposites for
Guo, J.; Lee, J.; Contescu, C. I.; Gallego, N. C.; Pantelides, S. T.; Pennycook, S. J.; Moyer, B. A.;
Izatt, R. M.; Terry, R. E.; Haymore, B. L.; Hansen, L. D.; Dalley, N. K.; Avondet, A. G.;
Garaj, S.; Hubbard, W.; Reina, A.; Kong, J.; Branton, D.; Golovchenko, J. A., Graphene as a
19
Nelson, P. H., A Permeation Theory for Single-File Ion Channels: One- and Two-Step Models. J.
Tybrandt, K.; Forchheimer, R.; Berggren, M., Logic Gates Based on Ion Transistors. Nat.
Yakobson, B. I., Asymmetric Diffusion in a Nonlinear Spatially Inhomogeneous Medium. J. Exp.
Johnson, K. A.; Goody, R. S., The Original Michaelis Constant: Translation of the 1913
Karnik, R.; Fan, R.; Yue, M.; Li, D.; Yang, P.; Majumdar, A., Electrostatic Control of Ions and
Hummer, G.; Pratt, L. R.; García, A. E., Free Energy of Ionic Hydration. J. Phys. Chem. 1996,
Ikuta, Y.; Maruyama, Y.; Matsugami, M.; Hirata, F., Probing Cations Recognized by a Crown
Abrashkin, A.; Andelman, D.; Orland, H., Dipolar Poisson-Boltzmann Equation: Ions and
He, Z.; Zhou, J.; Lu, X.; Corry, B., Bioinspired Graphene Nanopores with Voltage-Tunable Ion
14.
Michaelis–Menten Paper. Biochemistry 2011, 50, 8264-8269.
15.
Hub, J. S.; de Groot, B. L.; van der Spoel, D., G_Wham-a Free Weighted Histogram Analysis
Implementation Including Robust Error and Autocorrelation Estimates. J. Chem. Theory Comput. 2010,
6, 3713-3720.
16. Mähler, J.; Persson, I., A Study of the Hydration of the Alkali Metal Ions in Aqueous Solution.
Inorg. Chem. 2012, 51, 425-438.
17.
100, 1206-1215.
18.
Ether with the 3d-Rism Theory. Chem. Phys. Lett. 2007, 433, 403-408.
19.
Dipoles Close to Charge Interfaces. Phys. Rev. Lett. 2007, 99.
20.
Selectivity for Na+ and K+. ACS Nano 2013, 7, 10148-10157.
21.
Chem. Phys. 2011, 134, 165102.
22.
Zhou, H.; Ganesh, P.; Presser, V.; Wander, M. C. F.; Fenter, P.; Kent, P. R. C.; Jiang, D.-e.;
Chialvo, A. A.; McDonough, J.; Shuford, K. L.; Gogotsi, Y., Understanding Controls on Interfacial
Wetting at Epitaxial Graphene: Experiment and Theory. Phys. Rev. B 2012, 85, 035406.
23.
Molecules in Nanofluidic Transistors. Nano Lett. 2005, 5, 943-948.
24.
Commun. 2012, 3, 871.
25.
Theor. Phys. 1985, 89, 1830-1833.
26.
27.
Commun. 2011, 2, 506.
28.
Diodes Demonstrated in Conical Nanopores. J. Phys. Chem. C 2017, 121, 6170-6176.
29.
10 Nm Multiple Nanopores. Nano Lett. 2009, 9, 2044-2048.
30.
Modulation of Ionic Conductance of Cylindrical Silicon-on-Insulator Nanopore Array. J. Appl. Phys.
2010, 107.
31.
Order-Informed Harmonic Constraints. Nanotechnology 2014, 25, 485701.
32.
Graphene. RSC Adv. 2015, 5, 29179-29184.
33.
Functionalized Graphene Nanoribbons for Accurate High-Speed DNA Sequencing. Nanoscale 2016, 8,
1861-1867.
34.
Currents in Extended Reservoir Simulations. J. Chem. Phys. 2017, 147, 141102.
35.
Potential Functions for Proteins, Energy Minimizations for Crystals of Cyclic Peptides and Crambin. J.
Am. Chem. Soc. 1988, 110, 1657-1666.
36.
Jorgensen, W. L.; Maxwell, D. S.; Tirado-Rives, J., Development and Testing of the Opls All-
Atom Force Field on Conformational Energetics and Properties of Organic Liquids. J. Am. Chem. Soc.
1996, 118, 11225–11236.
Yakobson, B. I., Transport of H through Asymmetric Metal Layers. Phys. Scr. 1987, 36, 513.
Guan, W.; Fan, R.; Reed, M. A., Field-Effect Reconfigurable Nanofluidic Ionic Diodes. Nat.
Smolyanitsky, A., Molecular Dynamics Simulation of Thermal Ripples in Graphene with Bond-
Gruss, D.; Smolyanitsky, A.; Zwolak, M., Communication: Relaxation-Limited Electronic
Jorgensen, W. L.; Tirado-Rives, J., The Opls [Optimized Potentials for Liquid Simulations]
Smolyanitsky, A., Effects of Thermal Rippling on the Frictional Properties of Free-Standing
Paulechka, E.; Wassenaar, T. A.; Kroenlein, K.; Kazakov, A.; Smolyanitsky, A., Nucleobase-
Plett, T. S.; Cai, W.; Le Thai, M.; Vlassiouk, I. V.; Penner, R. M.; Siwy, Z. S., Solid-State Ionic
Nam, S.-W.; Rooks, M. J.; Kim, K.-B.; Rossnagel, S. M., Ionic Field Effect Transistors with Sub-
Joshi, P.; Smolyanitsky, A.; Petrossian, L.; Goryll, M.; Saraniti, M.; Thornton, T. J., Field Effect
20
Horn, H. W.; Swope, W. C.; Pitera, J. W.; Madura, J. D.; Dick, T. J.; Hura, G. L.; Head-Gordon,
Abascal, J. L. F.; Sanz, E.; García Fernández, R.; Vega, C., A Potential Model for the Study of
37.
T., Development of an Improved Four-Site Water Model for Biomolecular Simulations: Tip4p-Ew. J.
Chem. Phys. 2004, 120, 9665-9678.
38.
Ices and Amorphous Water: Tip4p/Ice. J. Chem. Phys. 2005, 122, 234511.
39.
Gromacs: Fast, Flexible, and Free. J. Comput. Chem. 2005, 26, 1701-1718.
40.
Efficient, Load-Balanced, and Scalable Molecular Simulation. J. Chem. Theory Comput. 2008, 4, 435-
447.
Van Der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; Berendsen, H. J. C.,
Hess, B.; Kutzner, C.; van der Spoel, D.; Lindahl, E., Gromacs 4: Algorithms for Highly
21
Supporting information
Aqueous Ion Trapping and Transport in Graphene-embedded 18-crown-6 Ether Pores
Alex Smolyanitsky, Eugene Paulechka, Kenneth Kroenlein
S1. Ionic fluxes
Figure S1. Typical cumulative fluxes obtained for different transmembrane voltages (0.5 M KCl and
NaCl). Zero Cl- fluxes were obtained in all simulations.
S2. Permeation and membrane potential in a 20-pore membrane
Figure S2. Simulated current and membrane potential as functions of the voltage across a 20-pore
membrane (a fully occupied state is shown in the inset) for 0.5 M KCl. Each point was obtained from a
200 ns long room-temperature MD simulation, as described in the main text.
S1
S3. Transistor sensitivity tuning
We rewrite Eq. (1) from the main text and replace the salt bulk concentration c with 𝑐 exp (−
𝑞𝑉𝑔
𝑘𝑏𝑇
):
𝑞𝑘𝑑𝑐 exp(−
𝑞𝑉𝑔
𝑘𝑏𝑇
)sinh (
𝐼 =
𝐾𝑑 cosh(
𝑞𝑉
2𝑘𝑏𝑇
)+𝑐 exp(−
)
𝑞𝑉
2𝑘𝑏𝑇
𝑞𝑉𝑔
𝑘𝑏𝑇
)
(S1)
The differential per-pore transconductance is
𝑔(𝑐) =
𝑑𝐼
𝑑𝑉𝑔
= −
𝑞2𝑘𝑑𝐾𝑑
𝑘𝑏𝑇
𝑓1(𝑉) 𝑐 (𝐾𝑑𝑓2(𝑉) + 𝑐 exp (−
−2
))
, (S2)
𝑞𝑉𝑔
𝑘𝑏𝑇
where 𝑓1(𝑉) = sinh (
𝑞𝑉𝑔
𝑘𝑏𝑇
at 𝐾𝑑𝑓2 = 𝑐 exp (−
), yielding
𝑞𝑉
2𝑘𝑏𝑇
) cosh (
𝑞𝑉
2𝑘𝑏𝑇
) and 𝑓2(𝑉) = cosh (
𝑞𝑉
2𝑘𝑏𝑇
). The absolute value of g is maximized
𝑐0 = 𝐾𝑑 exp (
𝑞𝑉𝑔
𝑘𝑏𝑇
) cosh (
𝑞𝑉
2𝑘𝑏𝑇
).
(S3)
With 𝑘𝑑 and 𝐾𝑑 used in the main text, we can evaluate Eqs. (S2) and (S3) for KCl. At an operating point
of V = 200 mV and the current modulated around 𝑉𝑔 = 0, we obtain 𝑐0= 0.26 M and 𝑔(𝑐0) ≈ -0.44 nS. At
a higher operating point of V = 300 mV, 𝑐0= 1.78 M and a considerably larger optimal transconductance
is obtained: 𝑔(𝑐0) ≈ -3.0 nS. Presented calculations are only accurate at the order-of-magnitude level, but
clearly demonstrate that transistor sensitivity exhibits resonant properties with respect to salt
concentration, and is realistically tunable for a selected operating transmembrane voltage V. KCl
concentration yielding maximal absolute value of transconductance as a function of V is shown in Fig. S2.
The inset demonstrates resonant behavior of g with respect to c at V = 200 mV.
Figure S3. Optimal KCl concentration c0 as a function of selected transmembrane voltage operating point
around Vg = 0. The inset shows absolute value of transconductance g as a function of bulk salt
concentration for V = 200 mV with a maximum at c0 = 0.26 M.
S2
S4. Structure and atomic charges in graphene-embedded 18-crown-6 pores
Figure S4. Atomic charges, according to the OPLS-AA forcefield; the circular charge distribution was
confirmed by quantum-mechanical calculations. Each resulting dipole is outlined by a dashed triangle and
the charges of each atomic species are equal between dipoles.
A direct replacement of sp2 carbons with oxygens in an infinite pristine graphene sheet will necessarily
introduce defects like uncompensated charge, or require extra hydrogenation to obtain system neutrality.
However, in a finite sheet interfaced with a conducting substrate, however large, this situation can be
avoided (see, for example, Fig. S5 with hydrogen-passivated edges).
A realistic graphene sample is expected to have naturally occurring defects, including C-H, C-O, or C=O
bonds, which, along with the presence of mobile charge in graphene, can compensate for the defects from
introducing the pore structure, thus maintaining electrical neutrality of the overall system. The correctness
of this assumption is supported by Figure 2 of Ref. 6 in the main text, where most of the experimentally
obtained oxygen-containing rings are shown to contain six atoms. As a result, a simplified model
presented here should be able to describe the main features of the crown-porous membrane in terms of its
interactions with aqueous ions. In addition, such a simplification is not expected to produce a noticeable
effect on the parameters used in the MD simulations. At the same time, more rigorous calculations (for
example, using density functional theory) require explicit consideration of the defects. All charges shown
in Fig. S4, along with the planarity of the pore region, were confirmed by independent quantum-
mechanical calculations (CHELPG scheme1 at the HF/6-31+G(d) theory level, using Gaussian 09
software2) performed on a structure with defects similar to those shown in Fig. S5.
Figure S5. Bond distribution in a finite graphene sample with hydrogen-terminated edges and an
embedded 18-crown-6 pore.
S3
References
Breneman, C. M.; Wiberg, K. B., Determining Atom-Centered Monopoles from Molecular
1.
Electrostatic Potentials. The Need for High Sampling Density in Formamide Conformational Analysis. J.
Comput. Chem. 1990, 11, 361-373.
2.
Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian,
H. P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenberg, J. L.; Hada, M.; Ehara, M., et al. Gaussian 09,
Gaussian, Inc.: Wallingford, CT, USA, 2009.
Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.;
S4
|
1909.13236 | 1 | 1909 | 2019-09-29T08:32:00 | A general Lewis acidic etching route for preparing MXenes with enhanced electrochemical performance in non-aqueous electrolyte | [
"physics.app-ph",
"cond-mat.mtrl-sci"
] | Two-dimensional carbides and nitrides of transition metals, known as MXenes, are a fast-growing family of 2D materials that draw attention as energy storage materials. So far, MXenes are mainly prepared from Al-containing MAX phases (where A = Al) by Al dissolution in F-containing solution, but most other MAX phases have not been explored. Here, a redox-controlled A-site-etching of MAX phases in Lewis acidic melts is proposed and validated by the synthesis of various MXenes from unconventional MAX phase precursors with A elements Si, Zn, and Ga. A negative electrode of Ti3C2 MXene material obtained through this molten salt synthesis method delivers a Li+ storage capacity up to 738 C g-1 (205 mAh g-1) with high-rate performance and pseudocapacitive-like electrochemical signature in 1M LiPF6 carbonate-based electrolyte. MXene prepared from this molten salt synthesis route offer opportunities as high-rate negative electrode material for electrochemical energy storage applications. | physics.app-ph | physics | A general Lewis acidic etching route for preparing MXenes with
enhanced electrochemical performance in non-aqueous electrolyte
Youbing Li 1,2†, Hui Shao 3,4†, Zifeng Lin 5*, Jun Lu 6, Per O. Å. Persson 6, Per Eklund
6, Lars Hultman 6, Mian Li 1, Ke Chen 1, Xian-Hu Zha1, Shiyu Du 1, Patrick Rozier 3,4,
Zhifang Chai 1, Encarnacion Raymundo-Piñero 4,7, Pierre-Louis Taberna 3,4, Patrice
Simon 3,4*, Qing Huang 1*
1 Engineering Laboratory of Advanced Energy Materials, Ningbo Institute of Materials
Technology and Engineering, Chinese Academy of Sciences, Ningbo, Zhejiang
315201, China
2 University of Chinese Academy of Sciences, 19 A Yuquan Rd, Shijingshan District,
Beijing 100049, China
3 CIRIMAT, Université de Toulouse, CNRS, France
4 Réseau sur le Stockage Electrochimique de l'Energie (RS2E), FR CNRS n°3459
5 College of Materials Science and Engineering, Sichuan University, Chengdu, 610065,
China
6 Thin Film Physics Division, Department of Physics, Chemistry, and Biology (IFM),
Linköping University, SE-581 83 Linköping, Sweden
7 CNRS, CEMHTI UPR3079, Univ. Orléans, F-4071 Orléans, France
*Correspondence to:
Prof. Zifeng Lin, E-mail: [email protected]
Prof. Patrice Simon, E-mail: [email protected]
Prof. Qing Huang, E-mail: [email protected]
†These authors contributed equally to this work.
One Sentence Summary: Lewis acidic molten salts etching is an effective and
promising route for producing MXenes with superior electrochemical performance in
non-aqueous electrolyte.
Abstract: Two-dimensional carbides and nitrides of transition metals, known as
MXenes, are a fast-growing family of 2D materials that draw attention as energy
storage materials. So far, MXenes are mainly prepared from Al-containing MAX
phases (where A = Al) by Al dissolution in F-containing solution, but most other MAX
phases have not been explored. Here, a redox-controlled A-site-etching of MAX phases
1
in Lewis acidic melts is proposed and validated by the synthesis of various MXenes
from unconventional MAX phase precursors with A elements Si, Zn, and Ga. A
negative electrode of Ti3C2 MXene material obtained through this molten salt synthesis
method delivers a Li+ storage capacity up to 738 C g-1 (205 mAh g-1) with high-rate
performance and pseudocapacitive-like electrochemical signature in 1M LiPF6
carbonate-based electrolyte. MXene prepared from this molten salt synthesis route offer
opportunities as high-rate negative electrode material for electrochemical energy
storage applications.
Main Text: Two-dimensional (2D) transition metal carbides or carbonitrides
(MXenes) are one of the latest additions to the family of 2D-materials. MXenes are
prepared by selective etching of the A layer elements in MAX phase precursors, where
M represents an early transition metal element (Ti, V, Nb, etc.), A is an element mainly
from the group 13-16 (Al, Si, etc.) and X is carbon and/or nitrogen (1). Their general
formula can be written as Mn+1XnTx (n=1-3), where Tx stands for the surface
terminations, generally considered to be -F, -O, and -OH. Thanks to their unique 2D
layered structure, hydrophilic surfaces and metallic conductivity (>6000 S cm-1),
MXenes show promise in a broad range of applications, especially in electrochemical
energy storage (2, 3).
Following the first report of Ti3C2 MXene synthesis in 2011, MXenes are mainly
prepared by selective etching of the A-layer of in MAX phases by aqueous solutions
containing fluoride ions such as aqueous hydrofluoric acid (HF) (1), mixtures of lithium
fluoride and hydrochloric acid (LiF+HCl) (4) or ammonium bifluoride ((NH4)HF2) (5).
To date, the high reactivity of Al with fluoride-based aqueous solutions has limited
synthesized MXenes to preferentially Al-containing MAX phase precursors. Although
2
Alhabeb et al. reported the synthesis of Ti3C2 MXene through oxidant-assisted etching
of Si from Ti3SiC2 MAX phase (5), the etching mechanism was still based on hazardous
HF solution. Thus, MXene synthesis is challenged 1) to find nonhazardous synthesis
routes for preparing MXene and 2) to enable a broader range of MAX-phase precursors.
Recently, Huang and et al. reported that Ti3C2Cl2 MXene can be prepared by etching
Ti3ZnC2 MAX phase in ZnCl2 Lewis acidic molten salt via a replacement reaction
mechanism (6). In the present paper, we generalize this synthesis route to a wide
chemical range of A-site elements featuring besides Zn also Al, Si, Ga from various
MAX phase precursors. This is accomplished by selective etching in Lewis acid molten
salts via a redox substitution reaction. With such processing we also show that, for
instance, MXene could be obtained from MAX phases with A = Ga. The etching
process is illustrated here using Ti3C2 prepared from Ti3SiC2 immersion in CuCl2
molten salt. The obtained MXene exhibits enhanced electrochemical performance with
high Li+ storage capacity combined with high-rate performance in non-aqueous
electrolyte, which makes these materials promising electrode materials for high-rate
battery and hybrid devices such as Li-ion capacitor applications (7, 8). This method
allows producing new 2D materials that are difficult or even impossible to be prepared
by using previously reported synthesis methods like HF etching. As a result, it expands
further the range of MAX phase precursors that can be used and offer important
opportunities for tuning the surface chemistry and the properties of MXenes.
3
Fig. 1. Schematic diagram of Ti3C2Tx MXene preparation by immersing Ti3SiC2 MAX
phase in CuCl2 Lewis molten salt at 750°C.
Fig. 1 shows a sketch of the Ti3C2 MXene synthesis from the reaction between
Ti3SiC2 and CuCl2 at 750°C; the reactions are listed below:
Ti3SiC2 + 2CuCl2 = Ti3C2 + SiCl4(g)+ 2Cu
Ti3C2 + CuCl2 =Ti3C2Cl2 + Cu
(1)
(2)
Ti3SiC2 MAX precursor is immersed at 750°C in molten CuCl2 (Tmelting=498oC). The
exposed Si atoms weakly bonded to Ti in the Ti3C2 sublayers are oxidized into Si4+
cation by Lewis acid Cu2+, resulting in the formation of the volatile SiCl4 phase
(Tboiling=57.6°C) and concomitant reduction of Cu2+ into Cu metal (equation 1). Similar
to what has been recently reported (6), extra Cu2+ partially reacts with the exposed Ti
atoms from Ti3C2 to form metallic copper, while the charge compensation is ensured
by Cl- anions to form Ti3C2Cl2 (equation 2). The formation mechanism of Ti3C2Cl2
from Ti3SiC2 is analog to that of chemical etching of Ti3AlC2 in HF solution (1): Cu2+
4
and Cl- act as H+ and F-, respectively. The as-prepared powder of Ti3C2Cl2 and Cu
metal, see Fig. S1, were further immersed in ammonium persulfate (APS) solution to
remove Cu particles from the Ti3C2Cl2 MXene surface, which also results in the
addition of O-based surface groups (Fig. S2). This final material prepared from this
molten salt route will be noted as MS-Ti3C2Tx MXene, where Tx stands for O and Cl
surface groups.
Fig. 2. Morphological and structural characterizations of MS-Ti3C2Tx MXene. (A)
XRD patterns of pristine Ti3SiC2 before (black line) and after (red line) reaction with
CuCl2, and final MS-Ti3C2Tx MXene obtained after washing in 1 M (NH4)2S2O8
solution (purple line). (B) SEM and (C) Cross-sectional STEM images showing the
nanolaminate nature of the material (scale bar in the atomically resolved image inset in
(C) is 1 nm), and (D) XPS spectra of the Ti 2p energy level from the MS-Ti3C2Tx
MXene sample.
X-ray diffraction (XRD) patterns of the pristine Ti3SiC2 before (black), and after
reaction with CuCl2 at 750°C for 24h (noted as Ti3SiC2-CuCl2, red) and final product
5
after APS washing (MS-Ti3C2Tx, purple) are shown in Fig. 2A. Compared to pristine
Ti3SiC2, most of the diffraction peaks disappear in the final product, leaving (00l) peaks
as well as several broad and low-intensity peaks in the 2 range from 5° to 75°; these
features indicate the successful reduction of Ti3SiC2 into layered Ti3C2 (MXene) (9).
Additionally, the shift of Ti3C2 (00l) diffraction peaks from 10.13° to 7.94° two theta
degree indicate an expansion of the interlayer distance from 8.8 Å to 10.9 Å. The sharp
and intense peaks located at 2 ≈ 43.29°, 50.43°, and 74.13° can be indexed as metallic
Cu (Fig. 2A, red plot), which confirms the proposed etching mechanism in Lewis acid
melt (equation 1). The XRD pattern of the final product (Fig. 2A, purple plot) exhibits
only the (00l) MXene peaks, confirming the removal of the Cu. SEM image of the final
MS-Ti3C2Tx sample is shown in Fig. 2B. After etching in molten salt, the Ti3SiC2
particle (Fig. S1A) turns into an accordion-like microstructure (Fig. S1B), similar to
previously reported for MXenes obtained by HF etching (1). The spherical particles
observed on the Ti3C2 before APS treatment (Fig. S1B) are assumed to be metallic Cu
produced during the etching process from equations (1, 2) (Fig. S1C), which become
removed by immersion in APS solution (Fig. S2).
The lamellar microstructure of the MS-Ti3C2Tx MXene is clearly visible in STEM
images, as shown in Fig. 2C. The SiCl4 gas molecules formed in situ during the etching
reaction (equation 1) is believed to act as an effective expansive agent to delaminate
the MXene, similar to the preparation of expanded graphite through the decomposition
of intercalated inorganic acids (10).
MS-Ti3C2Tx MXene sample surface was further characterized by XPS analysis. Fig.
S3A shows an overview XPS spectrum for the Ti3SiC2 precursor (black) and MS-
Ti3C2Tx MXene (red), where the signals of Si 2p, C 1s, Ti 2p, and O 1s are observed at
102, 285, 459, and 532 eV, respectively (11). The disappearance of the Si signal
6
confirms the effectiveness of Si removal by Lewis acid etching reaction (Fig. S3B).
Similarly, no significant amounts of Cu or S element were detected (Fig. S4A and S4B).
The deconvolution of the Ti 2p spectra (Fig. 2D) in the energy range between 454 and
460 eV was achieved following previous works (6, 12) and the details are given in
Table S1. The Ti 2p spectra show the existence of Ti-O and Ti-Cl chemical bonds, most
likely from O and Cl surface groups associated with partial surface oxidation. The
observed Ti-C bonds come from the core [TiC6] octahedral building blocks of the Ti3C2
MXene. The fitting of the O 1s (Fig. S4C) and C 1s (Fig. S4D) spectra show O-
terminated surface functional groups on MS-Ti3C2Tx sample, including the possible
hydroxides. The XPS signal of the Cl 2p energy level confirms the presence of Ti-Cl
bonds (Fig. S4E). The Cl groups are expected from equation (2), while O surface
functional groups are formed during the oxidation treatment in APS solution and
subsequent washing process (13). EDS analysis (Table S2) revealed an O-termination-
group content of about 20 at.% together with 16.5 at.% of Cl-termination-group content
in
the MS-Ti3C2Tx MXene,
resulting
in an approximate composition of
Ti3C1.3Cl1.15O1.39.
Temperature-programed desorption, coupled with mass spectroscopy measurements
(TPD-MS) have been achieved on MS-Ti3C2Tx MXene samples and MXene prepared
from conventional etching treatment in HF, noted as HF-Ti3C2Tx (Fig. S5 and Table
S3). H2O release observed below 400°C for both samples corresponding to surface
adsorbed and intercalated water coming from the washing with water after synthesis
(14). Differently from HF-Ti3C2Tx, MS-Ti3C2Tx, MXene shows substantial CO2 release
below 600°C, which could be ascribed to the partial carbon oxidation from APS
oxidizing treatment. Also noteworthy is the absence of any -OH surface groups release
for MS-Ti3C2Tx MXene, decreasing the hydrophilicity of the surface. Cl group is stable
7
on Ti3C2 at 750°C (15), but a trace of released Cl is still detected as well as SO2 below
600°C, the latter coming from APS treatment. Also interesting is the quantification of
the total amount of oxygen from CO and CO2 gases (8.2 wt.%, see Table S3), which is
close to 9.5 wt.% estimated from EDS analysis.
Fig. 3. Generalization of the Lewis acid etching route to a large family of MAX phase.
(A) Gibbs Free Energy mapping (700°C) guiding the selection of Lewis acid chloride
salts according to electrochemical redox potentials of A site elements in MAX phases
(X axis) and molten salt cations (Y axis) in chloride melts. Stars mark corresponding
MXenes that are demonstrated in the current study. SEM images reveal the typical
accordion morphology of MXenes from different MAX phases etched by varied Lewis
acid chlorides, such as Ti2AlC by CuCl2 (B), Ti3ZnC2 by CuCl2 (C), Ti3AlCN by CuCl2
(D), Ti3AlC2 by NiCl2 (E), Ti3ZnC2 by FeCl2 (F), and Ta2AlC by AgCl (G). Scale bars
are 2 μm.
The capability of Lewis acid to withdraw electrons from A element in the MAX
phase can be well reflected from their respective electrochemical redox potential in
halide melts. For instance, Si4+/Si couple has a redox potential as low as -1.38 V vs.
8
Cl2/Cl- at 750°C. As a result, CuCl2 molten salt (redox potential of -0.43 V vs. Cl2/Cl-)
can easily oxidize Si into Si4+ (etching/exfoliation of MAX phase into MXene). The
present Lewis acid etching process can be then generalized to prepare a broad family
of MXene materials. Fig. 3A shows a Gibbs Free Energy mapping prepared from
thermodynamics data (see equation 3 and Fig. S6) to guide the selection of effective
Lewis acids for MAX phases having different A elements (Fig. 3A). In these
calculations, the etching is independent to the composition of MX layer and n value of
Mn+1AXn. The color of each spot/star indicates the value of Gibbs free energy of the
reaction between selected A element in MAX phase and Lewis acid chloride melt at
700°C (Equation 3).
A + y/x BClx = ACly + y/x B
(3)
From these thermodynamic calculations, etching of A element from MAX can be
achieved by using a Lewis acid with higher redox potential. Based on this map, a series
of MAX phases - specifically Ti2AlC, Ti3AlC2, Ti3AlCN, Nb2AlC, Ta2AlC, Ti2ZnC,
and Ti3ZnC2 - was successfully exfoliated into corresponding MXenes (Ti3C2Tx,
Ti3CNTx, Nb2CTx, Ta2CTx, Ti2CTx, Ti3C2Tx) using various chlorides molten salts
(CdCl2, FeCl2, CoCl2, CuCl2, AgCl, NiCl2), as marked in star shape (Fig. 3A). SEM
images in Fig. 3B-3G show the lamellar microstructures of obtained MXenes. The
successful preparation of Ta2CTx and Ti3C2Tx MXenes from Ta2AlC and Ti3SiC2,
which were theoretically predicted hard to be exfoliated, evidences the effectiveness of
the Lewis acid molten salts route (16). Additional information about as-prepared
MXenes can be found in Fig. S7-S14. Taking account of the diversity and green
chemistry of Lewis acid in inorganic salts, there is unexplored parameter space to
optimize such etching methodology. At the same time, it broadens the selection scope
of MAX phase family for MXene fabrication and offers opportunities for tuning the
9
surface chemistry of MXene materials by using various molten salts based on other
anions (such as Br-, I-, SO4
2-, and NO3
-).
Layered MS-Ti3C2Tx MXene powders here derived from Ti3SiC2 (Fig. 2B) were
further used to prepare electrodes by mixing with carbon conducting additive and
binder (see the experimental section for details). Fig. 4A shows the cyclic voltammetry
(CV) profiles of the MS-Ti3C2Tx MXene electrode in 1M LiPF6/EC:DMC electrolyte
recorded at 0.5 mV s-1 with different negative cut-off potentials. The electrochemical
signature is remarkable as it differs from what is previously reported for MXene made
in non-aqueous electrolytes (17-20). Indeed, CV does not show redox peaks associated
with Li-ion intercalation, such as reported in the literature (21, 22). Instead, the charge
storage mechanism is achieved by a constant current versus applied potential, similarly
to what is observed in a pseudocapacitive material, with an almost constant current
during reduction and oxidation process in a potential range between 2.2 V vs. Li+/Li
and 0.2 V vs. Li+/Li. The discharge capacity of the MS-Ti3C2Tx MXene powder in this
non-aqueous Li-ion battery electrolyte reaches 738 C g-1 (205 mAh g-1) at 0.5 mV s-1
within the full potential window of 2.8 V, which translates into 323 F g-1 within 2 V
(see Fig. S15). These are the highest capacitance values reported for Ti3C2 MXene in
non-aqueous electrolytes, to the best of our knowledge (3, 17, 23, 24). Those
remarkable performances make MXene materials now suitable to be used as negative
electrodes in non-aqueous energy storage devices. Also important, and differently from
previous works where electrodes had to be prepared from filtration of delaminated
MXene suspensions to achieve high electrochemical performance (25), raw, non-
delaminated MXene powders (Fig. 2B) were used here to prepare the electrode films.
This broadens the range of application of the materials to prepare electrodes for energy-
storage devices.
10
Fig. 4. Electrochemical characterizations of MS-Ti3C2Tx MXene electrode in 1M LiPF6
in EC:DMC (1:1) electrolyte. (A) Cyclic voltammetry profiles (CVs) at a 0.5 mV s-1
potential scan rate with various cut-off negative potentials; CVs exhibits a mirror-like
shape with no redox peak during Li intercalation/deintercalation redox reaction; (B) In
situ XRD maps of the (002) peak during anodic and cathodic scans for 3 different
cycles; the peak position shift is less than 0.25Å during cycling; (C) Change of the
MXene electrode capacity versus the discharge time during CVs recorded at various
potential scan rates from anodic scans. The active material weight loading is 1.4 mg
cm-2; (D) Galvanostatic charge/discharge curves at current densities from 0.5 to 3 A g-
1.
The charge storage mechanism was investigated using in situ X-ray diffraction
technique during cyclic voltammetry experiments at 0.5 mV s-1. Fig. 4B shows the
change of the (002) peak position during anodic and cathodic scans for three different
cycles. The initial d-spacing was found to be 11.02 Å, and the peak position was found
11
to be roughly constant during the polarization with a maximum change of 0.25 Å. The
small value of the d-spacing indicates that MXene layers are separated by about 3 Å:
this supports the intercalation of de-solvated Li+ ions between the MXene layers, such
as recently reported (17), blocking the co-intercalation of solvent molecules and
resulting in improved electrochemical performance. During the cathodic scan (Fig. 4A),
Li+ ions are intercalated between the MXene layers; this is assumed to be associated
with the change in the oxidation state of Ti, such as observed in lithium-ion battery
during Li+ intercalation (26, 27). Li+ de-insertion from the MXene structure occurs
during the anodic potential scan, with a remarkable mirror-like CV shape. During the
first cycle upon reduction, an irreversible capacity is observed (Fig. S16A), as a result
of the formation of the solid electrolyte interphase layer (SEI) (28). As a result of these
high power performance, the electrochemical impedance spectroscopy plots recorded
at various bias potentials (Fig. S18A) show a charge-transfer resistance of about 25
Ω·cm² followed by a restricted-diffusion behavior with a fast increased of the imaginary
part at low frequency (29).
Fig. 4C shows the change of the Ti3C2 MXene capacity with discharge time calculated
from CVs achieved at various potential scan rates (Fig. S16B and Table S7). The
capacity reaches 738 C g-1 (205 mAh g-1) for a discharge time of 1.5 h (C/1.5 rate). This
value corresponds to a minimum of 1.28 F exchanged per mole of Ti3C2, which is about
0.42 electron transferred per Ti atom, much higher than previously reported values (17,
27). The electrode still delivers 142 mAh g-1 for 280 s discharge time (13 C rate) and
75 mAh g-1 for a time less than 30 s (128 C rate). Together with the galvanostatic plots
achieved at various current densities (Fig. S17A), these results highlight the high-power
performance of the present Ti3C2 MXene material as electrode during Li+ ion
intercalation reaction, occurring at lower potential vs Li+/Li compared to previously
12
reported pseudocapacitive materials (30, 31). Interestingly, an increase of the electrode
weight loading (4 mg cm-2) does not substantially affect the power capability (Fig.
S16C and D). Galvanostatic charge/discharge measurements (Fig. 4D) confirm the
unique electrochemical signature of the electrode in non-aqueous electrolyte with a
slopping voltage profile within a potential range of 0.2-2.2 V vs. Li+/Li, as expected
from the CVs shown in Fig. 4A. Last, but not least, cycle stability was impressive with
90% capacity retention after 2,400 cycles (Fig. S17B). Similar remarkable
electrochemical signature and performance were obtained for other MS-MXene studied
here, such as can be seen more specifically from the CVs and power performance of a
Ti3C2Tx electrode prepared from Ti3AlC2 MAX phase (see Fig. S19).
The combination of mirror-like electrochemical signature in non-aqueous Li-ion
containing electrolyte, together with high capacity, high-rate discharge and charge
performance (less than one minute) and the low operating potential range (0.2 -- 2.2 V
vs. Li+/Li) makes this Ti3C2 MXene prepared from molten salt derivation route relevant
as negative electrode in electrochemical energy storage devices (batteries and Li-ion
capacitors). As a result, the general Lewis acidic etching route proposed here expands
the range of MAX phase precursors that can be used to prepare new MXenes, and offer
unprecedented opportunities for tailoring the surface chemistry and consequently the
properties of MXene materials.
References and Notes
1.
2.
3.
4.
M. Naguib et al., Two-Dimensional Nanocrystals Produced by Exfoliation of
Ti3AlC2. Advanced Materials 23, 4248-4253 (2011).
M. R. Lukatskaya et al., Cation intercalation and high volumetric capacitance
of two-dimensional titanium carbide. Science 341, 1502-1505 (2013).
B. Anasori, M. R. Lukatskaya, Y. Gogotsi, 2D metal carbides and nitrides
(MXenes) for energy storage. Nat Rev Mater 2, Article number: 16098 (2017).
M. Ghidiu, M. R. Lukatskaya, M.-Q. Zhao, Y. Gogotsi, M. W. Barsoum,
Conductive two-dimensional titanium carbide 'clay' with high volumetric
capacitance. Nature 516, 78-81 (2014).
13
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
A. Feng et al., Fabrication and thermal stability of NH4HF2-etched Ti3C2
MXene. Ceramics International 43, 6322-6328 (2017).
M. Li et al., Element Replacement Approach by Reaction with Lewis Acidic
Molten Salts to Synthesize Nanolaminated MAX Phases and MXenes. Journal
of the American Chemical Society 141, 4730-4737 (2019).
K. Naoi et al., Ultrafast charge-discharge characteristics of a nanosized core-
shell structured LiFePO4 material for hybrid supercapacitor applications.
Energy Environ. Sci. 9, 2143-2151 (2016).
M. R. Lukatskaya, B. Dunn, Y. Gogotsi, Multidimensional materials and device
architectures for future hybrid energy storage. Nature communications 7, 12647
(2016).
M. Alhabeb et al., Selective Etching of Silicon from Ti3SiC2 (MAX) To Obtain
2D Titanium Carbide (MXene). Angewandte Chemie 130, 5542-5546 (2018).
S. Yang et al., Ultrafast Delamination of Graphite into High-Quality Graphene
Using Alternating Currents. Angew Chem Int Ed Engl 56, 6669-6675 (2017).
E. Kisi, J. Crossley, S. Myhra, M. Barsoum, Structure and crystal chemistry of
Ti3SiC2. Journal of Physics and Chemistry of Solids 59, 1437-1443 (1998).
J. Halim et al., X-ray photoelectron spectroscopy of select multi-layered
transition metal carbides (MXenes). Applied Surface Science 362, 406-417
(2016).
O. Çakır, Review of Etchants for Copper and its Alloys in Wet Etching
Processes. Key Engineering Materials 364-366, 460-465 (2008).
N. Shpigel et al., Direct Assessment of Nanoconfined Water in 2D Ti3C2
Electrode Interspaces by a Surface Acoustic Technique. Journal of the
American Chemical Society 140, 8910-8917 (2018).
J. Lu et al., Tin+1Cn MXenes with fully saturated and thermally stable Cl
terminations. Nanoscale Advances, (2019).
16. M. Khazaei et al., Insights into exfoliation possibility of MAX phases to
MXenes. Physical Chemistry Chemical Physics 20, 8579-8592 (2018).
X. Wang et al., Influences from solvents on charge storage in titanium carbide
MXenes. Nature Energy 4, 241-248 (2019).
X. Wang et al., Pseudocapacitance of MXene nanosheets for high-power
sodium-ion hybrid capacitors. Nature Communications 6, 6544 (2015).
J. Come et al., A Non-Aqueous Asymmetric Cell with a Ti2C-Based Two-
Dimensional Negative Electrode. Journal of The Electrochemical Society 159,
A1368-A1373 (2012).
J. Luo et al., Pillared Structure Design of MXene with Ultralarge Interlayer
Spacing for High-Performance Lithium-Ion Capacitors. ACS Nano 11, 2459-
2469 (2017).
C. E. Ren et al., Porous Two-Dimensional Transition Metal Carbide (MXene)
Flakes for High-Performance Li-Ion Storage. ChemElectroChem 3, 689-693
(2016).
R. Cheng et al., Understanding the Lithium Storage Mechanism of Ti3C2Tx
MXene. The Journal of Physical Chemistry C 123, 1099-1109 (2018).
J. Pang et al., Applications of 2D MXenes in energy conversion and storage
systems. Chemical Society Reviews 48, 72-133 (2019).
D. Xiong, X. Li, Z. Bai, S. Lu, Recent Advances in Layered Ti3C2Tx MXene for
Electrochemical Energy Storage. Small 14, Article number: 1703419 (2018).
14
17.
18.
19.
20.
21.
22.
23.
24.
25. M. R. Lukatskaya et al., Ultra-high-rate pseudocapacitive energy storage in
two-dimensional transition metal carbides. Nature Energy 2, Article number:
17105 (2017).
Y. Xie et al., Role of Surface Structure on Li-Ion Energy Storage Capacity of
Two-Dimensional Transition-Metal Carbides. Journal of the American
Chemical Society 136, 6385-6394 (2014).
26.
28.
29.
27. M. R. Lukatskaya et al., Probing the Mechanism of High Capacitance in 2D
Titanium Carbide Using In Situ X-Ray Absorption Spectroscopy. Adv Energy
Mater 5, Aricle number: 1500589 (2015).
C. R. Birkl, M. R. Roberts, E. McTurk, P. G. Bruce, D. A. Howey, Degradation
diagnostics for lithium ion cells. Journal of Power Sources 341, 373-386
(2017).
J. P. Diard, B. L. Gorrec, C. Montella, Linear diffusion impedance. General
expression and applications. Journal of Electroanalytical Chemistry 471, 126-
131 (1999).
V. Augustyn et al., High-rate electrochemical energy storage through Li+
intercalation pseudocapacitance. Nat Mater 12, 518-522 (2013).
H.-S. Kim et al., Oxygen vacancies enhance pseudocapacitive charge storage
properties of MoO3-x. Nat Mater 16, 454-460 (2017).
31.
30.
Acknowledgments: This study was supported financially by the National Natural
Science Foundation of China (Grant No. 21671195, 91426304, and 51902319), and
China Postdoctoral Science Foundation (Grant No. 2018M642498). HS was supported
by a grant from the China Scolarship Council. PS, PLT and HS thanks the Agence
Nationale de la Recherche (Labex STORE-EX) for financial support. ZL is supported
by the Fundamental Research Funds for the Central Universities (YJ201886). The
authors acknowledge the Swedish Government Strategic Research Area in Materials
Science on Functional Materials at Linköping University (Faculty Grant SFO‐Mat‐LiU
No. 2009 00971). The Knut and Alice Wallenberg Foundation is acknowledged for
support of the electron microscopy laboratory in Linköping, a Fellowship grant (P.E),
a Scholar Grant (L. H., 2016-0358).
Supplementary Materials:
Materials and Methods
Supplementary Text
Figs. S1 to S19
Tables S1 to S7
15
Supplementary Materials for
A general Lewis acidic etching route for preparing MXenes with
enhanced electrochemical performance in non-aqueous electrolyte
Youbing Li 1,2†, Hui Shao 3,4†, Zifeng Lin 5*, Jun Lu 6, Per O. Å. Persson 6, Per Eklund
6, Lars Hultman 6, Mian Li 1, Ke Chen 1, Xian-Hu Zha 1, Shiyu Du 1, Patrick Rozier 3,4,
Zhifang Chai 1, Encarnacion Raymundo-Piñero 4,7, Pierre-Louis Taberna 3,4, Patrice
Simon 3,4*, Qing Huang 1*
*Correspondence to:
Prof. Zifeng Lin, E-mail: [email protected]
Prof. Patrice Simon, E-mail: [email protected]
Prof. Qing Huang, E-mail: [email protected]
†These authors contributed equally to this work.
This PDF file includes:
Materials and Methods
Supplementary Text
Figs. S1 to S19
Tables S1 to S7
Full Reference List
16
Materials and Methods
Materials
High-purity Ti3AlC2, Ti3ZnC2, Ti3SiC2, Ti3AlCN, Ti2AlC, Ti2ZnC, Nb2AlC and
Ta2AlC MAX phases powders were synthesized as previously reported (1-6). Ti2GaC
MAX phase was synthesized in our laboratory via molten salt method. Zinc chloride
(anhydrous, ZnCl2, > 98 wt.% purity), cadmium chloride (anhydrous, CdCl2, > 98 wt.%
purity), ferrous chloride (anhydrous, FeCl2, > 98 wt.% purity), cobalt chloride
(anhydrous, CoCl2, > 98 wt.% purity), copper chloride (anhydrous, CuCl2, > 98 wt.%
purity), nickel chloride (anhydrous, NiCl2, > 98 wt.% purity), and silver chloride
(anhydrous, AgCl, > 98 wt.% purity), sodium chloride (anhydrous, NaCl, > 98 wt.%
purity), potassium chloride (anhydrous, KCl, > 98 wt.% purity), ammonium persulfate
((NH4)2S2O8, > 98 wt.% purity) and absolute ethanol (C2H6O, > 98 wt.%) were
purchased from Aladdin Chemical Reagent, China.
Preparation of MXenes from Lewis acid molten salt route
Various MAX phases and Lewis acid salts were used to prepare MXenes, as
summarized in Table S6. We here take Ti3SiC2 MAX phase and CuCl2 as an example:
1 g of Ti3SiC2 MAX phase powders and 2.1 g of CuCl2 powders were mixed (with a
stoichiometric molar ratio of 1:3) and grinded for 10 minutes. Then 0.6 g of NaCl and
0.76 g of KCl were added into the above mixtures and grinded for another 10 minutes.
Afterward, the mixture was placed into an alumina boat, and the boat was then put into
an alumina tube with argon-flow. The powder mixture was heated to 750°C with a
heating ramp of 4°C min-1, and hold for 24 h. Afterward, the obtained products were
washed with deionized water (DI H2O) to remove salts, and MXene/Cu mixed particles
were obtained. The mixtures of MXene/Cu were then washed by 1 M (NH4)2S2O8
solution (APS) to remove the residual Cu particles (7). The resulting solution was
further cleaned by deionized water (DI H2O) and alcohol for five times and filtered with
a microfiltration membrane (polyvinylidene fluoride, 0.45 μm). Finally, the MXene
powders (denoted as MS-Ti3C2Tx) were dried under vacuum at room temperature for
24 h.
17
Materials characterizations
The phase composition of the samples was analyzed by X-ray diffraction (D8 Advance,
Bruker AXS, Germany) with Cu K radiation. X-ray diffraction patterns were collected
with a step of 0.02° 2θ with a collection time of 1 s per step. The microstructures and
chemical compositions were analyzed by scanning electron microscopy (SEM,
QUANTA 250 FEG, FEI, USA) at 20 kV, with an energy-dispersive spectrometer
(EDS); EDS values were fitted by XPP. The chemical composition and bonding states
were measured by X-ray photoelectron spectroscopy (XPS) using a Kratos AXIS
ULTRA DLD instrument with a monochromic Al K X-ray source (hv = 1486.6 eV).
The power was 96 W, and the X-ray spot size set to 700 x 300 um. The pass energy of
the XPS analyzer was set at 20 eV. The pressure of the analysis chamber was kept
below 5 x 10-9 Torr. All spectra were calibrated using the binding energy (BE) of C 1s
(284.8eV) as a reference. The XPS atomic sensitivity factors involved in the atomic
concentration calculation were 0.278 (C 1s), 1.833 (Ca 2p), 2.001 (Ti 2p) and 0.78 (O
1s), respectively, according to Kratos Vision Processing software. Etch conditions were
defined by a beam energy of 4 kV, a current of 100 μA, and a raster size of 3 mm).
Transmission electron microscopy and high-resolution TEM images were obtained
using a Tecnai F20 (FEI, USA) electron microscope at an acceleration voltage of 200
kV. Structural and chemical analysis was carried out by high-resolution STEM high
angle annular dark field (HRSTEM-HAADF) imaging and STEM affiliated energy
dispersive X-ray spectroscopy (EDS) within Linköping's double Cs corrected FEI
Titan3 60-300 microscope operated at 300 kV, and STEM-EDX was recorded with the
embedded high sensitivity Super-X EDX detector. Temperature-programmed
desorption (TPD) was performed under inert atmosphere (Ar, 100ml min-1). The sample
(10-20 mg) was placed in a thermo-balance and heat treated up to 1300oC at a rate of
10oC min-1. The decomposition products (gas evolved) were monitored by on-line mass
spectrometry (Skimmer, Netzsch, Germany). Cl and S quantification could not be
achieved due to the absence of standards.
Electrochemical measurements
MS-Ti3C2Tx MXene self-standing electrodes were prepared by mixing the MXene
powder with 15 wt.% carbon black and 5 wt.% PTFE binder, and laminated many times
to obtain films with different thickness. The active material weight loading was
18
calculated by dividing the mass (mg) of MXene active material by the electrode area
(cm2). Metallic lithium foil was used as the counter and reference electrode, LP30 (1 M
LiPF6 in ethylene carbonate/dimethyl carbonate with 1:1 volume ratio) as the
electrolyte and 2 slides of 25-μm thick cellulose as the separator. Swagelok cells were
assembled in the Ar-filled glovebox with oxygen and water content less than 0.1 ppm.
All electrochemical tests were performed using a VMP3 potentiostat (Biologic). Cyclic
voltammetry and galvanostatic test were conducted in 2-electrode mode versus Li
electrode. Electrochemical impedance spectroscopy (EIS) was carried out with a
potential amplitude of 10 mV in the range from 10 mHz to 200 kHz.
In situ XRD was conducted on a Bruker D8 Advance diffractometer using Cu Kα
radiation source. Two-electrode Swagelok cell system (8), using MS-Ti3C2Tx MXene
film as the working electrode, beryllium window as the current collector, and Li metal
as the counter electrode, was used to perform the electrochemical test for the in-situ
XRD measurements. All XRD patterns were recorded during cyclic voltammetry test
at a potential scan rate of 0.5 mV s-1. The (002) peak located between 6° to 10° was
recorded to calculate the interlayer d-spacing (Fig. 4B).
In cyclic voltammetry, the capacity (C g-1) and average capacitance (F g-1) of a single
electrode are evaluated from the anodic scan using
Q = ∫ 𝑖 dt
m
(1)
C =
Q
V
(2)
Where i is the current changed by time t, m is the mass of active material, V is the
potential window.
In galvanostatic charge/discharge plots, the capacity (C g-1) is given by:
Q =
𝑖∆t
m
(3)
Where Δt is charging/discharging time.
19
Supplementary Text
Fig. S1. (A) SEM images of Ti3SiC2 MAX phase precursor before (A) and after (B)
reaction with CuCl2 at 750C. (C) EDS analysis of the MXene after reaction with CuCl2
at 750C, before immersion in (NH4)S2O8 (APS) solution. The presence of Cu metal
and Cl agrees with equation 1 and 2 presented in the manuscript. O element comes from
washing treatment in water. The successful removal of Si from Ti3SiC2 MAX phase is
evidenced by the significant weakening of the Si signal.
20
Fig. S2. SEM image of MS-Ti3C2Tx MXene after treatment by APS solution (0.1 mol/L)
to remove Cu particles and (B) corresponding EDS point analysis (B). After treatment
by APS solution at room temperature, MXene keeps its original layered structure. EDS
result shows the presence of Ti, Cl, O, C element of the MXene. The successful removal
of Cu after treatment by APS solution is evidenced by the presence of only a residual
weak signal. (C) Element mapping of MS-Ti3C2Tx MXene treatment by APS solution.
XPS analysis of Ti3SiC2 MAX phase and MS-Ti3C2Tx MXene
XPS analysis of the Ti3SiC2 MAX phase precursor (black) and MS-Ti3C2Tx MXene
(red) after reaction in CuCl2 at 750C and further immersion in APS solution are
presented in Fig. S3. Fig. S3A shows an overview XPS spectrum for the Ti3SiC2
precursor (black) and MS-Ti3C2Tx MXene (red) after APS treatment, respectively. For
Ti3SiC2, the signals of Si 2p, C 1s, Ti 2p, and O 1s were found at 101.2, 282.9, 458.6,
and 531.9 eV, respectively (9). The XPS of Si 2p in Ti3SiC2 (Fig. S3B, black) shows a
peak at 101.8 eV assigned to SiO2, which indicates the existence of oxide layer on Si
and a peak at 98.3 eV attributed to Ti-Si bonds (9). After etching by CuCl2 and further
immersion in APS solution, only the signals of Ti 2p, O 1s, Cl 2p, and C 1s were
detected. No Si signal could be detected on the final MS-Ti3C2Tx
MXene, which
confirms the Si removal.
21
Moreover, no significant amounts of Cu and S element were detected (Fig. S4A and
S4B). Fig. S4C shows the O 1s spectrum, where the peaks at 530.0 eV, 531.3 eV, and
533.3 eV are assigned to the Ti-O, Ti-C-Ox, and H2O (10, 11), respectively. The C 1s
signal in MS-Ti3C2Tx MXene (Fig. S4D) shows peaks at 281.2 eV, 284.5 eV, 286.2 eV,
and 288.5 eV assigned to the Ti-C, C-C, C-O and C=O bond (10, 12), respectively. The
peaks at 198.8 eV and 200.4 eV are associated with Cl-Ti (2p1/2) and Cl-Ti (2p3/2) bonds
(3, 13), which indicated the presence of Ti-Cl bonds in MS-Ti3C2Tx MXene (Fig. S4E).
The XPS signal of Ti 2p in MS-Ti3C2Tx MXene is shown in Fig. 2D. The peaks at 454.5
eV and 460.5 eV are assigned to the Ti-C (I) (2p3/2) and Ti-C (I) (2p1/2) bond (9, 10).
The peaks at 456.0 eV and 461.8 eV are assigned to the Ti-C (II) (2p3/2) and Ti-C (II)
(2p1/2) bond (9, 10). The peaks at 458.2 eV and 464.0 eV attributed to high-valency Ti
compound, are assigned to the Ti-Cl (2p3/2) and Ti-Cl (2p1/2) bonds, respectively (3, 13).
The peaks at 459.7 eV and 464.9 eV are associated with the Ti-O (2p3/2) and Ti-O (2p1/2)
(10, 11), respectively. The results are summarized in Table S1.
Fig. S3. XPS analysis of the Ti3SiC2 MAX phase precursor (black) and MS-Ti3C2Tx
MXene (red) after reaction in CuCl2 at 750C and further immersion in APS solution.
(A) The global view of the XPS spectra. (B) Spectra of Si 2p energy level. No Si signal
could be detected on the final MS-Ti3C2Tx
MXene product.
22
Fig. S4. XPS analysis of the MS-Ti3C2Tx MXene after reaction in CuCl2 at 750C and
further immersion in APS solution. Spectra of Cu 2p (A), S 2p (B), O 1s (C), C 1s (D)
and Cl 2p (E) energy level. No Cu or S signals could be detected on the final MS-
Ti3C2Tx MXene product.
23
Table S1. XPS analysis of MS-Ti3C2Tx MXene after APS treatment.
Region
BE(eV)
FWHM(eV)
Fraction
Assigned
reference
Ti
2p3/2(2p1/2)
454.5(460.5)
1.3(1.4)
456.0(461.8)
2.3(2.3)
458.2(464.0)
2.2(2.3)
459.7(464.9)
1.4(1.8)
35.5
31.6
20.4
12.5
C 1s
O 1s
Cl
281.1
284.5
286.2
288.5
529.8
531.0
533.6
0.8
1.4
1.2
1.3
1.5
1.6
1.2
2p3/2(2p1/2)
198.8(200.4)
1.0(0.9)
8.9
79.3
6.3
5.5
35.6
58.4
6.0
100
to
Ti-C
Ti-C
(9, 10)
(9, 10)
Ti-Cl
(3, 13)
Ti-O
(10, 11)
Ti-C-Tx
C-C
C-O
C=O
(10)
(3)
(10)
(12)
Ti-O
(10, 11)
Ti-C-Ox
(10)
H2O
(10, 11)
Ti-Cl
(3, 13)
24
EDS analysis of Ti3SiC2 MAX phase and MS-Ti3C2Tx MXene
Table S2 shows the chemical compositions of Ti3SiC2, Ti3C2Cl2-Cu, and MS-Ti3C2Tx
(after immersion in APS solution). EDS results revealed a Cl and O element content of
about 16.49 at.% and 19.79 at.% in the final MS-Ti3C2Tx MXene, respectively. After
APS treatment, the Cl content remains unchanged, and the Cu element content is
reduced from 10.04 at.% to <0.9 at.%, while the S element content is 0.57 at.% obtained
from the (NH4)2S2O8 solution treatment, respectively. The O content is increased to
19.79 at.%, and this may be attributed to water adsorbed during the oxidation treatment
in APS solution; importantly, this value is consistent with mass spectroscopy
measurements (TPD-MS) results.
Table S2. Average chemical composition (at.%) of Ti3SiC2, Ti3C2Cl2-Cu, and MS-
Ti3C2Tx MXene.
EDS
Ti
Si
C
O
Cl
Cu
S
analysis
Ti3SiC2
52.24
17.69
20.75
Ti3C2-Cu
42.69
MS-Ti3C2Tx
42.77
1.25
1.07
6.57
8.88
21.89
10.04
15.25
18.43
19.79
16.49
0.88
0.57
25
Temperature programmed desorption coupled with mass spectroscopy (TPD-MS)
analysis of HF-Ti3C2Tx and MS-Ti3C2Tx MXenes
Fig. S5 shows the TPD-MS analysis results of HF-Ti3C2Tx (Fig. S5A and C) and MS-
Ti3C2Tx MXenes (Fig. S5B and D). The previous study has shown that HF-Ti3C2Tx
MXene decomposes beyond 800°C (Fig. S5C) (14). The decomposition of the surface
groups present on the HF-Ti3C2Tx MXene surface occurs in the 25°C -- 800°C
temperature range. Different from HF-Ti3C2Tx, MS-Ti3C2Tx MXene does not show the
presence of -OH surface groups (Fig. S5A and B). An important CO2 gas release
observed for the MS-Ti3C2Tx MXene is assumed to originate from the oxidation by the
APS of carbon from Ti3C2. The quantification of CO2, H2O and CO was achieved, and
the total content in oxygen was found to be 8.2 wt.%, that is similar to that calculated
from EDS analysis (9.5 wt.%). For the -- Cl groups there are two different species, one
small amount evolving together with hydrogen at temperatures around 800°C
(corresponding to around 3 wt%) and others more thermally stable desorbing at higher
temperatures with a maximum at around 1100°C. Beyond 800°C, where the MXene
decomposes (Fig. S5C and D), the MS analysis shows the presence of other species
including, TiO and SiO, as well as some decomposition products from APS (H2O, N2,
NH3 and SO2 at a lower temperature). However, the quantification of these species was
not possible because of the absence of standards. In the 25 -- 600°C temperature range,
the weight loss associated with the gas evolution of CO2, CO and H2O (Fig. S5B)
accounts for about 15 wt%, showing that oxygenated groups and adsorbed/intercalated
water, together with -Cl groups and -SO2 terminations are the main components of the
MS-Ti3C2Tx MXene surface.
26
Fig. S5. TPD-MS measurements at temperature range up to 800C (A) and full
temperature range (C) of Ti3C2Tx MXene samples (HF-Ti3C2Tx) prepared from
conventional etching treatment in HF; at temperature range up to 600C (B) and full
temperature range (D) for MS-Ti3C2Tx MXene samples. Species marked with asterisks
in (D) were other gases for MS-Ti3C2Tx MXene samples, where no quantification was
possible because of the lack of standards. Weight loss in % and gas evolution in
µmol/g/s are obtained after quantification for H2O, CO, CO2, -OH, and F. (A) and (C)
are adapted from Ref. (14).
27
Table S3. Mass spectroscopy measurements (TPD-MS) analysis of from HF etching
(HF- Ti3C2Tx) and from CuCl2 molten salt route after APS treatment (MS-Ti3C2Tx).
H2O
H2O
-OH
-OH
CO
CO
CO2
CO2
O(total
μmol/
g
wt.
%
μmol/
g
wt.
%
μmol/
g
wt.
%
HF-
Ti3C2T
3600
6.5
3995
6.8
723
2.0
μmol/
g
-
wt.
%
)
wt.%
-
13.3
x
MS-
Ti3C2Tx
2950
5.3
-
-
308
0.9
934
4.1
8.2
28
Guidelines for preparing various MXenes from Lewis acidic molten salts etching
route
The Gibbs free energy and redox potentials were calculated to guide the selection of
suitable MAX phase precursors / Lewis salts to prepare MXene materials. Generally,
the covalent M-X bonding in the MAX phase is very strong, while the M-A boning is
much weaker (15). Hence, we assume that the Ti-C bonding in manuscript equation (1)
remains unchanged during the etching reaction. The equation (1) in the manuscript is
simplified as:
Si + 2CuCl2 = SiCl4 (gas) + 2Cu
(4)
Which can be generalized as (5)
aA + bBCln = aAClm + m.a/nB + (b-m.a/n)BCln (5)
The Gibbs free energies (Gr) between A elements from the MAX phases and Lewis
salts (reaction 4) were calculated by HSC Chemistry software (HSC 6.0). Specifically,
for the equation (5) at 750°C, the values of Hf
(f stands for formation) and Sf can
be obtained from the HSC software, given as Hr (r stands for reaction) of -67.877
kcal and Sr of 20.479 cal K-1. Then Gr is given by:
Gr = Hr -- TSr (6)
Gr value of -371.74 kJ was calculated for equation (4), which indicates that the
reaction is thermodynamically spontaneous. We then generalized the calculations of the
Gibbs free energy by changing A-site element in the MAX phases (such as Al, Zn, In,
Ga, Si, Sn, and Ge, et al.) and cations of the Lewis salts (such as Mn, Zn, Cd, Fe, Co,
Cu, Ni, and Ag). The details are listed in Table S4.
29
Table S4. Gibbs Free Energy Gr of the reaction of different Lewis molten salts with
A-site elements in MAX phases at 700°C.
Gibbs Free Energy (∆G) of different A-site elements (kJ mol-1)
Salts
Al
Zn
In
Ga
Si
Sn
Ge
MnCl2
-0.88
77.64
221.38
164.53
178.21
366.16
335.18
ZnCl2
-117.36
-
104.91
48.06
22.91
210.87
179.89
CdCl2
-166.94
-33.05
55.33
-1.52
-43.18
144.76
113.78
FeCl2
-246.62
-82.29
-24.48
-80.88
-147.35
41.23
9.31
CoCl2
-263.13
-97.18
-40.85
-97.71
-171.44
16.50
-14.47
CuCl2
-410.02
-195.11
-187.74
-244.60
-367.30
-179.35
-210.33
NiCl2
-295.73
-118.91
-73.45
-130.31
-214.91
-26.96
-57.94
AgCl
-290.52
-115.44
-68.25
-125.10
-207.97
-20.02
-51.00
The electrochemical redox potentials of redox couple in halide melts can serve as
another tool to predict the feasibility the Lewis acidic molten salts etching reaction.
Taking Ti3SiC2 in CuCl2 molten salt reaction as an example, the potential of the molten
salt Cu2+/Cu (-0.43 V vs. Cl2/Cl-) is higher than its counterpart Si4+/Si (-1.38 V vs.
Cl2/Cl-) at 700°C. The Si-Si bonding of the Ti3SiC2 phase can be easily broken by the
strong oxidized Cu2+ while the strong covalent Ti-C bonding remains unchanged. The
redox potentials of the molten salts (V vs. Cl2/Cl-) were calculated from equation (7)
and (8) in a temperature range of 400 -- 900oC:
BCln (l) = B(s) + n/2 Cl2 (g) (7)
Where B represents elements such as Al, Fe, Zn, In, Ga, Ge, Si, Sn, Mn, Cu, Co, Ni,
Cd, and Ag, n is the number of exchanged electrons. Gibbs free energy of reaction (7)
was calculated by HSC Chemistry 6.0 (16), and the potential of the reaction (7) was
obtained from (8):
𝐸(𝑉) = −
𝛥𝐺𝑟
𝑛𝐹
(8)
where Gr is the Gibbs free energy per mole of reaction (7) in J mol-1 and F the Faraday
constant, 96,485 C mol-1. The potential E(V) of the reaction (7) corresponds to the
30
potential difference between Bn+/B and Cl2/Cl- redox couples. All the potential values
are shown in Fig. S6 and Table S5.
In this paper, six different MXenes are successfully prepared from eight different MAX
phase precursors etching by various halide molten salts under the predictions of the
Gibbs free energy and redox potentials (Table S6).
Table S5. Redox potentials of the molten salts (V vs. Cl2/Cl-) at the temperature range
of 400-900 oC.
T
Al3+
Zn2+
In3+
Ga3+
Ge4+
Si4+
Sn4+
Mn2+
Fe2+
Cu2+
Co2+
Ni2+
Cd2+
Ag+
(oC)
/Al
/Zn
/In
/Ga
/Ge
/Si
/Sn
/Mn
/Fe
/Cu
/Co
/Ni
/Cd
/Ag
400
-1.90 -1.64 -1.14 -1.35 -1.07 -1.48 -1.00 -1.97
-0.90 -0.61
-1.13
-1.05 -1.45
-0.92
500
-1.88 -1.57 -1.12 -1.32 -1.04 -1.45 -0.97 -1.93
-0.92 -0.54
-1.06
-0.97 -1.39
-0.89
600
-1.86 -1.50 -1.10 -1.30 -1.00 -1.41 -0.93 -1.88
-0.95 -0.48
-1.00
-0.90 -1.32
-0.87
700
-1.84 -1.44 -1.08 -1.27 -0.97 -1.38 -0.89 -1.84 -0.975 -0.43
-0.94
-0.82 -1.27
-0.84
800
-1.82 -1.38 -1.05 -1.25 -0.94 -1.35 -0.86 -1.80
-0.99 -0.37
-0.89
-0.75 -1.21
-0.82
900
-1.80 -1.32 -1.03 -1.22 -0.91 -1.31 -0.82 -1.76
-1.01 -0.32
-0.85
-0.68 -1.16
-0.79
31
Fig. S6. Redox potentials of the molten salts (V vs. Cl2/Cl-) as a function of temperature.
32
Characterizations of various MXenes prepared from Lewis acid molten salts
method
Fig. S7-14 presents the XRD patterns of various MAX phases and the products obtained
after reaction with various Lewis acid salts (Table. S6). It also gives the SEM images
and the corresponding EDS analysis of the series of products. As shown in the XRD
patterns, most of Bragg peaks of the pristine MAX phases disappear after the molten
salt etching process, leaving (00l) peaks and several broad and low intensity peaks,
indicating the successful obtention of layered MXene materials from MAX phases by
Lewis acid molten salts etching route. SEM images show that the pristine particle-like
MAX phases turn into an accordion-like open structure, suggesting the successful
synthesis of MXene such as previous reported for MXenes prepared by HF etching
method (3). EDS analysis indicates the successful removal of A element from the MAX
phases, as well as the presence of Cl and O surface groups on layered MXenes. These
results demonstrate that the Lewis acidic molten salts etching method not only can be
employed as a universal way to prepare these layered materials, but also offers
opportunities for tuning the surface chemistry of MXene.
33
Table S6. The reaction conditions of MAX phases with Lewis acid salts.
MAX Phases
Salts
Composite of starting materials (mol)
T (°C)
Ti2AlC
Ti3AlC2
Ti3AlC2
Ti3AlCN
Ti2AlC
Ti3AlC2
Ti3AlC2
Nb2AlC
Ta2AlC
Ti3ZnC2
Ti3ZnC2
Ti3ZnC2
Ti3ZnC2
Ti3ZnC2
Ti3ZnC2
Ti3SiC2
Ti2GaC
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:5:2:2
MAX:Salt:NaCl:KCl = 1:5:2:2
MAX:Salt:NaCl:KCl = 1:2:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:4:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
MAX:Salt:NaCl:KCl = 1:3:2:2
650
700
700
700
650
700
700
700
700
650
700
700
700
700
700
750
600
CdCl2
FeCl2
CoCl2
CuCl2
CuCl2
NiCl2
CuCl2
AgCl
AgCl
CdCl2
FeCl2
CoCl2
CuCl2
NiCl2
AgCl
CuCl2
CuCl2
34
Fig. S7. Ti2AlC-CuCl2: (A) XRD patterns of Ti2AlC MAX phase before (black) and
after (red) reaction with CuCl2, (B) SEM image and (C) EDS point analysis of the
product after etching process. Ti3ZnC2-CuCl2: (D) XRD patterns of Ti3ZnC2 MAX
phase before (black) and after (red) reaction with CuCl2, (E) SEM image and (F) EDS
point analysis of the product after etching process. Scalebars correspond to 2 μm.
35
Fig. S8. Ti3AlCN-CuCl2: (A) XRD patterns of Ti3AlCN MAX phase before (black) and
after (red) reaction with CuCl2, (B) SEM image and (C) EDS point analysis of the
product after etching process. Ti3AlC2-NiCl2: (D) XRD patterns of Ti3AlC2 MAX phase
before (black) and after (red) reaction with NiCl2, (E) SEM image and (F) EDS point
analysis of the product after etching process. Scalebars correspond to 2 μm.
36
Fig. S9. Ti3ZnC2-FeCl2: (A) XRD patterns of Ti3ZnC2 MAX phase before (black) and
after (red) reaction with FeCl2, (B) SEM image and (C) EDS point analysis of the
product after etching process. Ta2AlC-AgCl: (D) XRD patterns of Ta2AlC MAX phase
before (black) and after (red) reaction with AgCl, (E) SEM image and (F) EDS point
analysis of the product after etching process. Scalebars correspond to 2 μm.
37
Fig. S10. Ti2AlC-CdCl2: (A) XRD patterns of Ti2AlC MAX phase before (black) and
after (red) reaction with CdCl2, (B) SEM image and (C) EDS point analysis of the
product after etching process. Ti3AlC2-FeCl2: (D) XRD patterns of Ti3AlC2 MAX phase
before (black) and after (red) reaction with FeCl2, (E) SEM image and (F) EDS point
analysis of the product after etching process. Scalebars correspond to 2 μm.
38
Fig. S11. Ti3AlC2-CoCl2: (A) XRD patterns of Ti3AlC2 MAX phase before (black) and
after (red) reaction with CoCl2, (B) SEM image and (C) EDS point analysis of the
product after etching process.Nb2AlC-AgCl: (D) XRD patterns of Nb2AlC MAX phase
before (black) and after (red) reaction with AgCl, (E) SEM image and (F) EDS point
analysis of the product after etching process. Scalebars correspond to 2 μm.
39
Fig. S12. Ti3AlC2-CuCl2: (A) XRD patterns of Ti3AlC2 MAX phase before (black) and
after (red) reaction with CuCl2, (B) SEM image and (C) EDS point analysis of the
product after etching process. Ti3ZnC2-CdCl2: (D) XRD patterns of Ti3ZnC2 MAX
phase before (black) and after (red) reaction with CdCl2, (E) SEM image and (F) EDS
point analysis of the product after etching process. Scalebars correspond to 2 μm.
40
Fig. S13. Ti3ZnC2-CoCl2: (A) XRD patterns of Ti3ZnC2 MAX phase before (black) and
after (red) reaction with CoCl2, (B) SEM image and (C) EDS point analysis of the
product after etching process. Ti3ZnC2-NiCl2: (D) XRD patterns of Ti3ZnC2 MAX
phase before (black) and after (red) reaction with NiCl2, (E) SEM image and (F) EDS
point analysis of the product after etching process. Scalebars correspond to 2 μm.
41
Fig. S14. Ti3ZnC2-AgCl: (A) XRD patterns of Ti3ZnC2 MAX phase before (black) and
after (red) reaction with AgCl, (B) SEM image and (C) EDS point analysis of the
product after etching process. Ti2GaC-CuCl2: (D) XRD patterns of Ti2GaC MAX phase
before (black) and after (red) reaction with CuCl2, (E) SEM image and (F) EDS point
analysis of the product after etching process. Scalebars correspond to 2 μm.
42
Electrochemical performance
As shown in Fig. S15A, the main capacity contribution comes from the low potential
range region, which highlights the interest of such material to be used as a negative
electrode in Li-ion containing electrolyte. A maximum capacity of 738 C g-1 (205 mAh
g-1) is achieved within a full potential range of 2.8 V (from 0.2 to 3 V vs. Li+) with a
capacitance to 264 F g-1. 646 C g-1 (180 mAh g-1) can be still delivered within potential
window of 2 V (from 0.2 to 2.2 V vs. Li+/Li) together with a record capacitance of 323
F g-1 for MXene in non-aqueous electrolytes.
Fig. S15. (A) CVs at 0.5 mV s-1 of MS-Ti3C2Tx MXene in 1M Li-PF6 in EC/DMC (1:1)
electrolyte with various positive cut-off potentials; (B) Capacitance and capacity values
in the different potential ranges from the anodic scan.
The coulombic efficiency is 50% in the first cycle (Fig. S16A); the irreversible capacity
at the first cycle corresponds to the SEI layer formation. After several cycles, the
coulombic efficiency stabilizes at 98% for a scan rate of 1 mV s-1 (Fig. S16B). Details
of the discharge capacity and capacitance values of MS-Ti3C2Tx MXene electrode
(active material weight loading of 1.4 mg cm-2) are listed in Table S7. The capacitance
of the MXene electrode at a scan rate of 0.5 mV s-1 is 264 F g-1 (205 mAh g-1) with the
full potential window of 2.8 V. The capacitance remains at 97 F g-1 (75 mAh g-1) when
the scan rate increases to 100 mV s-1 (discharge time of 28 s), which corresponds to a
capacitance retention of 37% as compared to the value of 0.5 mV s-1. Moreover,
increasing the active material weight loading up to 4 mg cm-2 does not hinder the power
capability of the Ti3C2Tx material as can be seen from Fig. S16C and D. Fig. S16D
shows the discharge capacity values calculated from the CVs. The thicker electrode
delivers 680 C g-1 (areal capacity of 2.72 C cm-2) at a scan rate of 0.5 mV s-1 with a
capacity retention of 35% at 100 mV s-1. The high rate performance of the MS-Ti3C2Tx
43
MXene electrode is further confirmed by the galvanostatic test at the full potential range
(Fig. S17A). Specifically, it can deliver 210 mAh g-1 within 1 h and 80 mAh g-1 within
20 s (capacity retention of 38%). Those results suggest that MS-Ti3C2Tx MXene
materials can serve as a high rate anode electrode for the Li-ion storage. 90% capacity
retention was achieved after 2,400 galvanostatic cycles (Fig. S17B).
Fig. S16. (A) First three cycles of CV at 0.5 mVs-1 of a MS-Ti3C2Tx MXene electrode
in LP30 electrolyte. (B) CVs at various potential scan rates of a MS-Ti3C2Tx MXene
electrode in LP30 electrolyte. The active material weight loading is 1.4 mg cm-2. (C)
CVs at various potential scan rates of a MS-Ti3C2Tx MXene electrode with active
material weight loading of 4 mg cm-2 in LP30 electrolyte. (D) Change of the MXene
electrode capacity versus the discharge time calculated from (B) and (C).
44
Table S7. Discharge capacity and capacitance values of a MS-Ti3C2Tx MXene
electrode calculated from the anodic scan of the CVs (Fig. S16B). The active material
weight loading is 1.4 mg cm-2.
Scan rate /
Capacitance /
Capacity / C g-1
C-rate
Coulomb
mV s-1
F g-1
(and mAh g-1)
efficiency / %
0.5
1
2
5
10
20
50
100
264
240
230
205
183
159
122
97
738 (205)
672 (187)
645 (179)
576 (160)
511 (142)
445 (124)
340 (94)
0.6
1.3
2.6
6.4
13
26
64
271 (75)
128
98
98
97
98
99
100
100
100
Fig. S17. (A) Charge/discharge capacities calculated from galvanostatic test at different
C-rate, with the potential range from 0.2 to 3 V vs. Li+/Li. The active material weight
loading is 1.1 mg cm-2 (B) Long cycling at 30 C-rate, 90% of capacity retained after
2400 cycles.
Electrochemical impedance spectroscopy measurements were made at various bias
potentials vs. Li+/Li to understand the electrochemical performance of the MS-Ti3C2Tx
MXene material (Fig. S18A). All the Nyquist plots show similar features with a high
45
frequency semi-circle followed by a fast increase of the imaginary part of the
impedance at low frequencies. The high frequency semi-circle loop is assigned to the
charge-transfer resistance of about 25 Ω cm², which is almost three times larger than
the one observed of a porous MXene electrode in propylene carbonate-based electrolyte
(17). The near-vertical imaginary parts at low frequency range indicate a capacitive-
like charge storage kinetics instead of a diffusion dominated process, as can be seen
from the absence of a Warburg region in the mid frequency range (45° line). Moreover,
the charge storage kinetics are further investigated by determining b-value (see Fig.
S18B) following equation:
i=a vb (9)
It has been suggested that a b-value of 1 relates to the capacitive (surface-like) process,
while a b-value of 0.5 identifies the diffusion-controlled (bulk) process (18, 19). Fig.
S18B shows the (i) versus scan rate plot in log scale from 0.5 to 100 mV s-1. A linear
relationship with a slope of 1 is observed in a scanning potential rate range from 0.5 to
20 mV s-1, indicating a capacitive-like charge storage kinetics. The deviation from this
linear at higher scan rates (>20 mV s-1) may be assigned to kinetics (restricted-
diffusion) or/and ohmic limitations at high current density.
Fig. S18. (A) The electrochemical impedance spectroscopy plots recorded at various
bias potentials. The Nyquist plots show a high frequency loop of about 25 Ω cm²
associated with the charge transfer resistance, and a diffusion-restricted behavior at low
frequency. (B) Change of the peak current with the potential scan rate in log scale. A
slope of 1 stands for a surface-controlled process, while a slope of 0.5 indicates a
diffusion-controlled reaction.
46
Moreover, Ti3C2Tx MXene prepared from Ti3AlC2 MAX precursor exhibits similar
electrochemical behavior of MS-Ti3C2Tx in LP30. Almost identical CV signatures were
observed and presented in Fig. S19A. This Al-MAX derived MXene electrode delivers
730 C g-1 at a scan rate of 0.5 mV s-1 and possesses a capacity retention of 36% at a
scan rate of 100 mV s-1 (Fig. S19B). These results indicate that the Lewis acidic molten
salts etching route is a promising method to prepare high rate electrode MXene
materials.
Fig. S19. (A) CVs at various potential scan rates of a Ti3C2Tx MXene electrode
prepared from Ti3AlC2 MAX phase in LP30 electrolyte. (B) Change of the Ti3C2Tx
MXene electrode capacity versus the discharge time during CVs recorded at various
potential scan rates. The active material weight loading is 1.1 mg cm-2.
47
References and Notes
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
M. Li et al., Copper -- SiC whiskers composites with interface optimized by
Ti3SiC2. Journal of Materials Science 53, 9806-9815 (2018).
Q. Wang et al., Synthesis of High-Purity Ti3SiC2 by Microwave Sintering.
International Journal of Applied Ceramic Technology 11, 911-918 (2014).
M. Li et al., Element Replacement Approach by Reaction with Lewis Acidic
Molten Salts to Synthesize Nanolaminated MAX Phases and MXenes. Journal
of the American Chemical Society 141, 4730-4737 (2019).
B. Manoun et al., Synthesis and compressibility of Ti3(Al,Sn0.2)C2 and
Ti3Al(C0.5,N0.5)2. Journal of applied physics 101, 113523 (2007).
C. Hu et al., Microstructure and properties of bulk Ta2AlC ceramic synthesized
by an in situ reaction/hot pressing method. Journal of the European Ceramic
Society 28, 1679-1685 (2008).
W. Zhang, N. Travitzky, C. Hu, Y. Zhou, P. Greil, Reactive hot pressing and
properties of Nb2AlC. Journal of the American Ceramic Society 92, 2396-2399
(2009).
O. Çakır, Review of Etchants for Copper and its Alloys in Wet Etching
Processes. Key Engineering Materials 364-366, 460-465 (2007).
M. Morcrette et al., In situ X-ray diffraction techniques as a powerful tool to
study battery electrode materials. Electrochimica Acta 47, 3137-3149 (2002).
E. Kisi, J. Crossley, S. Myhra, M. Barsoum, Structure and crystal chemistry of
Ti3SiC2. Journal of Physics and Chemistry of Solids 59, 1437-1443 (1998).
J. Halim et al., X-ray photoelectron spectroscopy of select multi-layered
transition metal carbides (MXenes). Applied Surface Science 362, 406-417
(2016).
12.
14.
13.
11. M. Han et al., Ti3C2 MXenes with modified surface for high-performance
electromagnetic absorption and shielding in the X-band. ACS applied materials
& interfaces 8, 21011-21019 (2016).
Q. Xue et al., Mn3O4 nanoparticles on layer-structured Ti3C2 MXene towards
the oxygen reduction reaction and zinc -- air batteries. Journal of Materials
Chemistry A 5, 20818-20823 (2017).
C. Mousty‐Desbuquoit, J. Riga, J. J. Verbist, Solid state effects in the electronic
structure of TiCl4 studied by XPS. The Journal of Chemical Physics 79, 26-32
(1983).
O. Mashtalir et al., The effect of hydrazine intercalation on the structure and
capacitance of 2D titanium carbide (MXene). Nanoscale 8, 9128-9133 (2016).
15. M. Magnuson, M. Mattesini, Chemical bonding and electronic-structure in
MAX phases as viewed by X-ray spectroscopy and density functional theory.
Thin Solid Films 621, 108-130 (2017).
S. Guo, J. Zhang, W. Wu, W. Zhou, Corrosion in the molten fluoride and
chloride salts and materials development for nuclear applications. Progress in
Materials Science 97, 448-487 (2018).
X. Wang et al., Influences from solvents on charge storage in titanium carbide
MXenes. Nature Energy 4, 241 (2019).
J. Wang, J. Polleux, J. Lim, B. Dunn, Pseudocapacitive Contributions to
Electrochemical Energy Storage in TiO2 (Anatase) Nanoparticles. The Journal
of Physical Chemistry C 111, 14925-14931 (2007).
16.
17.
18.
48
H. Shao, Z. Lin, K. Xu, P.-L. Taberna, P. Simon, Electrochemical study of
pseudocapacitive behavior of Ti3C2Tx MXene material in aqueous electrolytes.
Energy Storage Materials 18, 456-461 (2019).
19.
49
|
1909.11174 | 1 | 1909 | 2019-08-05T16:10:40 | Coherent WDM transmission using quantum-dash mode-locked laser diodes as multi-wavelength source and local oscillator | [
"physics.app-ph",
"eess.SP",
"physics.optics"
] | Quantum-dash (QD) mode-locked laser diodes (MLLD) lend themselves as chip-scale frequency comb generators for highly scalable wavelength-division multiplexing (WDM) links in future data-center, campus-area, or metropolitan networks. Driven by a simple DC current, the devices generate flat broadband frequency combs, containing tens of equidistant optical tones with line spacings of tens of GHz. Here we show that QD-MLLDs can not only be used as multi-wavelength light sources at a WDM transmitter, but also as multi-wavelength local oscillators (LO) for parallel coherent reception. In our experiments, we demonstrate transmission of an aggregate data rate of 4.1 Tbit/s (23x45 GBd PDM-QPSK) over 75 km standard single-mode fiber (SSMF). To the best of our knowledge, this represents the first demonstration of a coherent WDM link that relies on QD-MLLD both at the transmitter and the receiver. | physics.app-ph | physics | Coherent WDM transmission using quantum-dash mode-locked laser
diodes as multi-wavelength source and local oscillator
Juned N. Kemal,1 Pablo Marin-Palomo,1 Vivek Panapakkam,2 Philipp Trocha,1 Stefan Wolf,1
Kamel Merghem,2 Francois Lelarge,3 Abderrahim Ramdane,2 Sebastian Randel,1 Wolfgang
Freude,1 And Christian Koos 1,4,*
1Institute of Photonics and Quantum Electronics (IPQ), Karlsruhe Institute of Technology (KIT), Germany
2Centre de Nanosciences et de Nanotechnologies (CNRS), Univ. Paris-Sud, C2N - Université Paris-Saclay, France
3Now with Almae Technologies, Marcoussis, France
4Institute of Microstructure Technology (IMT), Karlsruhe Institute of Technology (KIT), Germany
*[email protected]
Quantum-dash (QD) mode-locked laser diodes (MLLD) lend themselves as chip-scale frequency comb generators
for highly scalable wavelength-division multiplexing (WDM) links in future data-center, campus-area, or
metropolitan networks. Driven by a simple DC current, the devices generate flat broadband frequency combs,
containing tens of equidistant optical tones with line spacings of tens of GHz. Here we show that QD-MLLDs can
not only be used as multi-wavelength light sources at a WDM transmitter, but also as multi-wavelength local
oscillators (LO) for parallel coherent reception. In our experiments, we demonstrate transmission of an aggregate
data rate of 4.1 Tbit/s (23×45 GBd PDM-QPSK) over 75 km standard single-mode fiber (SSMF). To the best of
our knowledge, this represents the first demonstration of a coherent WDM link that relies on QD-MLLD both at
the transmitter and the receiver.
1.
Introduction
Driven by the explosive growth of data traffic on all levels
of optical communication networks [1], wavelength-
division multiplexing (WDM) has become a necessity not
only for long-haul transmission, but also for shorter links
that connect, e.g., data centers across metropolitan or
campus-area networks. These links crucially rely on
compact WDM transceivers that offer multi-terabit/s
aggregate data rates and that can be deployed in large
quantities at low cost. In this context, optical frequency
combs have emerged as a promising approach towards
compact and efficient light sources that provide a
multitude of tones for parallel WDM transmission [2-9].
Unlike the carriers derived from a bank of individual
lasers, the tones of a comb are intrinsically equidistant in
frequency and are defined by just two parameters the
center frequency and the free spectral range (FSR).
Among
the different frequency comb sources,
quantum-dash mode-locked laser diodes (QD-MLLD) are
particularly interesting due to their technical simplicity:
Driven by a DC current, these devices emit broadband
frequency combs consisting of tens of optical carriers,
spaced by free spectral ranges of, e.g., 25 GHz or 50 GHz,
as determined by
length. In previous
experiments [6,9], the potential of QD-MLLD has been
demonstrated by using the devices as multi-wavelength
optical sources for parallel transmission on tens of WDM
the cavity
channels, achieving aggregate lines rates of more than
10 Tbit/s. In these experiments, symbol-wise phase-
tracking [9] or self-injection locking [6] was used to
overcome the limitations imposed by the rather high
intrinsic linewidth of the QD-MLLD tones. However,
while these demonstrations show the great potential of
QD-MLLD as multi-wavelength light sources at the
WDM transmitter, demodulation of the coherent signals
still relied on a single-wavelength external-cavity laser
(ECL), which acts as a local oscillator (LO) and needs to
be tuned to the respective channel of interest. The
scalability advantages of QD-MLLD for massively
parallel WDM transmission have hence only been
exploited at the transmitter, but not at the receiver side.
In this work, we demonstrate that QD-MLLD can also
act as multi-wavelength local oscillators for intradyne
reception of a multitude of optical channels [10]. In our
experiments, we use a pair of QD-MLLDs with similar
free spectral ranges (FSR) -- one device to generate the
optical carriers for the WDM transmitter, and the other
device to provide the required LO tones for parallel
coherent reception. We use 23 carriers separated by
50 GHz to transmit data at a symbol rate of 45 GBd using
QPSK signaling, thereby obtaining data streams with an
aggregate line rate of 4.14 Tbit/s and an aggregate net data
rate of 3.87 Tbit/s which are transmitted over 75 km of
standard single mode fiber (SSMF). To the best of our
knowledge, this is the first demonstration of a QD-MLLD
acting as a multi-wavelength LO
in coherent
transmission. When combined with highly integrated
coherent transceiver circuits [11-14] and advanced digital
signal processing (DSP) schemes implemented, e.g., in
7 nm CMOS technology [15,16], our approach could
open a path towards coherent WDM transceivers with
unprecedented compactness and scalability.
2. Optical frequency combs for WDM
transmission and reception
The basic concept of comb-based WDM transmission and
reception is shown in Fig. 1. The tones from the
transmitter comb
(Tx comb) are demultiplexed
(DEMUX), and data is encoded onto the individual
carriers by in-phase/quadrature modulators (IQ mod.).
The optical channels are combined using a frequency
multiplexer (MUX), amplified if needed, and transmitted
through the transport fiber. At the WDM receiver, the
of the MLLD via p-doped and n-doped InP layers,
respectively. Lateral confinement of carriers is obtained
by proton implantation [17]. Multiple longitudinal modes
can coexist in the laser cavity due to the inhomogeneously
broadened gain of the active region originating from the
shape distribution of the QDs. Mode locking due to
mutual sideband injection [18] produces an optical
frequency comb output centered at 1545 nm with a line
spacing of 25 GHz.
In coherent reception, it is desired to maintain a low
frequency offset between the carrier of each channel and
the corresponding LO tone at the receiver. This requires a
Tx comb and an LO comb with similar center frequencies
and free spectral ranges. To this end, we used two QD-
MLLDs that are taken from the same wafer and are
carefully cleaved to have very similar cavity lengths, thus
resulting in overlapping spectra for appropriately set
operating currents and temperatures. To select the
operating conditions of
two QD-MLLDs, we
the
Fig. 1. Setup of a coherent WDM transmission link using optical frequency combs as light sources both at the transmitter and
the receiver. WDM transmitter: Carriers from a transmitter comb (Tx comb) are demultiplexed (DEMUX) and modulated
separately using in-phase/quadrature modulators (IQ mod.). The channels are then multiplexed (MUX) to form the WDM
signal. Link: The WDM signal is amplified and transmitted. WDM receiver: The channels are demultiplexed and sent to an
array of independent coherent receivers (Coh. Rx). The demultiplexed spectral lines of a local oscillator comb (LO comb)
serve as LO tones for the various coherent receivers. EDFA: Erbium-doped fiber amplifier.
channels are demultiplexed and sent to an array of
independent coherent
(Coh. Rx). The
demultiplexed spectral lines of a second frequency comb
(LO comb) serve as local oscillator tones for the various
coherent receivers.
receivers
For our data transmission experiment, we use two
QD-MLLDs -- one as the Tx comb and the second as the
LO comb. Figure 2(a) shows the basic structure of a QD-
MLLD with an active region that consists of three stacked
layers of InAs quantum dashes separated by InGaAsP
barriers. Each QD layer is 2 nm thick and the barriers
have a thickness of 40 nm. Two 80 nm-thick separate
confinement heterostructure (SCH) layers of InGaAsP
terminate the dash-barrier stack towards the top and
bottom InP layers [17,18]. The optical mode is guided by
a buried ridge waveguide of 1.0 µm width. Cleaved chip
facets form a Fabry-Perot laser cavity with a length of
1.71 mm, leading to an FSR of 25 GHz. Top and bottom
gold layers provide electrical contacts to the active region
characterized the devices using the setup in Fig. 2(b). The
emitted light is coupled to a lensed fiber, which is
equipped with anti-reflection (AR) coating for 1550 nm
and connected to an optical isolator to reduce optical
back-reflection
cavity. Different
instruments, denoted by C, D, and E, are connected to the
output of the isolator for measuring the data of Figs. 2 (c),
2(d), and 2(e), respectively.
laser
into
the
Figure 2(c) shows the total output optical power of the
Tx comb and the LO comb as a function of the injection
current. Insertion losses of coupling interface to the
lensed fiber (2 dB) and the isolator (1 dB) are taken into
account. At a pump current of 300 mA, the output power
radiated from the facet of the Tx comb exceeds 18 mW
(12.5 dBm). The LO comb power peaks at 12 mW
(10.8 dBm) for an injection current of 200 mA. We
attribute the smaller output power of the LO comb to
imperfect mounting and
the
associated chip in our experiment.
thermal coupling of
Fig. 2. Concept and characterization of quantum-dash mode-locked laser diodes (QD-MLLD).
(a) Three-dimensional
schematic of a QD-MLLD. The active region consists of three stacked layers of InAs quantum dashes, which are separated by
InGaAsP barriers. The stack is terminated by a separate confinement heterostructure (SCH) layer to the top and the bottom.
The buried ridge waveguide has a width of 1.0 µm. Cleaved chip facets form a Fabry-Perot laser cavity with a length of 1.71
mm, leading to an FSR of 25 GHz. Top and bottom gold layers provide electrical contacts to the active region of the MLLD
via p-doped and n-doped InP layers.
(b) Basic optical setup for using the QD-MLLD comb source. The device is driven by
a constant injection current (not shown). The emitted light is collected by a lensed fiber, and an optical isolator is used to
reduce back-reflections from the optical setup into the QD-MLLD. Setups C, D, and E are used for measuring the data of
Subfigures (c), (d), and (e), respectively. PM: Power meter; ESA: Electrical spectrum analyzer; OSA: Optical spectrum
(c) Measured total comb power versus injection current of the Tx and the LO combs. Due to imperfect mounting
analyzer.
of the LO QD-MLLD chip, the LO comb performs worse than the Tx comb.
(d) Full-width half maximum (FWHM) of the
radio-frequency (RF) beat note (upper plot) as a function of the drive current for the Tx comb (blue) and for the LO comb
(red), and free spectral range (FSR, lower plot) as a function of the drive current for the Tx comb (blue), and for the LO comb
(red). For some drive currents, we observe distinctively higher RF linewidths, see, e.g., Point I. In these cases, we often observe
more than one beat note in the RF spectrum, Inset I, which we attribute to the coexistence of two sub-combs with slightly
different FSR [20]. For such operating states, the depicted large RF linewidth values are to be understood as an indication of
unstable operation by the MLLD. Inset II shows the measured RF beat note corresponding to stable single-mode operation of
the QD-MLLD. In the insets, the spectral power is normalized to the dominant peak, and the frequency axis is defined with
respect to the spectral position of this peak.
(e) Optical power spectra of the Tx comb (blue) and the LO comb (red). The LO
comb has a slightly lower bandwidth than the Tx comb. RBW: Resolution bandwidth of the OSA.
Figure 2(d), upper plot shows the measured FWHM
of the RF beat note, commonly referred to as the RF
linewidth, as a function of the injection current, both for
the Tx comb and the LO comb. The RF beat note results
from mixing neighboring comb lines in a photodiode and
is measured by connecting the photodiode to an electrical
spectrum analyzer (ESA), see Fig. 2(b), Setup D. A
narrow RF linewidth indicates a reduced optical phase
noise of the individual comb lines [19], which is
advantageous for coherent optical communication. For
certain drive currents, we observe comparatively large RF
linewidths, see, e.g., Point I in Fig. 2(d). In these cases,
we often observe more than one beat note in the RF
spectrum, Inset I in Fig. 2(d), which we attribute to the
coexistence of two sub-combs with slightly different FSR
in the QD-MLLD cavity [20]. For such operating states,
the large RF linewidth values refer to the dominant
spectral peak and should only be taken as indicators for
unstable operation by the QD-MLLD. Figure 2(d), Inset
II shows the measured RF beat note corresponding to a
stable single-mode operating state of the QD-MLLD. The
lower plot in Fig. 2(d) shows the measured FSR of the Tx
comb and the LO comb as a function of the injection
current. During the above current sweeps, the sub-mounts
of the Tx and LO MLLD chips are temperature-stabilized
at 23.6 oC and 19.0 oC, respectively. These temperatures
were chosen such that the tones of the two combs roughly
coincide in center frequency. As can be seen in Fig. 2(d),
the Tx comb has a small RF linewidth for a wide range of
injection currents above 130 mA, whereas the LO comb
Fig. 3. Phase-noise characterization of QD-MLLD.
(a) Experimental setup. Narrow-band tunable band-pass filters (TBF)
are used to select distinct lines with comparable frequencies from the Tx comb and the LO comb. A polarization controller
(PC) is used to match the polarizations of the selected tones at the input of a 90° optical hybrid. Balanced photodetectors
(BPD) deliver the electrical in-phase (I) and quadrature (Q) signals, which are sampled with a real-time oscilloscope (RTO)
for offline processing.
(b) Power spectrum of the beat note between the two tones selected from the Tx comb and the LO
comb as a function of the offset from the beat note frequency (green trace). The duration of the corresponding time-domain
signals is 0.4 µs. We use the IQ data to independently extract a short-term Lorentzian linewidth of 5.1 MHz. The associated
line shape (red trace) coincides well with the measured spectrum.
exhibits a small RF linewidth only for injection currents
between 90 mA and 130 mA. This difference in behavior
is again attributed to the imperfect mounting of the LO
MLLD chip. For the subsequent WDM experiments with
the two combs, we used injection currents of 278 mA and
125 mA for the Tx comb and the LO comb, respectively,
giving nearly the same FSR (25 GHz) for both combs
while maintaining low RF linewidths of less than 20 kHz.
The optical spectra of the combs operated at the selected
injection currents are shown in Fig. 2(e). The LO comb
has a lower 3 dB bandwidth than the Tx comb, thereby
limiting the total number of WDM channels that can be
received.
The modulation formats and the symbol rates
employed in a coherent communication link depend on
the combined short-term linewidth of the transmitter
carrier and the receiver LO [21]. Figure 3(a) shows the
experimental setup used for estimating the linewidth of
the beat note generated by mixing carrier and LO tones.
Tunable band-pass filters (TBF) select single comb tones
with comparable optical frequencies from the Tx comb
and the LO comb. The selected tones are coupled to a 90o
optical hybrid with a polarization controller in the Tx
comb path to align the polarizations of the tones. The in-
phase and quadrature outputs of the optical hybrid are
detected with balanced photodetectors (BPD), and the
electrical signals are sampled with two-channels of a real-
time oscilloscope (RTO). Figure 3(b), green trace, shows
the power spectrum of the beat note between a carrier tone
of the Tx comb and the corresponding tone of the LO
comb. The plot is obtained by a Fourier transform of the
sampled time-domain data, which was recorded over a
duration of 0.4 µs. The horizontal axis refers to the
frequency offset from the center of the beat note.
For further analysis of the phase-noise characteristics,
of the recorded I/Q signal and
the combined short-term
we extract the phases
use
it
for determining
(Lorentzian) linewidth of the tones from the Tx and LO
combs. To this end, the differences of the phases
are calculated for
various delay times
is
computed. From this, the short-term Lorentzian linewidth
at
is extracted from the slope of the variance
and the variance
small time delays [22],
.
(1)
This technique leads to a 5.1 MHz Lorentzian linewidth
for the beat note. The same result is obtained by analyzing
the FM noise spectrum and extracting the spectrally
constant white-noise component [23]. The corresponding
Lorentzian lineshape is plotted in Fig. 3(b), red trace, and
shows good agreement with the directly measured
lineshape, green trace. This indicates that, for the chosen
observation time of 0.4 µs, the lineshape is dominated by
high-frequency phase noise rather than by low-frequency
drift. Since the estimated linewidth of the QD-MLLD
tones is relatively high compared to that obtained from
benchtop-type continuous-wave (CW) ECL, we use high
symbol rates and employ symbol-wise phase tracking as
in [9] for the transmission experiments discussed in
Section 3.
3. Coherent optical communications using
QD-MLLDs
3.1 Experimental setup for WDM transmission
For an experimental demonstration of coherent WDM
transmission with QD-MLLDs as light sources both at the
transmitter (Tx) and receiver (Rx), we use the setup
depicted in Fig. 4(a). The total Tx comb output power of
9 dBm is boosted by an erbium-doped fiber amplifier
(EDFA-1). The spacing between the amplified comb lines
t2itttft02f220()lim2f
Fig. 4. Coherent WDM transmission experiment using QD-MLLD comb generators both as a multi-wavelength light source at the
transmitter (Tx) and as a multi-wavelength LO at the receiver (Rx).
(a) Experimental setup: The Tx comb is de-interleaved into
even and odd carriers using a programmable filter (PF-1). The two sets of carriers are then modulated with independent data streams
to emulate a real WDM signal with independent data on neighboring channels. Polarization-division multiplexing (PDM) is
emulated by splitting the single-polarization WDM signal into two paths, delaying one of them, and recombining the two
decorrelated signals on orthogonal polarizations of the transmission fiber. At the receiver, we subsequently select individual lines
of an LO comb and use them for coherent detection of the corresponding WDM channel. The WDM channels are spaced by
approximately 50 GHz and carry QPSK signals at symbol rates of 45 GBd, limited by the bandwidth of the arbitrary-waveform
generator (AWG) used to generate the modulator drive signals at the Tx. P1, P2, and P3 denote points in the experimental setup at
which the optical spectra in Subfigure (b) have been taken. EDFA: Erbium-doped fiber amplifier, IQ: In-phase/quadrature
modulator; DL: Delay line; VOA: Variable optical attenuator; PBC: Polarization beam combiner; TBF: Tunable band-pass filter;
SSMF: Standard single-mode fiber; ESA: Electrical spectrum analyzer; CD: Chromatic dispersion; MIMO: Multiple-input and
(b) Comb spectra of both even and odd carriers before modulation (upper plot, P1), spectrum of 23 modulated
multiple-output.
carriers (lower plot, P2), and spectrum of the LO comb (lower plot, P3).
(c) RF beat note spectrum resulting from the self-beating
of nearest-neighbored comb lines of each comb when detected directly with a photodetector. The spectrum was obtained by
impinging the output of the two QD-MLLDs on a photodetector simultaneously as depicted in Subfigure (a).
is increased to 50 GHz by selecting every second line
with a programmable filter (PF-1, Finisar) such that a
symbol rate of 45 GBd can be used, see spectrum in
Fig. 4(b) which was taken at point P1 of the experimental
setup. The programmable filter PF-1 is additionally used
to flatten the comb spectrum. We emulate WDM traffic
by encoding independent data streams on neighboring
carriers. This is done by de-interleaving 23 spectral lines
from the Tx comb into even and odd carriers using PF-1.
The even and odd carriers are amplified by EDFA-2 and
carriers
are
respectively. Both
EDFA-3,
then
independently modulated with pseudo-random bit
sequences (PRBS) of length 211 1. This bit sequence is
mapped to QPSK symbols with a rate of 45 GBd and a
raised-cosine spectrum having a 10 % roll-off. In our
experiments, the symbol rate was limited by the
bandwidth of the arbitrary-waveform generator (AWG)
used to generate the drive signals for the IQ modulators at
the transmitter. The modulated even and odd carriers are
then recombined by a polarization-maintaining (PM)
3 dB coupler to form a WDM signal with uncorrelated
data streams on neighboring channels. To emulate
polarization-division multiplexing (PDM), the WDM
signal is split into two paths. In one of the paths, the signal
is delayed by about 5.34 ns (240 symbols) with an optical
delay line (DL) for decorrelation. A variable optical
attenuator (VOA) in the other path equalizes the signal
powers in the two paths. The signals are then recombined
in orthogonal polarization states using a polarization
beam combiner (PBC). The dual-polarization WDM
signal is amplified and transmitted over a 75 km SSMF
link, see spectrum in Fig. 4(b), which was taken at point
P2 of the experimental setup. At the receiver, the WDM
channel of interest is selected using a 0.6 nm tunable
band-pass filter (TBF) and amplified by EDFA-5. A
second 1.5 nm-wide TBF is used to suppress out-of-band
amplified spontaneous emission (ASE) noise of EDFA-5.
The selected channel is detected by a dual-polarization
coherent receiver (Keysight N4391A). The corresponding
reference tone for coherent detection is selected from the
LO comb using a second programmable filter (PF-2),
amplified by EDFA-6, and filtered by another 0.6 nm
TBF for ASE noise suppression. The output of the
coherent receiver is digitized using four oscilloscope
channels with a sampling rate of 80 GSa/s each (Keysight
DSOX93204A) and evaluated offline using digital signal
processing (DSP). In the DSP, we first up-sample the
received signal to two samples per symbol. This is
followed by blind chromatic-dispersion compensation
based on the Godard clock-tone algorithm [24] after
which timing recovery is carried out [25,26]. An adaptive
multiple-input and multiple-output (MIMO) equalizer
using the constant-modulus algorithm (CMA) then
demultiplexes the two orthogonal polarizations of the
signal and compensates linear transmission impairments
[27]. We use periodogram-based
to
compensate for frequency offsets between the signal
carrier and the LO tone. Subsequently, carrier phase noise
compensation is carried out based on a blind phase search
(BPS) algorithm [21]. This technique allows for symbol-
wise phase tracking and has been shown to effectively
cope with the phase noise of QD-MLLD carriers in
techniques
previous experiments [9]. Figure 4(b) depicts
the
superposition of the LO comb tones with the WDM
signal. As discussed in Section 2, the FSR of the combs
are nearly identical and the modulated channels are hence
in good frequency alignment with the corresponding LO
tones. Figure 4(c) shows the RF beat notes of the Tx comb
and the LO comb captured simultaneously using an
electrical
(ESA, Keysight
PXA N9030A), see "FSR monitor" in Fig. 4(a). The two
combs have narrow RF linewidths (< 20 kHz) and FSRs
with a very small difference ( FSR = 1.4 MHz).
spectrum
analyzer
3.2 WDM transmission over 75 km at a net data
rate of 3.8 Tbit/s
The results of our transmission experiment are shown in
Fig. 5. Throughout our experiments, the BER was too low
to be measured within a limited record length of 1×06 bits.
We therefore rely on the error-vector magnitude (EVM,
normalized to the maximum constellation amplitude) to
quantify signal quality and estimate the BER assuming
that, after phase-noise compensation, the QPSK signals
are impaired by additive white Gaussian noise only [28].
The measured EVM for each WDM channel is depicted
in the upper plot of Fig. 5(a), and the corresponding BER
estimates are shown in the lower plot. All channels fall
well below the BER threshold of 4.5×103 for hard-
decision forward-error correction (HD-FEC) with 7%
overhead [29]. This is confirmed by the fact that none of
our signal recordings of 1×106 bits each showed any error.
This leads to an aggregate line rate of 4.14 Tbit/s and a
net data rate of 3.87 Tbit/s, transmitted over 75 km of
SSMF. Example constellation diagrams obtained for the
channel at 194.15 THz, both for back-to-back and for
fiber transmission are shown in Fig. 5(b). The PRBS
length used in these experiments was 211 1.
3.3 Comparison between QD-MLLD LO and ECL
LO
In this section, we investigate whether or not there is a
penalty in received signal quality when using a QD-
MLLD comb line as LO rather than an ECL LO. The
experimental setup is depicted in Fig. 6(a) and relies on
an ECL as light source at the Tx. The ECL carrier, having
a linewidth of less than 100 kHz and a power of 6 dBm,
is amplified by EDFA-1 prior to modulation with a QPSK
signal at a symbol rate of 45 GBd. In this experiment, we
used PRBS of length 211 1 and 215 1, both of which
lead to identical results. In the following, we show only
the results for the longer sequence. PDM is emulated as
discussed in Section 3.1. The PDM signals are amplified
by EDFA-2, filtered by a TBF to remove ASE noise, and
sent to the signal input of a dual-polarization coherent
Fig. 5. Coherent data transmission performance obtained by using a QD-MLLD comb generator both as the Tx light source
and as the LO.
(a) Upper plot: Measured EVM as a function of the carrier frequency of the transmitted channels for both
back-to-back (btb, ▲) and 75 km SSMF transmission (■). Lower plot: BER estimated from EVM assuming that the QPSK
signals are impaired by additive white Gaussian noise only [28].
(b) Example constellation diagrams for the signal carried
by the comb line at 194.15 THz, both for back-to-back (btb) and for 75 km fiber transmission.
receiver. The modulated signal is received by using either
a QD-MLLD comb line or a tone obtained from a second
ECL as LO. The LO power is boosted by EDFA-3 and
filtered by PF, which is mainly needed for the comb input.
We vary the optical carrier-to-noise power ratio (OCNR)
of the LO tone using a variable optical attenuator (VOA)
in front of another EDFA (EDFA-4). The OCNR for a
reference noise bandwidth of 12.5 GHz (0.1 nm) is
measured at the monitor output of the dual-polarization
coherent receiver (Keysight N4931A). Offline DSP as
described in Section 3.1 is employed to analyze the
received data. Figure 6(b) shows the measured BER with
1106 analyzed bits versus the OCNR of the LO. Note
that, due to the limited number of 106 analyzed bits, the
BER values below 105 might be subject to statistical
inaccuracies. Different LO tones from the QD-MLLD are
tested by using the corresponding Tx carrier frequency.
As can be seen from Fig. 6(b), LO tones from a QD-
MLLD (solid lines) perform similarly to an ECL LO
(dotted lines). However, when using QD-MLLD LO
tones at low OCNR, cycle slips were observed in the
recovered symbols due to incorrect estimation of the
carrier phase by a multiple of π/2. The impact of this cycle
slips can be reduced by using differential coding along
Fig. 6. Comparison of transmission performance obtained by using a QD-MLLD and a conventional external-cavity lasers
(ECL) as LO tone generators.
(a) Experimental setup. At the Tx, we use an ECL having a linewidth below 100 kHz as light
source. The ECL carrier is boosted by an EDFA (EDFA-1) and modulated with a QPSK signal at a symbol rate of 45 GBd.
The length of the underlying pseudo-random bit sequence amounts to 215 1. Polarization-division multiplexing (PDM) is
emulated by the same method discussed in Section 3.1. The PDM signals are amplified by an EDFA (EDFA-2), filtered by a
tunable bandpass filter (TBF), and sent to the signal input of a dual-polarization coherent receiver. The signal is detected by
using either a QD-MLLD comb tone as an LO, or by using a second ECL. The LO is amplified by EDFA-3 and filtered by a
programmable filter. A variable optical attenuator (VOA) in front of EDFA-4 is used to vary the optical carrier-to-noise power
ratio (OCNR) of the LO tone.
(b) Plot of BER vs. OCNR of the LO tone. Results obtained from the ECL are depicted as
dotted lines, whereas the solid lines refer to the QD-MLLD LO tones. The OCNR is indicated with respect to a reference noise
bandwidth of 12.5 GHz (0.1 nm). We find that LO tones from a QD-MLLD (solid lines) perform similarly to an ECL LO
(dotted lines).
with phase-slip tolerant FEC schemes [30] or by adding
pilot symbols for accurate phase referencing [31].
4. Summary
We demonstrate that QD-MLLD can be used both as
multi-wavelength light sources at the Tx and as multi-
wavelength LO at the Rx of massively parallel WDM
systems. In out experiments, we use 23 carriers to
transmit a line rate (net data rate) of 4.140 Tbit/s (3.869
Tbit/s) over a 75 km standard single mode fiber (SSMF)
link. To the best of our knowledge, this is the first time
that an optical frequency comb generated by a QD-MLLD
has been used as a multi-wavelength LO in a coherent
transmission system. Due to their compactness and
simple operation requiring a DC drive current only, QD-
MLLD
interesting comb
generator concept for efficient and highly scalable WDM
transceivers.
represent a particularly
Funding and Acknowledgments
This work was supported by the European Research
Council (ERC Starting Grant 'EnTeraPIC', # 280145,
ERC Consolidator Grant 'TeraSHAPE', # 773248), the
EU project BIG PIPES (# 619591), by the Alfried Krupp
von Bohlen und Halbach Foundation, by the Helmholtz
International Research School for Teratronics (HIRST),
and by the Karlsruhe School of Optics & Photonics
(KSOP). We gratefully acknowledge support from the
Erasmus Mundus
program
EUROPHOTONICS (Grant No. 159224-1-2009-1-FR-
ERA MUNDUS-EMJD) and
the Deutsche
Forschungsgemeinschaft (DFG) through CRC 1173.
Doctorate
Joint
from
References
1. Cisco Visual Networking. "Cisco global cloud index: forecast
and methodology, 2016-2021, (white paper), " Cisco, 2018
2. D. Hillerkuss, R. Schmogrow, M. Meyer, S. Wolf, M. Jordan, P.
Kleinow, N. Lindenmann, P. C. Schindler, A. Melikyan, X. Yang,
S. Ben-Ezra, B. Nebendahl, M. Dreschmann, J. Meyer, F.
Parmigiani, P. Petropoulos, B. Resan, A. Oehler, K. Weingarten,
L. Altenhain, T. Ellermeyer, M. Moeller, M. Huebner, J. Becker,
C. Koos, W. Freude, and J. Leuthold, "Single-laser 32.5 {Tbit/s
Nyquist WDM} transmission," J. Opt. Commun. Netw. 4, 715 --
723 (2012).
3. C. Weimann, P. C. Schindler, R. Palmer, S. Wolf, D. Bekele, D.
Korn, J. Pfeifle, S. Koeber, R. Schmogrow, L. Alloatti, D. Elder,
H. Yu, W. Bogaerts, L. R. Dalton, W. Freude, J. Leuthold, and C.
Koos, "Silicon-organic hybrid (SOH) frequency comb sources for
terabit/s data transmission," Opt. Express 22(3), 3629 -- 3637
(2014).
4. V. Ataie, E. Temprana, L. Liu, E. Myslivets, B. P.-P. Kuo, N. Alic,
and S. Radic, "Ultrahigh count coherent WDM channels
transmission using optical parametric comb-based frequency
synthesizer," J. Lightw. Technol. 33(3), 694 -- 699 (2015).
J. Pfeifle, V. Vujicic, R. T. Watts, P. C. Schindler, C. Weimann,
R. Zhou, W. Freude, L. P. Barry, and C. Koos, "Flexible terabit/s
5.
6.
7.
Nyquist-WDM super-channels using a gain-switched comb
source," Opt Express 23(2), 724 -- 738 (2015).
J. N. Kemal, P. Marin-Palomo, K. Merghem, G. Aubin, C. Calo,
R. Brenot, F. Lelarge, A. Ramdane, S. Randel, W. Freude, and C.
Koos, "32QAM WDM transmission using a quantum-dash
passively mode-locked laser with resonant feedback," in Optical
Fiber Communication Conference Postdeadline Papers (OSA,
2017), paper Th5C.3.
P. Marin-Palomo, J. N. Kemal, M. Karpov, A. Kordts, J. Pfeifle,
M. H. P. Pfeiffer, P. Trocha, S. Wolf, V. Brasch, M. H. Anderson,
R. Rosenberger, K. Vijayan, W. Freude, T. J. Kippenberg, and C.
Koos, "Microresonator-based solitons for massively parallel
coherent optical communications," Nature 546(7657), 274 -- 279
(2017).
8. A. Fülop, M. Mazur, A. Lorences-Riesgo, O. B. Helgason, P.-H.
9.
Wang, Y. Xuan, D. E. Leaird, M. Qi, P. A. Andrekson, A. M.
Weiner, and V. Torres-Company, "High-order coherent
communications using mode-locked dark-pulse Kerr combs from
microresonators." Nat. Commun. 9, 1598 (2018).
P. Marin-Palomo, J. N. Kemal, P. Trocha, S. Wolf, K. Merghem,
F. Lelarge, A. Ramdane, W. Freude, S. Randel, and C. Koos,
"Comb-based WDM transmission at 10 Tbit/s using a DC-driven
quantum-dash mode-locked laser diode," arXiv:1904.11952v2
(2019).
10. J. N. Kemal, P. Marin-Palomo, V. Panapakkam, P. Trocha, S.
Wolf, K. Merghem, F. Lelarge, A. Ramdane, S. Randel, W.
Freude, C. Koos, C. Koos, and C. Koos, "WDM transmission
using quantum-dash mode-locked
laser diodes as multi-
wavelength source and local oscillator," in Optical Fiber
Communication Conference (OSA, 2017), paper Th3F.6.
11. C. Doerr and L. Chen, "Silicon Photonics in Optical Coherent
Systems," Proc. IEEE 106, 2291 -- 2301 (2018).
12. A. Novack, M. Streshinsky, T. Huynh, T. Galfsky, H. Guan, Y.
Liu, Y. Ma, R. Shi, A. Horth, Y. Chen, A. Hanjani, J. Roman, Y.
Dziashko, R. Ding, S. Fathololoumi, A. E.-J. Lim, K. Padmaraju,
R. Sukkar, R. Younce, H. Rohde, R. Palmer, G. Saathoff, T. Wuth,
M. Bohn, A. Ahmed, M. Ahmed, C. Williams, D. Lim, A.
Elmoznine, A. Rylyakov, T. Baehr-Jones, P. Magill, D. Scordo,
and M. Hochberg, "A silicon photonic transceiver and hybrid
tunable laser for 64 Gbaud coherent communication," in Optical
Fiber Communication Conference Postdeadline Papers (OSA,
2018), paper. Th4D.4.
13. R. W. Going, M. Lauermann, R. Maher, H.-S. Tsai, A. Hosseini,
M. Lu, N. Kim, P. Studenkov, S. W. Corzine, J. Summers, M.
Anagnosti, M. Montazeri, J. Zhang, B. Behnia, J. Tang, S.
Buggaveeti, T. Vallaitis, J. Osenbach, M. Kuntz, X. Xu, K.
Croussore, V. Lal, P. Evans, J. T. Rahn, T. Butrie, A. Karanicolas,
K.-T. Wu, M. Mitchell, M. Ziari, D. Welch, and F. Kish, "1.00
(0.88) Tb/s per wave capable coherent multi-channel transmitter
(receiver)
integrated SiGe
electronics," IEEE J. Quantum Electron. 54(4), 1 -- 10 (2018).
InP-based PICs with hybrid
14. L. M. Augustin, R. Santos, E. den Haan, S. Kleijn, P. J. A. Thijs,
S. Latkowski, D. Zhao, W. Yao, J. Bolk, H. Ambrosius, S.
Mingaleev, A. Richter, A. Bakker, and T. Korthorst, "InP-based
generic foundry platform for photonic integrated circuits," IEEE J.
Sel. Top. Quantum Electron. 24, 1 -- 10 (2018).
15. J.-P. Elbers, N. Eiselt, A. Dochhan, D. Rafique, and H. Griesser,
"PAM4 vs coherent for DCI applications," in Advanced Photonics
2017 (IPR, NOMA, Sensors, Networks, SPPCom, PS) (OSA,
2017), paper SpTh2D.1.
16. T. Pfau, H. Zhang, J. Geyer, and C. Rasmussen, "High
performance coherent ASIC," in 2018 European Conference on
Optical Communication (ECOC) (IEEE, 2018), pp. 1 -- 3.
17. F. Lelarge, B. Dagens, J. Renaudier, R. Brenot, A. Accard, F. Dijk,
D. Make, O. L. Gouezigou, J.-G. Provost, F. Poingt, J. Landreau,
O. Drisse, E. Derouin, B. Rousseau, F. Pommereau, and G.-H.
Duan, "Recent advances on InAs/InP quantum dash based
semiconductor lasers and optical amplifiers operating at 1.55 μm,"
IEEE J. Sel. Top. Quantum Electron. 13, 111 -- 124 (2007)
18. R. Rosales, K. Merghem, A. Martinez, A. Akrout, J.-P. Tourrenc,
A. Accard, F. Lelarge, and A. Ramdane, "InAs/InP quantum-dot
passively mode-locked lasers for 1.55-μ m applications," IEEE J.
Sel. Top. Quantum Electron. 17(6), 1292 -- 1301 (2011).
19. R. Rosales, K. Merghem, A. Martinez, A. Akrout, J.-P. Tourrenc,
A. Accard, F. Lelarge, and A. Ramdane, "Timing jitter from the
optical spectrum in semiconductor passively mode locked lasers."
Opt. Express 20(8), 9151-9160 (2012).
20. K. Merghem, C. Calo, R. Rosales, X. Lafosse, G. Aubin, A.
Martinez, F. Lelarge, and A. Ramdane, "Stability of optical
frequency comb generated with InAs/InP quantum-dash-based
passive mode-locked lasers," IEEE J. Quantum Electron. 50, 275 --
280 (2014).
21. T. Pfau, S. Hoffmann, and R. Noe, "Hardware-efficient coherent
digital receiver Concept with feedforward carrier recovery for M-
QAM constellations," J. Light. Technol. 27(8), 989 -- 999 (2009).
22. K. Kikuchi, "Characterization of semiconductor-laser phase noise
and estimation of bit-error rate performance with low-speed
offline digital coherent receivers," Opt. Express 20(5), 5291
(2012).
23. T. N. Huynh, L. Nguyen, and L. P. Barry, "Phase noise
characterization of SGDBR lasers using phase modulation
detection method with delayed self-heterodyne measurements," J.
Light. Technol. 31, 1300 -- 1308 (2013).
24. C. Malouin, P. Thomas, B. Zhang, J. O'Neil, and T. Schmidt,
"Natural expression of the best-match search Godard clock-tone
algorithm for blind chromatic dispersion estimation in digital
coherent receivers," in Advanced Photonics Congress (OSA,
2012), p. SpTh2B.4.
25. D. N. Godard, "Passband timing recovery in an all-digital modem
receiver," IEEE Trans. Commun. 26, 517 -- 523 (1978).
26. C. W. Farrow, "A continuously variable digital delay element," in
IEEE International Symposium on Circuits and Systems, Espoo,
Finland, 2641 -- 2645 (1988).
27. S. J. Savory, "Digital filters for coherent optical receivers," Opt.
Express 16, 804 -- 817 (2008).
28. R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D.
Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C.
Koos, J. Becker, W. Freude, and J. Leuthold, "Error vector
magnitude as a performance measure for advanced modulation
formats," IEEE Photonics Technol. Lett. 24, 61 -- 63 (2012).
Erratum: ibid. 24(23), 2198 (2012).
29. F. Chang, K. Onohara, and T. Mizuochi, "Forward error correction
for 100 G transport networks," IEEE Commun. Mag. 48(3), S48-
S55 (2010).
30. L. Schmalen, S. Brink, and A. Leven, "Spatially-coupled LDPC
protograph codes for universal phase slip-tolerant differential
decoding." in Optical Fiber Communication Conference, OSA
Technical Digest (Optical Society of America, 2015), paper
Th3E.6.
31. M. Magarini, L. Barletta, A. Spalvieri, F. Vacondio, T. Pfau, M.
Pepe, M. Bertolini, and G. Gavioli, "Pilot-symbols-aided carrier-
phase recovery for 100-G PM-QPSK digital coherent receivers,"
IEEE Photonics Technol. Lett. 24, 739 -- 741 (2012).
|
1812.01740 | 1 | 1812 | 2018-12-04T22:59:39 | Time-resolved X-ray diffraction study of the structural dynamics in a ferroelectric thin film induced by sub-coercive fields | [
"physics.app-ph"
] | The electric field-dependence of structural dynamics in a tetragonal ferroelectric lead zirconate titanate thin film is investigated under sub-coercive and above-coercive fields using time-resolved X-ray diffraction. During the application of an external field to the pre-poled thin film capacitor, structural signatures of domain nucleation and growth include broadening of the in-plane peak width of a Bragg reflection concomitant with a decrease of the peak intensity. This disordered domain state is remanent and can be erased with an appropriate voltage pulse sequence. | physics.app-ph | physics |
Time-resolved X-ray diffraction study of the structural dynamics in a ferroelectric
thin film induced by sub-coercive fields
C. Kwamen,1 M. Rossle,1 W. Leitenberger,2 M. Alexe,3 and M. Bargheer2, 4, ∗
1Helmholtz Zentrum Berlin, Albert-Einstein-Str. 15, 12489 Berlin, Germany
2Institut fur Physik & Astronomie, Universitat Potsdam,
Karl-Liebknecht-Str. 24-25, 14476 Potsdam, Germany
3Department of Physics, University of Warwick, Coventry CV4 7AL
4Helmholtz Zentrum Berlin fur Materialien und Energie,
Albert-Einstein-Str. 15, 12489 Berlin, Germany
(Dated: December 6, 2018)
The electric field-dependence of structural dynamics in a tetragonal ferroelectric lead zirconate
titanate thin film is investigated under sub-coercive and above-coercive fields using time-resolved
X-ray diffraction. During the application of an external field to the pre-poled thin film capacitor,
structural signatures of domain nucleation and growth include broadening of the in-plane peak width
of a Bragg reflection concomitant with a decrease of the peak intensity. This disordered domain
state is remanent and can be erased with an appropriate voltage pulse sequence.
Ferroelectrics (FE) are not only technologically inter-
esting because of their electromechanical properties that
enable their application in transducer devices. The re-
versible spontaneous polarization of FE has been used in
memory devices with a reported data retention of about
10 years [1]. FE memory devices make use of the remnant
polarization state obtained after poling or reversing the
FE polarization by an electric field to store the boolean
algebraic logic states "0" and "1". Operating FEs above
their coercive field leads to fatigue, which limits device
lifetime [2 -- 6]. Also, the quest for low power consumption
requires operating such devices under the lowest possible
bias. Hence there is growing interest in sub-coercive field
applications and associated remnant states [7 -- 11].
It is well known that FE thin films do not reverse their
polarity as a whole when an external field is applied. Af-
ter domain nucleation, the regions with opposite polar-
ization are separated by domain walls (DW) [12, 13]. One
of the domain polarization directions prevails as the ex-
ternal field approaches the saturation field, however, the
time required to fully suppress domains with the opposite
polarization is determined by the domain wall velocity.
The DWs are mobile, add structural disorder within the
sample volume, and reduce the observed maximum po-
larization. DWs contribute to the observed piezoelectric
strain in FE devices already at very low fields [14, 15].
This has lead to new devices in magnetic and FE ma-
terials for DW logic and DW diode applications [16 -- 18].
It has recently been demonstrated that multiple memory
states can be created due to the coupling between rem-
nant strain and domain states [7, 19].
Some actuator applications require fast actuation with
reproducible length changes that are not affected by
drifts on longer time-scale. To study the time-dependent
response of FE ceramics, time-resolved X-ray diffraction
(TR-XRD) has been used to probe the dynamics under
electrical loading, offering the possibility to distinguish
between the intrinsic piezoelectric response of individ-
ual domains and extrinsic contributions originating from
changes of the volume fraction of the domains. [14, 15].
In the last years, in-situ synchrotron XRD has been used
to quantify the electromechanical response and fatigue
behavior of FE thin films and powders under loading,
with the advantage of yielding appropriate temporal and
spatial resolution as compared to conventional X-ray se-
tups [20 -- 23]. In-situ high-resolution XRD has for exam-
ple been used to study the switching dynamics of 90◦
domains in epitaxial PZT [24]. More recent measure-
ments aim at combining laser excitation and electrical
excitation for x-ray structural dynamics investigations
[25]. Several synchrotron-based experiments reported in-
tensity and position changes of the diffraction maxima in
PZT thin films, however, electrical characterization was
studied independently [26, 27]. The interpretation of very
large intensity changes up to 40% for the two polarization
states solely ascribed to the anomalous scattering must
be reconsidered since contributions from disorder cannot
be neglected. We have recently reported the structural
dynamics accompanying the reproducible repeated elec-
trical switching of FE thin films, relating the observations
to the simultaneously measured electrical charging with
saturation fields of opposite polarity [28].
In the present TR-XRD study of a 250 nm thin
Pb(Zr0.2Ti0.8)O3 (PZT) FE capacitor, we compare the
time-dependent structural dynamics during sub-coercive
field pulses to the full switching dynamics under satu-
ration fields. We present the field- and time-dependent
lattice response, infer the domain dynamics, and demon-
strate the presence of remnant disordered states. The ex-
periments were realized on a FE capacitor composed of
PZT as FE layer and a metallic SrRuO3 (SRO) electrode
epitaxially grown on a (001) oriented SrTiO3 (STO) sub-
strate by pulsed laser deposition. Platinum top elec-
trodes of hexagonal shape with 0.3 mm edge length and
a thickness of about 30 nm were deposited onto the PZT
film by sputtering. The sample structure is sketched in
the inset of Figure 1. An external field was applied be-
tween one Pt electrode and the SRO layer along the [001]
direction, which is the FE polarization axis of PZT. A
pulse generator (Keithley 3390) was used to apply volt-
age pulses up to ±8 V with the help of a tungsten nee-
dle [28]. X-rays of 9 keV photon energy provided at the
XPP-KMC3 endstation at the storage ring BESSY II at
the Hemholtz Zentrum in Berlin were employed for the
TR-XRD. The experiments were performed in the so-
called multibunch operation mode [29, 30]. The X-ray
beam was focused onto the sample surface such that only
the region under one electrode was probed. For recording
full reciprocal space maps (RSM) a PILATUS 100K pixel
detector (DECTRIS) was used. To obtain time-resolved
signals, the detector gate was triggered appropriately by
the electric field pulse sequence.
In order to increase
the acquisition speed, ω scans through the RSM were
recorded with a home-built detector made of a photo-
multiplier tube (Hamamatsu) and a fast scintillator with
<1 ns response time, which collects an angular diver-
gence of the diffracted x-rays of 0.9◦, which exceeds the
width of the RSM in either angular direction by a factor
of three. This detection geometry is often referred to as
rocking curve, an established method to determine lattice
plane spacings in mosaic crystals via Bragg's law [31 -- 33].
The typical full-width-at-half-maximum (FWHM) of the
rocking curves was 0.3◦.The X-ray signal is sent to a time-
correlated single photon counting system (PicoHarp 300),
which allows reconstructing the rocking curves at differ-
ent delays by synchronizing the detector trigger with the
rising edge of the electrical pulse. [30] The high resolu-
tion X-ray diffraction measurement showed that the film
is c-axis oriented [28]. The reciprocal space map without
applied external field was reported in a recent publica-
tion for a different electrode of the same sample [28].
The excellent FE properties of the device with an RC
time constant of about 1 µs and a coercivity of about
4.5 V have been discussed in ref. [28]: Polarization rever-
sal observed by electronic measurements was shown to be
accompanied by a intensity changes of about ∆I = 5%
resulting form difference in structure factor between the
(002) and (00¯2) Bragg reflections [28]. This is consistent
with a relative motion of the Ti and Zr atoms relative
to the Oxygen octahedra of ∆ξPb-Ti = 0.0187 nm and
∆ξTi-O = 0.03 nm along the c-axis [34]. Now we apply
an asymmetric square pulse sequence as shown in Fig-
ure 1(a) to a test device and analyze the dynamics of
the 002 PZT reflection. The strain η = ∆c(t)/c(t = 0)
corresponding to the relative c-axis change is shown in
Figure 1(b). In standard models of polarization switch-
ing in thin films under saturation fields [12, 13, 35 -- 37],
the polarization reversal is mediated by the nucleation
of new domains with reversed polarization, their subse-
quent propagation across the film thickness, and a sub-
sequent lateral motion of DWs perpendicular to the field
direction that finally leads to a merging of the reversed
2
Figure 1. Time-dependence of c-axis strain η, normalized
peak width, w, and normalized integrated intensity, I, during
sub-coercive field loading. (a) Applied waveform, (b) out-of-
plane lattice response to the applied field, (c) angular broad-
ening of the measured peak, and (d) corresponding integrated
intensity. The inset schematically shows the sample structure,
the application of the voltage, and scattering geometry.
domains. The structural dynamics observed in our sam-
ple for two different voltages, U , are shown in Figure 1.
The pulses marked P1 and P4 are the poling pulses used
to initialize the film's polarization state in a reproducible
manner. Pulse P2 and P3 are chosen to be symmetrical
and below the coercive field Uc. Note that the time t = 0
is defined as onset of pulse P2, since we are interested in
the sub-coercive field dynamics.
We first discuss the darkblue lines in Figure 1 represent-
ing an electric field of -2 V, well below Uc of the device.
The PE response to pulse P2 is purely compressive (Fig-
ure 1b), indicating that the majority of the thin film has
retained its initial polarization. The relative peak width
w = (w(t) − w(t = 0))/w(t = 0) represented in Fig-
ure 1(c) increases slowly and reaches a saturation value
after ∼ 150 µs with an overall width change of ∼ 20%. We
also note that the value of w remains constant even after
switching the external field off at about 200 µs. Simul-
taneously, the normalized integrated intensity I shown
in Figure 1(d) decreases by ∼ 7.5% of its initial value.
−505(a)P1P2P3P4U(V)012(b)η·10311.11.2(c)w−20002004006000.90.951(d)t(µs)I 3
In Figure 2 we concentrate on the temporal sample re-
sponse during Pulse 2 with various field amplitudes. In
Figure 2(a), the different applied fields are plotted as
function of time after the application of the pulse. Be-
fore each of these pulses were applied, the poling pulse
P1 from Figure 1(a) with opposite polarity and ampli-
tude 8 V set the initial condition of the sample. We var-
ied the amplitude of pulse P2 from −2 to −6 V and kept
its duration constant. For the experiment at −8 V we re-
duced the pulse length to avoid degradation of the sam-
ple. The X-ray analysis of the structural dynamics of
the PZT film is plotted in Figures 2(b-d). Figure 2(b)
confirms the PE contraction of the film for −2 V which
can now be compared to the much larger inverse piezo-
effect obtained under larger fields. In addition, we would
like to highlight the steeper and steeper gradient of the
expansion dynamics for larger fields. The domain nucle-
ation and propagation is more efficient and leads to a
more rapid positive PE expansion. The initial increase
of w of about 20% shown in Figure 1(c) indicates disor-
der in the film that increases by domain nucleation and
propagation. For fields U2 ≥ 3 V w decreases again as
domains merge and grow, thus decreasing the disorder.
For U2 = −3 V some disorder remains when pulse P2
comes to an end and this disorder remains under zero
field for t > 200µs. For U2 = −2 V the pulse ends in
the state of maximum peak width, and also this signa-
ture of large domain disorder is remnant. In fact, a small
increase of the peak width remains after switching off
Pulse P2 even for the largest applied field. This proba-
bly indicates a slight asymmetry of the film's structural
properties. For the high field of −6 V and −8 V the nor-
malized diffracted intensity plotted in Figure 2(d) con-
firms the expected structure factor change for full po-
larization reversal. However, for low field switching, the
intensity decreases by more than the ∆I = 5% expected
for the structure factor change. The polarization is only
reversed in a small fraction of the sample, and hence
it is not the structure factor but rather disorder associ-
ated with inhomogeneous domain nucleation and motion
that induces the intensity reduction. Now we extract the
significant signatures of sub-coercive field induced polar-
ization reversal from Figure 2(b-d) and plot these signa-
tures as a function of the field U in Figure 3. Each of the
strain transients above U2 ≥ 3 V in Figure 2(b) shows
a pronounced kink where the rate of expansion, dη/dt,
changes. It occurs at the time tkink which shows a very
similar field dependence as the time twmax at which the
maximum of the peak width is reached. This indicates
maximum structural disorder and the point where the
reversed domains start merging. This time precedes the
time tkink of the sudden increase of the lattice constant
by a constant factor of tkink/twmax = 2. Figure 3 also
reports the initial rate dw/dt at which the peak width
changes. In order to emphasize the identical voltage de-
pendence we plot the time t∆w=10% where the peak width
Figure 2. Time-dependence of the c-axis strain, η, the nor-
malized peak width, w, and normalized integrated intensity I
under switching pulses of varying amplitude U . (a) Schematic
of the applied waveform, (b) strain due to the applied field as
a function of time, (c) normalized FWHM from the recorded
rocking curves, and (d) corresponding integrated intensity
normalized to its value at t = 0.
The continuous change of w at constant field shows that
domains nucleate and move within the FE film towards
a state with approximately equal amounts of both po-
larization states as this increases the inhomogeneity of
the film. Pulse P3 with opposite polarity is applied af-
ter a short field-free time and recovers w and I, thereby
resetting the domain state in the film. The stability of
w during the time between the electrical pulses, i.e. at
zero fields, proves that the disordered domain configura-
tion induced by sub-coercive fields is remnant. In con-
trast, the small PE c-axis contraction, which is induced
by pulse P2, relaxes back to its zero-field value. Pulse P3
with opposite polarity then leads to an expansion with
similar amplitude, which is consistent with the small ap-
plied field below Uc, which leaves the majority of the FE
film in the initial polarization state. We now discuss the
blue lines in Figure 1 that represent a somewhat larger
electric field pulse with U2 = −3V, which is just below Uc
for a standard hysteresis loop measured at 1 kHz. Now
a PE expansion is observed during both pulse P2 and
P3, indicating that a bit more than 50% of the thin film
have reversed polarization during pulse P2, which is then
switched back by pulse P3. Thus for this pulse sequence,
the coercive field is slightly below 3 V.
−8−6−4−20(a)U(V)−2V−3V−4V−5V−6V−8V012(b)η·10311.11.21.3(c)w0501001502000.90.951(d)t(µs)I 4
Figure 3. Maximum strain ηmax at t = 200 µs as a measure of
the FE polarization as function of the applied voltage U . The
entire dynamics speed up with increasing voltage: We com-
pare the times tw10% it takes to initially increase the peak
width by 10% and the twmax (blue solid squares) character-
izing the time for the maximum peak width to tkink (open
squares) at which the strain η suddenly increases (filled black
symbols). The times have been scaled by appropriate param-
eters given in the legend.
has increased by 10% (full black circles) as a measure
of the inverse rate. There is another constant factor of
tkink/tw10% = 8.
We would like to emphasize that the different remnant
values of w after 200 µs define domain states that are
stable as long as no electric fields are applied. This dis-
order could be used for multi-level memory applications
based on remnant domain configurations [7]. Such do-
main states can be created by the pulse sequence dis-
cussed above.
In the following, we demonstrate that
these states can be erased by using an appropriate pulse
sequence in order to define a reference state for multi-
level memory applications. A simple "erasing" pulse se-
quence is indicated in the inset of Figure 4(a). We mon-
itored this reference domain structure and the changes
induced by the writing pulse shown in Figure 4(a) with
8 V amplitude by reciprocal space mapping (RSM). Fig-
ure 4(e) visualizes the structural changes directly in the
RSM for selected time delays. Figure 4(b) shows the
peak shift, which is only present while the electric field is
applied. Figure 4(c) indicates the normalized width w⊥
along qz indicating an inhomogeneous electric field dis-
tribution across the electrode area.[28] This peak width
change disappears after the external field is switched off.
In contrast, the large FWHM of the in-plane component
of the Bragg peak w(cid:107) after the "erasing" sequence indi-
cates a high density of small domains, which is rapidly
reduced by the external field pulse. This ordering of the
domain structure again is remnant and states of differ-
ent degrees of disorder are stable states in the thin film
device.
We demonstrated that states of remnant ferroelectric
Figure 4. Structural response to an above-coercive field pulse
directly after the "erasing" pulse sequence shown in the in-
sert of panel a). (a) Applied pulse, (b) strain obtained form
the change of the out-of-plane lattice constant. (c) Normal-
ized FWHM of the out-of-plane Bragg peak. (d) Normalized
FWHM of the corresponding scattering vector parallel to film
(e) RSMs at selected times. The width and the
surface.
−1
height of the displayed RSM corresponds to w⊥ = 0.09A
and w(cid:107) = 0.07A
−1 respectively.
domain disorder can be written in a nanometric PZT thin
film device at sub-coercive fields. The structural changes
of the domain states were observed by time-resolved re-
ciprocal space mapping with hard X-rays, looking at a
working device that was operated for more than 108
switching cycles during these measurements. This sub-
coercive field switching could be used in low power ferro-
electric memory applications operating at low field. We
showed that the intensity of Bragg reflections changes
during the switching not only because the structure fac-
tor changes but also because the domain disorder broad-
ens the Bragg peaks.
We acknowledge the BMBF for the financial support
via grant No. 05K16IPA. M.A. acknowledges support
of Royal Society through the Wolfson Research Merit
and Theo Murphy Blue-sky Awards and EPSRC (UK)
through grants No. EP/P031544/1 and EP/P025803/1.
−8−6−4−20.511.522.5U(V)η·103−8−6−4−20100200300400U(V)tkink(µs)ηmaxtkink2∗twmax8∗t∆w=10% ∗ [email protected];
http://www.udkm.physik.uni-potsdam.de
[1] J. F. Scott and C. A. Paz de Araujo, Science 246, 1400
(1989), ISSN 0036-8075, URL http://www.sciencemag.
org/cgi/doi/10.1126/science.246.4936.1400.
[2] H. M. Duiker, P. D. Beale, J. F. Scott, C. A. Paz
de Araujo, B. M. Melnick, J. D. Cuchiaro, and L. D.
McMillan, Journal of Applied Physics 68, 5783 (1990),
ISSN 0021-8979, URL http://aip.scitation.org/doi/
10.1063/1.346948.
[3] W. Yi, K. H. Wong, and W. Wen-Bin, Chinese Physics
Letters 19, 566 (2002), URL http://stacks.iop.org/
0256-307X/19/i=4/a=336.
[4] Z. Luo, S. Pojprapai, J. Glaum, and M. Hoffman, Jour-
nal of the American Ceramic Society 95, 2593 (2012),
ISSN 00027820, URL http://doi.wiley.com/10.1111/
j.1551-2916.2012.05232.x.
[5] J. Shieh, J. E. Huber, and N. A. Fleck, Jour-
nal of the European Ceramic Society 26, 95 (2006),
ISSN 09552219, URL http://linkinghub.elsevier.
com/retrieve/pii/S0955221904004820.
[6] Q. Jiang, W. Cao, and L. E. Cross, Journal of the
American Ceramic Society 77, 211 (1994), ISSN 0002-
7820, URL https://doi.org/10.1111/j.1151-2916.
1994.tb06979.x.
[7] B. Kundys, V. Iurchuk, C. Meny, H. Majjad, and
B. Doudin, Applied Physics Letters 104, 232905 (2014),
ISSN 0003-6951, URL https://doi.org/10.1063/1.
4883375.
[8] R. Yimnirun, N. Wongdamnern, N. Triamnak, M. Un-
ruan, A. Ngamjarurojana, S. Ananta, and Y. Laosirita-
worn, Journal of Applied Physics 103, 086105 (2008),
ISSN 0021-8979, URL http://aip.scitation.org/doi/
10.1063/1.2907755.
[9] A. Pramanick, D. Damjanovic, J. E. Daniels, J. C. Nino,
and J. L. Jones, Journal of the American Ceramic Soci-
ety 94, 293 (2011), ISSN 00027820, URL http://doi.
wiley.com/10.1111/j.1551-2916.2010.04240.x.
[10] J. L. Jones, E. Aksel, G. Tutuncu, T.-M. Usher, J. Chen,
X. Xing, and A. J. Studer, Physical Review B 86, 024104
(2012), ISSN 1098-0121, URL https://link.aps.org/
doi/10.1103/PhysRevB.86.024104.
[11] N. Wongdamnern, N. Triamnak, M. Unruan, K. Kanchi-
ang, A. Ngamjarurojana, S. Ananta, Y. Laosiritaworn,
and R. Yimnirun, Physics Letters A 374, 391 (2010),
ISSN 03759601, URL http://dx.doi.org/10.1016/j.
physleta.2009.11.019.
[12] W. J. Merz, Physical Review 95, 690 (1954), ISSN
0031-899X, URL https://link.aps.org/doi/10.1103/
PhysRev.95.690.
[13] D. J. Kim, J. Y. Jo, T. H. Kim, S. M. Yang, B. Chen,
Y. S. Kim, and T. W. Noh, Applied Physics Letters 91,
132903 (2007), ISSN 0003-6951, 0707.3674, URL http:
//aip.scitation.org/doi/10.1063/1.2790485.
[14] A. Pramanick, D. Damjanovic, J. C. Nino, and J. L.
Jones, Journal of the American Ceramic Society 92, 2291
(2009), ISSN 00027820, URL http://doi.wiley.com/
10.1111/j.1551-2916.2009.03218.x.
[15] A. Pramanick, J. E. Daniels, and J. L. Jones, Jour-
nal of the American Ceramic Society 92, 2300 (2009),
ISSN 00027820, URL http://doi.wiley.com/10.1111/
5
j.1551-2916.2009.03219.x.
[16] D. A. Allwood, G. Xiong, and R. P. Cowburn, Applied
Physics Letters 85, 2848 (2004), ISSN 0003-6951, URL
https://doi.org/10.1063/1.1802388.
[17] D. A. Allwood, Science 309, 1688 (2005), ISSN 0036-
8075, URL http://www.sciencemag.org/cgi/doi/10.
1126/science.1108813.
[18] J. Whyte and J. Gregg, Nature Communications 6, 7361
(2015), ISSN 2041-1723, URL http://dx.doi.org/10.
1038/ncomms8361.
[19] W. Zhou, Y. Xiong, Z. Zhang, D. Wang, W. Tan, Q. Cao,
Z. Qian, and Y. Du, ACS Applied Materials & Interfaces
8, 5424 (2016), ISSN 1944-8244, URL http://pubs.acs.
org/doi/10.1021/acsami.5b11392.
[20] A. Davydok, T. Cornelius, C. Mocuta, E. Lima,
E. Araujo, and O. Thomas, Thin Solid Films 603,
29 (2016), ISSN 00406090, URL http://linkinghub.
elsevier.com/retrieve/pii/S0040609016000675.
[21] D.-H. DO, A. GRIGORIEV, D. M. KIM, C.-B. EOM,
P. G. EVANS, and E. M. DUFRESNE, Integrated Ferro-
electrics 101, 174 (2008), ISSN 1058-4587, URL https:
//doi.org/10.1080/10584580802470975.
[22] S. Gorfman, H. Choe, V. V. Shvartsman, M. Ziolkowski,
M. Vogt, J. Strempfer, T. (cid:32)Lukasiewicz, U. Pietsch, and
J. Dec, Physical Review Letters 114, 097601 (2015),
ISSN 0031-9007, URL http://link.aps.org/doi/10.
1103/PhysRevLett.114.097601.
[23] T. W. Cornelius, C. Mocuta, S. Escoubas, A. Merabet,
M. Texier, E. C. Lima, E. B. Araujo, A. L. Kholkin,
and O. Thomas, Journal of Applied Physics 122, 164104
(2017), ISSN 0021-8979, URL http://aip.scitation.
org/doi/10.1063/1.4994939.
[24] J. L. Jones, M. Hoffman, J. E. Daniels, and A. J. Studer,
Applied Physics Letters 89, 092901 (2006), ISSN 0003-
6951, URL http://aip.scitation.org/doi/10.1063/
1.2338756.
[25] H. Akamatsu, Y. Yuan, V. A. Stoica, G. Stone,
T. Yang, Z. Hong, S. Lei, Y. Zhu, R. C. Haislmaier,
J. W. Freeland,
et al., Phys. Rev. Lett. 120,
096101 (2018), URL https://link.aps.org/doi/10.
1103/PhysRevLett.120.096101.
[26] A. Grigoriev, D.-H. Do, D. M. Kim, C.-B. Eom,
B. Adams, E. M. Dufresne, and P. G. Evans, Phys. Rev.
Lett. 96, 187601 (2006), URL https://link.aps.org/
doi/10.1103/PhysRevLett.96.187601.
[27] A. Grigoriev, R. J. Sichel, J. Y. Jo, S. Choudhury, L.-
Q. Chen, H. N. Lee, E. C. Landahl, B. W. Adams,
E. M. Dufresne, and P. G. Evans, Phys. Rev. B 80,
014110 (2009), URL https://link.aps.org/doi/10.
1103/PhysRevB.80.014110.
[28] C. Kwamen, M. Rossle, M. Reinhardt, W. Leitenberger,
F. Zamponi, M. Alexe, and M. Bargheer, Physical Re-
view B 96, 134105 (2017), ISSN 2469-9950, URL https:
//doi.org/10.1103/PhysRevB.96.134105.
[29] K. Holldack, R. Ovsyannikov, P. Kuske, R. Muller,
A. Schalicke, M. Scheer, M. Gorgoi, D. Kuhn, T. Leit-
ner, S. Svensson, et al., Nature Communications 5, 4010
(2014), ISSN 2041-1723, URL http://www.nature.com/
articles/ncomms5010.
[30] H. Navirian, R. Shayduk, W. Leitenberger, J. Gold-
shteyn, P. Gaal, and M. Bargheer, Review of Scientific
Instruments 83, 063303 (2012), ISSN 0034-6748, URL
https://doi.org/10.1063/1.4727872.
[31] J. Ayers, Journal of Crystal Growth 135, 71 (1993),
ISSN 00220248, URL https://www.sciencedirect.
com/science/article/pii/0022024894907277.
[32] J. Seraf´ınczuk, Crystal Research and Technology 51, 276
(2016), ISSN 02321300, URL http://doi.wiley.com/
10.1002/crat.201500309.
[33] K. Kobayashi, A. A. Yamaguchi,
S. Kimura,
H. Sunakawa, A. Kimura, and A. Usui, Japanese Journal
of Applied Physics 38, L611 (1999), ISSN 00214922,
URL http://stacks.iop.org/1347-4065/38/L611.
[34] B. Noheda, J. A. Gonzalo, L. E. Cross, R. Guo, S.-E.
Park, D. E. Cox, and G. Shirane, Physical Review B 61,
8687 (2000), ISSN 0163-1829, URL https://doi.org/
10.1103/PhysRevB.61.8687.
[35] M. Dawber, K. M. Rabe, and J. F. Scott, October
77, 1083 (2005), 0503372, URL http://dx.doi.org/10.
1103/RevModPhys.77.1083.
[36] Y. W. So, D. J. Kim, T. W. Noh, J.-G. Yoon, and
T. K. Song, Applied Physics Letters 86, 092905 (2005),
ISSN 0003-6951, URL http://aip.scitation.org/doi/
10.1063/1.1870126.
[37] E. Fatuzzo and W. J. Merz, Physical Review 116, 61
(1959), ISSN 0031-899X, URL https://link.aps.org/
doi/10.1103/PhysRev.116.61.
6
|
1810.01216 | 3 | 1810 | 2018-10-06T12:31:01 | All-optical Nonlinear Activation Function for Photonic Neural Networks | [
"physics.app-ph",
"cond-mat.dis-nn",
"physics.optics"
] | With the recent successes of neural networks (NN) to perform machine-learning tasks, photonic-based NN designs may enable high throughput and low power neuromorphic compute paradigms since they bypass the parasitic charging of capacitive wires. Thus, engineering data-information processors capable of executing NN algorithms with high efficiency is of major importance for applications ranging from pattern recognition to classification. Our hypothesis is therefore, that if the time-limiting electro-optic conversion of current photonic NN designs could be postponed until the very end of the network, then the execution time of the photonic algorithm is simple the delay of the time-of-flight of photons through the NN, which is on the order of picoseconds for integrated photonics. Exploring such all-optical NN, in this work we discuss two independent approaches of implementing the optical perceptrons nonlinear activation function based on nanophotonic structures exhibiting i) induced transparency and ii) reverse saturated absorption. Our results show that the all-optical nonlinearity provides about 3 and 7 dB extinction ratio for the two systems considered, respectively, and classification accuracies of an exemplary MNIST task of 97% and near 100% are found, which rivals that of software based trained NNs, yet with ignored noise in the network. Together with a developed concept for an all-optical perceptron, these findings point to the possibility of realizing pure photonic NNs with potentially unmatched throughput and even energy consumption for next generation information processing hardware. | physics.app-ph | physics | ALL-OPTICAL NONLINEAR ACTIVATION FUNCTION FOR
PHOTONIC NEURAL NETWORKS
MARIO MISCUGLIO1, ARMIN MEHRABIAN1, ZIBO HU1, SHAIMAA I. AZZAM2,
JONATHAN GEORGE1, ALEXANDER V. KILDISHEV2, MATTHEW PELTON3,
VOLKER J. SORGER1*
1 Department of Electrical and Computer Engineering, George Washington University, Washington,
DC 20052, USA
2 School of Electrical & Computer Engineering and Birck Nanotechnology Center, Purdue University,
West Lafayette, Indiana, 47907, USA
3 Department of Physics, UMBC (University of Maryland, Baltimore County), Baltimore, Maryland
21250, USA
With the recent successes of neural networks (NN) to perform machine-learning tasks,
photonic-based NN designs may enable high throughput and low power neuromorphic
compute paradigms since they bypass the parasitic charging of capacitive wires. Thus,
engineering data-information processors capable of executing NN algorithms with high
efficiency is of major importance for applications ranging from pattern recognition to
classification. Our hypothesis is therefore, that if the time-limiting electro-optic
conversion of current photonic NN designs could be postponed until the very end of the
network, then the execution time of the photonic algorithm is simple the delay of the
time-of-flight of photons through the NN, which is on the order of picoseconds for
integrated photonics. Exploring such all-optical NN, in this work we discuss two
independent approaches of implementing the optical perceptron's nonlinear activation
function based on nanophotonic structures exhibiting i) induced transparency and ii)
reverse saturated absorption. Our results show that the all-optical nonlinearity provides
about 3 and 7 dB extinction ratio for the two systems considered, respectively, and
classification accuracies of an exemplary MNIST task of 97% and near 100% are found,
which rivals that of software based trained NNs, yet with ignored noise in the network.
Together with a developed concept for an all-optical perceptron, these findings point to
the possibility of realizing pure photonic NNs with potentially unmatched throughput
and even energy consumption for next generation information processing hardware.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Introduction
1.
The past decades have been marked by an exponential increase in demand for high speed and
energy efficient computer architectures. The established microelectronics technology faces its
biggest limitations with respect to handling real-time processing, large data volumes and
complex systems. The exponential growth in transistor count and integration, which marked
the previous century, cannot be pushed further, highlighting the twilight of Moore's law for
CMOS technology. In this changing landscape, multiple companies and research groups are
pursuing novel solutions to provide high-performance computing to the next level, using
better performing algorithms and novel technological approaches, where the work presented
here falls under the latter. A widely investigated solution is to replace general-purpose
processors (von Neumann architecture) with more specialized and task-specific processors.
Graphic Process Units (GPUs) or Field Programmable Gate Array (FPGAs), architectures,
which are optimized for data computation and parallelism provide efficiencies for their
specialized tasks up to an order of magnitude higher than generic-task CPUs [1-3]. However,
these approaches still rely on electronic transport and are bound by the speed and power
limits of the interconnects inside the circuits due to RC parasitic effects.
A subclass of algorithms which receives
benefits when implemented in non von Neumann
architecture is represented
by Neural Networks (NN). NNs are designed to simultaneously
process large arrays of data in order to learn representations of them with multiple levels of
abstraction.
These architectures consist of neurons that perform
the following basic
functions: 1)
Interpret multiple incoming
signals
in a form of multiple arrays (e.g. images)
through weighted addition (Multiply and Accumulate, MAC); 2) Apply a
nonlinear (NL)
activation function for discriminating the data; 3) Transmit
the result to multiple destination
neurons (i.e., fan-out).
Recently emerging Photonic Neural Networks (PNN) demonstrated the potential to increase
computing speed by 2-3 orders of magnitude [4]. In order to implement the functions of the
NN into a PNN, two classes of devices and their respective functions need to be engineered,
the weighted sum and the NL activation. The weighted sum, as previously investigated [5],
can be obtained by a combination of balanced photodetection, Mach-Zehnder interferometer,
or ring resonator, which constitutes the weights of the signal, and y-junctions or MMI
(Multimode interference coupler) for summation. The weighted addition in analog photonics,
which is the equivalent of the MAC, uses optical interference, while the coherent
electromagnetic waves propagate through the photonic integrated circuit (PIC), and offers
MAC energy consumption that does not trade-off with MAC speed [6]. Moreover, photonic
interconnects can have large bandwidth and low power consumption.
On the other hand, the absence of a straightforward and efficient NL in optics has severely
limited its usefulness in DeepLearning computing. Several attempts [7-9] showed that a NL
modulation of the optical signal could be achieved through Electro-optic Absorption
Modulators (EOMs). These rely on the on-chip implementation of an electronic platform that
is able to vary the optical effective mode index by tuning the voltage. Despite their relatively
straightforward implementation and controllability, EOM results in substantial trade-offs in
terms of speed and power efficiency of the otherwise intrinsically instantaneous transmission
of the signal through the PIC.
Other approaches [10] consist of devices that would need first to convert an optical signal into
an electrical signal by means of a detection mechanism and afterwards convert it back into an
optical signal, a solution that would hamper the speed and cascadability of the network due to
both the movement of charge carriers and noise (e.g. shot and thermal noise of a
photodetector). For this reason, the implementation of fast and energy efficient all-optical
nonlinearity becomes a key task for boosting the throughput of a neural network and
consequently lowering the latency and power dissipation.
In this work, we aim to overcome these limitations by implementing novel devices and
approaches based on light matter interaction (LMI) in assembly of nanoparticles in PICs. A
first device prototype relies on a reversible transparency induced by a Fano resonance in a
plasmon-exciton system. This system consists of a semiconductor nanocrystal, i.e. quantum
dot (QD), within two gold nanorods. We also studied the NL response induced by reverse
saturable absorption in films of buckyballs (C60). The optical response of the two hybrid
systems shows a strong non-linearity suitable for being employed as activation functions in
PNN circuits as investigated here. We find the optimal configurations that optimized the
mode coupling between the signal transmitted in a waveguide and the hybrid structures using
numerical tools. Finally, a NN architecture that exploits the proposed activation functions has
been emulated on an open source machine-learning framework, i.e. Tensorflow. We then
compared the performances in terms of speed and accuracy of the proposed system with
others, which employ conventional nonlinear functions (e.g. Tanh, Sigmoid, ReLu), for a
specific DeepLearning tasks. Ultimately, this work points towards further use of NL
activation functions of optically triggered devices, integrated with silicon photonics, for the
next generation of photonic processors.
2. Discussion
As a first step in our study we investigate the LMIs in hybrid systems consisting of a pair of
gold nanorods separated by a 10-nm gap that contains a single CdSe Quantum Dot (QD) Fig.
1a. Despite the apparent complexity, these quantum-assemblies could be either fabricated
through a bottom-up approach, i.e. colloidal synthesis, as well as fabricated through a
combination of top-down and bottom-up approach, by combining electron beam lithography
and guided colloidal deposition [11]. As reported before [12-15], and illustrated in Fig. 1a, the
hybrid system could be modeled as two coupled oscillators, one representing the dipole of the
QD transition, and the other representing the dipole of the plasmon. The surface plasmon is
driven by the incident field, and the QD is driven due to its coupling with the plasmon.
Destructive interference between the two oscillators leads to cancelation of the optical
response of the coupled system, known as an induced transparency or a Fano resonance.
a)
a)
c)
c)
Excited state
FSP
3>
b)
b)
1>
Ground state
2>
Metastable state
d)
d)
)
2
m
c
1
1
-
0
1
(
n
o
i
t
c
e
s
s
s
o
r
c
n
o
i
t
c
n
i
t
x
E
8
6
4
2
l
i
d
e
F
c
i
r
t
c
e
MNPs
QD
20 nm
l
E
d
e
z
i
l
a
m
r
o
N
MAX
0
0
-2
k
,
x
e
d
n
I
e
v
i
t
c
a
r
f
e
R
-4
-6
107
102
log10
103
106
Input Power Density (W/cm-2)
104
105
Fig.1. Physical modeling and material characterization: a) Phenomenological description system. Two
coupled oscillators that provides a classical analogue of the plasmon-exciton coupling induced
transparency in the represented system and schematic representation of the MNPs/QD system
consisting of a single quantum dot (QD) between a pair of gold nanoparticles (MNP). b) Normalized
electric field distribution of the hybrid system computed for a 2.04 eV impinging plane wave. Scale bar
20 nm. c) Calculated absorption spectra for a system illustrated in the inset of (a) taken from [16]. The
different curves show the response of the system for different amounts of energy in an incident laser
pulse. d) Extinction cross section (left y-axis) and imaginary part of the refractive index (right y-axis)
of the MNPs/QD system, as function of the input power density.
Figure 1b displays the normalized electric near field distribution computed at 2.04 eV, in
resonance with the gold rods of the proposed MNPs/QD system. The model parameters for
the metal MNPs/QD system are obtained from FDTD simulations [14]. The complex
refractive index of Gold and CdSe QD were obtained from refs [17] and [18]. The large
electric field in the feed-gap of a gold optical antenna [19], generated by the plasmonic dipole
resonances of the nano-rods, completely engulfs the CdSe QD, inducing the coupling between
the plasmon and the QD that ultimately gives rise to the Fano resonance, which shows the
obtained absorption spectra of the hybrid system for different incident optical powers, taken
from ref [14] (Fig. 1c). The Fano resonance, or transparency dip, is apparent for low incident
optical power; this linear response has recently been confirmed by experimental results [12].
Fig. 2. Hybrid-Nanoparticle (NP) waveguide integration for an all-optical nonlinearity for
photonic neural networks: a-c) Schematic representation of waveguide and MNPs/QD system
coupling. The MNPs/QD system placed in the middle (a) and top (c). b-d) Normalized electric
field distribution for a middle horizontal cut plane of the waveguide considering the assembly
being placed in the middle (b) and on top (d), for its maximum absorption (highest k). c)
Computed transmittance of the waveguide as function of the tunable absorption of a MNPs/QD
system placed in the middle (Black solid line) and on top of the waveguide (Red solid line). d)
Computed waveguide transmittance as function of the input power density and respective
nonlinear modulation ranges in dB.
As the incident power increases, the Fano resonance dip disappears, due to saturation of the
QD transition. Thus, the amount of energy dissipation in the coupled system is a NL function
of the input power. This nonlinearity can be seen in the extinction cross section (left y-axis)
spectra of the assembly computed at 2.04eV, as function of the input power density (Fig. 1d),
suggesting a narrow spectral operation band. In order to later embed the system in a photonic
waveguide, the imaginary part of the refractive index of the assembly as function of the input
power density, was derived using a simplified geometry and Mie Scattering model [20] (right
y-axis), which manifests a specular behavior.
We proceed further in our study by embedding the MNPs/QD system in a silicon photonic
waveguide and studying their interaction through a finite-difference time-domain (FDTD)
solver. The previously described physical model was investigated in two different device
configurations, placing the assembly in the middle and on top of the waveguide, as
schematized in Fig. 2 a and c, respectively. When placed on top, a 50 nm thin gold layer was
used for enhancing the mode overlap between the assembly and the transmitted waveguide
signal.
Fig. 2b and d, maps the normalized electric field travelling within the waveguide in the
horizontal plane. Here, the absorbance of the assembly is set to its maximum; i.e., the
imaginary part of the refractive index, k, is maximized. In this case, the overall transmitted
power is similar for both configurations. As shown in Fig. 2e-f, we evaluate these
configurations by analyzing the modulation range of the transmitted power as a function of
the imaginary part of the refractive index (e), which, for its part, depends on the input power
transmitted in the waveguide (f). It may be observed that in both the configurations, the sole
presence of an assembly produces a tangible NL modulation as function of the input optical
power, while showing a sigmoidal shape. As expected, the configuration with the MNPs/QD
system in the middle gives rise to a larger modulation range, reaching ~1.5 dB compared to
the 0.2 dB obtained with a particle placed on top.
l
i
d
e
F
c
i
r
t
c
e
l
E
d
e
z
i
l
a
m
r
o
N
d
e
l
i
F
l
c
i
r
t
c
e
E
d
e
z
i
l
a
m
r
o
N
Top
Middle
TOP
c)
e
c
n
a
t
t
i
m
s
n
a
r
T
0.65
0.6
0.55
0.5
0.45
0.4
MIDDLE
0.35
102
103
104
105
106
107
B
B
d
d
2
2
0
.
.
2
0
B
B
d
d
8
9
8
.
.
2
2
Fig. 3. Engineering the all-optical nonlinearity via array: a-b) Normalized electric field
distribution for a middle horizontal cut plane of the waveguide, considering an array of closely
spaced hybrid-structures, being in the middle (a) and on top (b) of the waveguide, for their
maximum absorption (largest k). (c) Transmittance as function of the input power density.
a)
b)
Proceeding further with our analysis, in order to enhance the modulation range and
consequently to introduce larger NLs in the transmitted signal, we consider the system
composed by arrays of MNPs/QD assemblies that are distributed either on top or in the
middle of the waveguide. Similarly to what is shown for single assemblies, Fig. 3(a,b) shows
the electric field distribution in the horizontal cut-plane of the waveguide, when either 4 or 8
assemblies are placed in the middle and on top. The assemblies are spaced with a 200 nm
pitch in both configurations. In both the configurations, the electric field transmitted at the
end of the waveguide is significantly attenuated, when the assemblies retain their maximum k.
More specifically, the modulation depth now approaches 50% for the top configuration and
65% for the middle configuration. Fig. 3(c) shows the NL modulation range of the proposed
devices as function of the input power density. The NL can affect the transmitted power by
more than 2 dB in the top array configuration and almost 3 dB in the middle array
configuration.
a)
a)
c)
b)
1
0.65
l
i
d
e
F
c
i
r
t
c
e
b)
l
E
d
e
z
i
l
a
m
r
o
N
d
e
l
i
F
l
c
i
r
t
c
e
E
d
e
z
i
l
a
m
r
o
N
e
c
)
n
V
a
(
t
t
t
i
u
m
p
s
t
u
n
O
a
r
T
0.6
0.8
0.55
0.6
0.5
0.4
0.45
0.2
0.4
0
0.35
Top
Middle
TOP
Probe
Pump
Probe
Pump
MIDDLE
102
10-2
103
Input fluence (µJ/cm2)
100
10-1
104
105
Input (V)
101
106
102
107
B
d
2
0
.
2
B
d
6
6
.
B
d
8
8
.
2
Figure 4. Optical NL in a Reverse Saturable Absorber. Nonlinear transmission of a reverse
saturable absorber made of high-concentration C60 in a PVA host thin film. (a) The simplified
band-diagram of the reverse saturable absorber modeled by a five-level system. (b)
Transmission vs. input fluence at 532 nm. Pump/Probe parameters: full-width at half
maximum = 1 ps and of the probe = 5 fs. Both pump and probe are centered at a wavelength of
532 nm. The lifetimes, τS1 = 30 ns, τT1 = 280 µs, τISC = 1.2 ns, τS2 = τT2 = 1ps.
Another possibility for inducing optical NL in a PIC is exploiting reverse saturable absorption
(RSA). RSA is a property of a material for which the absorption increases as the light
intensity increases. The response of the coupled MNPs/QD assemblies is an ultra strong,
"artificial" RSA that arises from the Fano interference between the components. However, a
broad range of materials show "natural," albeit weaker, RSA, such as organic
compounds [21,22], heavy metals [23], and clusters of metallic particles [24]. In analogy to
the discussion before, we next consider the use of C60 (Buckminsterfullerene or Buckyball)
dispersed in polyvinyl alcohol (PVA) as a reverse saturable absorber to obtain an intensity-
dependent transmission as a NL activation function. The band diagram of the majority of
reverse saturable absorbing materials can be approximated by five energy levels representing
two singlet and one triplet states, as depicted in Fig. 4a. We use rate equations to model the
transitions among the different energy levels and to capture the carrier kinetics in the system
and model the nonlinear light-matter interaction. Full details of the model can be found in
Ref. [25]. The model is integrated with a finite-difference time-domain (FDTD) solver using
an auxiliary differential equation approach to perform full-wave multiphysics analysis of the
structures using the same structured grid and time-stepping.
A pump-probe analysis is used to extract the nonlinear transmission where a strong pump
signal first illuminates the structure followed by a weak probe signal to probe the system
response (Fig. 4a). The respective lifetimes are taken from optical characterization [22]. The
concentration of the C60 molecules is chosen to be 10 mM to provide sufficient absorption
from a thin film. The transmission of 1 µm-thin film of C60 in PVA (refractive index ~ 1.45)
is calculated as a function of the input fluence and shown in Fig. 4b. The linear transmission
of the film is 0.84, and the saturation fluence is about 0.67 µJ/cm2. We can also conclude that
the NL modulation associated with the input power is approximately 7 dB.
After obtaining a small library consisting of 3 different NL optical responses, we introduce
these as nonlinear activation functions on a standardized NN training set, MNIST classifiers
of handwritten digits.
a)
a)
b)
b)
Fully-Connected
(100x100)
c)
c)
Activation Function
(10000)
Fully-Connected
(100x10)
Activation Function
(1000)
Fully-Connected
(10)
Softmax
PHASE SHIFTERS
(cid:9)(cid:3)
(cid:9)(cid:4)
(cid:9)(cid:5)
(cid:9)(cid:6)
(cid:10)(cid:3)
(cid:10)(cid:4)
(cid:10)(cid:5)
(cid:10)(cid:6)
COMBINER
NL ACTIVATION
FUNCTION
(cid:10)(
((cid:9)(cid:8)(cid:10)(cid:8)
(cid:8)
(cid:7)(cid:1)(cid:10)((cid:2)
Prediction accuracy for unseen
validation data
e)
100
98
)
%
(
y
c
a
r
u
c
c
A
)
96
%
(
y
c
a
r
u
c
c
A
94
1
0.98
0.96
0.94
Prediction accuracy for training
data
e)
d)
d)
100
1
98
0.98
)
%
(
y
c
a
r
u
c
c
A
)
96
%
(
y
c
a
r
u
c
c
A
94
0.96
0.94
92
0.92
90
0.9
0
0
10
10
Tanh
Sigmoid
ReLu
QD Top
QD Middle
C60
20
30
30
20
Epoch (#)
Epoch (#)
40
40
50
50
92
0.92
0.9
90
0
0
10
10
20
30
Epoch (#)
20
30
Epoch (#)
Tanh
Sigmoid
ReLu
QD Top
QD Middle
C60
40
40
50
50
Fig. 5. Design of an AO neuron and evaluation of the Neural Network performance a)
Schematic of an all-optical neuron. b) Representation of the emulated NN (b). c) Different
activation functions, VIN to VOUT, ReLU (light blue dashed line), Sigmoid (blue dashed line),
hyperbolic tangent (dark blue line) and the 2 proposed NL function, top (solid orange) and
middle (solid red) d-e) Prediction accuracy as function of epoch for the training (d) and
validation (e) phase of the network.
Figure 5a shows the overall hardware implementation of the AONN using photonic integrated
circuits. As previously reported [5,26], the input optical signals are weighted by phase shifters
and integrated by photonic combiners. Hereby, the NL activation functions (AF) can be
achieved either by means of induced transparency through plasmon-exciton coupling or by
reverse saturable absorption of C60. The proposed all-optical NL modulation of the signal
presents several advantages over an electro-optical counter part such as no charge fluctuation-
induced noise or ps-short response times.
In the successive steps, we evaluate the functionality of the proposed NL optical response of
the studied approaches as neural AF, by emulating their behavior in a 3-layer fully connected
NN implemented in the Google Tensorflow tool and for the MNIST dataset. The software
implementation MNIST is a known machine-learning dataset comprised of 60,000 grayscale
images of handwritten digits. The task is to identify the representing numbers for each image.
In the MNIST dataset each image has 28x28 pixel resolution. In order to feed each image to a
fully-connected layer we flatten each 2D image into a vector of 784 pixels.
The first layer of the network is comprised of 100 neurons. As the name suggests, fully-
connected networks have all-to-all connections among the neurons in each layer and the
incoming inputs. Thus, the first layer requires 784x100 connections. These are identity
connections and do not perform any type of weighting to the inputs. The second layer in our
network also has 100 neurons, which receive inputs from the first layer. With all-to-all
connections, that translates into 100x100 connections. It should be noted that the NL AFs are
placed between two consecutive layers on each input connection. As a result, we will have
100x100 NL AFs operating between the first and the second layer. The third layer in our
network has only 10 neurons, corresponding to 10 classes of output representing 10 digits
from 0 to 9. This last layer can be regarded as a summarizing layer that reduces the dimension
of the network output to the 10 required dimensions. For all the layers explained so far, an
optical realization is envisioned. However, almost all modern NNs take advantage of a
"Softmax" function implemented at the final layer. Softmax converts the incoming input into
a probability distribution function with one incoming value receiving a high probability and
the remainders receiving much lower probabilities. We have yet to envision an optical
realization for this layer, but it is worth noting that this layer can be replaced with similar AFs
at the cost of lower accuracy.
We train our network with two sets of AFs, namely, optical AF and commonly used AF in
DeepLearning applications. The optical AF are those calculated for MNPs/QD assemblies on
top of the waveguide (QD-Top), assemblies in the middle of the waveguide (QD-Middle),
and C60 films. Respectively, the commonly used software-based AF we used to compare our
photonic ones against the Rectified Linear Unit (ReLU), Sigmoid, and Tanh (Fig. 5b) [27]. It
is worth mentioning that the actual input and output units of our optical AFs are µWatt. The
shown optical AF are for devices of size of 80nm x 30nm. In order to obtain the mathematical
model transfer function of the AFs to be used in Tensorflow, we fit quadratic curves to data
points from the device simulations, with maximum Root Mean Squared Error (RMSE) due to
fitting of 0.23 µWatt. The quadratic AFs are then reconstructed in Tensorflow. It should be
noted that the NL AFs acts on the optical power of the incoming electromagnetic radiation
associated with the photon flux, which quadratically dependents on the electric field.
Moreover, in this study we aim to validate the NN performance by focusing on the shape of
the AFs, without taking into account the noise on both system and device level, which will be
the subject of an in-depth future investigation. Amplitude and phase noise induced by the NL
device, as well as noise in the input signal and weights, could indeed seriously compromise
the ability of the AF to discriminate amongst the data, hindering the accuracy of the solution.
We randomly split the MNIST dataset into two subsets of 50,000 and 10,000 images for
training and validation respectively. The network was initially set to train for 50 epochs;
however, we can see that for QD-middle and C60, the model started to over-fit and the training
and the validation accuracy dropped. Fig. 5c depicts the accuracy as the network is being
trained for 50 epochs, showing that all of the optical AFs are comparable to those of the
software-based networks in terms of the accuracy. QD-Middle is the only one performing
relatively poorly, converging to a maximum accuracy of 96%. The validation data accuracy
corroborates that of training data, as depicted in Fig. 5d. For all the optical NL activation
functions a ~100% accuracy is reached, apart from the QD-middle configuration, which
reaches a plateau at 99%. It is worth mentioning that the accuracy curves for the validation
data are smoother due to the fact that at each validation epoch, the network is evaluated over
the whole set of validation images. In contrast, during training we used batch processing and
a variation of gradient descent termed Stochastic Gradient Descent (SGD) for training. As a
result, at each epoch during the training the network only receives a subset of training images.
This results in slight variation of accuracy for different batches of data for different epochs.
Fig. 6. Comparison of computational energy efficiency and processing speed between existing
electronic neuromorphic demonstrations and our proposed programmable photonic platform.
NN = neural network, AONN = all-optical NN, GPU = graphical processing unit.
The proposed activation mechanisms based on NL LMI, showed compatible performance, in
terms of accuracy as function of epoch with respect to the well-established NL activation
functions known from the software-based machine-learning community. The proposed
architecture offers possible benefits of the absence of parasitic switching and short delays,
since the run-time would simply be given by time-of-flight of a photon through the network.
Therefore, an estimate on the processing time can be given by considering the physical length
of the components of this photonic integrated circuit NN and its effective waveguide index;
for instance the weighted addition obtained through combiners and phase shifters has a
physical length of about 100 µm, a passive waveguide synaptic summation has an on-chip
coverage of around 200 µm, and an estimated NL activation module is less than 10 µm long.
Thus, photonics allows a single neuron to be integrated well within 100's of µm in length,
leading to ~ps computation time-scales [28]. This AONN thence would have a delay of about
few ps, or 1012 MAC/s, and <1017 MAC/J efficiency given, for example, the power levels of
the Fano resonance discussed above. Such performance of the proposed AONN would
potentially be several orders of magnitude more efficient and faster than GPU and electro-
optical neural networks, provided noise is negligible (Fig. 6) [29].
Technology
NVIDIA GPU [30]
Electro-optical NN [30]
All-optical NN [this work]
Efficiency
(MAC/J)
3x106
4x1012
4x1016
Speed
(MAC/s)
107
1010
1012
Table 1. Comparison of different neuromorphic technologies in terms of computing speed and power
efficiency. All optical NN performances are estimated and needs to be considered as an upper-bound.
3. Conclusion
In summary, we have investigated a MNPs/QD system, based on two metal NPs sandwiching
a QD, which showed a coherent nonlinear optical response. This phenomenon was due to
interference between the dipoles of the plasmon oscillation in the metal nanoparticles and the
exciton transition in the QD. Furthermore, we modeled integration of the assembly in a
waveguide platform and optimized the modulation range of the NLs associated with the
transmitted signal, reaching a fully NL optical modulation of the transmitted signal up to 3
dB. Moreover, we also studied the reverse saturable absorption mechanism in a film of C60.
The film displayed a clear NL optical response as function of the impinging power density,
with a modulation range of approximately 7 dB. Moreover, the studied platforms can provide
insights into the speed of a complete NN architecture based on integrated photonics and all
optical activation functions. The proposed NL optical responses were used as activation
functions for fully-connected neural networks, emulated in Tensor Flow. We tested these
nonlinear activation functions on a standardized NN training set, MNIST classifiers of
handwritten digits. Our results show that the accuracy of the ONN can match others
commonly employed for up to 50 number of reconfigurations in both training and validation
phase. From an architecture point of view, our all-optical NN has the potential to significantly
outperform in terms of computing speed and energy efficiency the established architectures
based on either electronics or electro-optics. We estimated an efficiency of <1017 MAC/J and
a speed of 1012 MAC/s for a fully optical NN, which is several order of magnitude higher than
the electro-optic and FPGA counterparts. These new insights could contribute to the design
and fabrication of optical NL modulators, which could pave the way for all-optical high-speed
and efficient NNs. Future experimental work is needed to validate this potential.
4. Methods
Simulations
We use a commercial solver (FDTD Solutions from Lumerical Inc.) built on the finite-
difference time-domain method, which solves the Maxwell equations on a discrete spatio-
temporal grid, for all the simulations related to the MNPs/QD system to waveguide coupling.
The MNPs/QD system is modeled as a 3-D box with the absorptance being swept for
modeling the effect of the induced transparency as function of the input power. The adaptive
mesh algorithm is used to refine the grid in the QD domain. The RSA model of C60 response
is realized with a proprietary FDTD-ADE multiphysics code that brings in the carrier kinetics.
5. List of Abbreviations
AF (nonlinear) Activation Function of the perceptron
ADE Auxiliary Differential Equation(s)
AONN All-optical Neural Network
CMOS Complementary Metal Oxide Semiconductor
GPU Graphic Process Unit
FDTD Finite Difference Time Domain
FPGA Field Programmable Gate Array
MAC Multiply Accumulate
MNP Metal Nanoparticle
PIC Photonic Integrated Circuit
NP Nano-particle
QD Quantum Dot
NL Nonlinear
NN Neural Network
LMI Light matter Interaction
PNN Photonic Neural Network
RSA Reverse Saturable Absorption
6. Funding, acknowledgments, and disclosures
6.1 Acknowledgments
The authors acknowledge fruitful discussion with
team of Prof. Prucnal.
S. I. A. and A. V. K. acknowledge the financial support by DARPA/DSO Extreme Optics and
Imaging (EXTREME) Program, Award HR00111720032
the
B. Marr, B. Degnan, P. Hasler, and D. Anderson, "Scaling energy per operation via
B. J. Shastri, A. N. Tait, T. F. de Lima, M. A. Nahmias, H.-T. Peng, and P. R.
J. Hasler and H. B. Marr, "Finding a roadmap to achieve large neuromorphic
7. References
1.
F. Akopyan, J. Sawada, A. Cassidy, R. Alvarez-Icaza, J. Arthur, P. Merolla, N.
Imam, Y. Nakamura, P. Datta, G. Nam, B. Taba, M. Beakes, B. Brezzo, J. B. Kuang, R.
Manohar, W. P. Risk, B. Jackson, and D. S. Modha, "TrueNorth: Design and Tool Flow of a
65 mW 1 Million Neuron Programmable Neurosynaptic Chip," IEEE Transactions on
Computer-Aided Design of Integrated Circuits and Systems 34, 1537 -- 1557 (2015).
2.
hardware systems," Front. Neurosci. 7, (2013).
3.
an asynchronous pipeline," IEEE Trans. Very Large Scale Integr. Syst. 21, 147 -- 151 (2013).
4.
Prucnal, "Principles of Neuromorphic Photonics," arXiv:1801.00016 [physics] 1 -- 37 (2018).
5.
A. N. Tait, T. F. de Lima, E. Zhou, A. X. Wu, M. A. Nahmias, B. J. Shastri, and P.
R. Prucnal, "Neuromorphic photonic networks using silicon photonic weight banks,"
Scientific Reports 7, 7430 (2017).
6
"Neuromorphic Photonics," Optics & Photonics News, OPN 29, 34 -- 41 (2018).
7.
R. Amin, J. George, J. Khurgin, T. El-Ghazawi, P. R. Prucnal, and V. J. Sorger,
"Attojoule Modulators for Photonic Neuromorphic Computing," in Conference on Lasers and
Electro-Optics (2018), Paper ATh1Q.4 (Optical Society of America, 2018), p. ATh1Q.4.
8.
R. Amin, S. Khan, C. J. Lee, H. Dalir, and V. J. Sorger, "110 Attojoule-per-bit
Efficient Graphene-based Plasmon Modulator on Silicon," in Conference on Lasers and
Electro-Optics (2018), Paper SM1I.5 (Optical Society of America, 2018), p. SM1I.5.
9.
J. George, R. Amin, A. Mehrabian, J. Khurgin, T. El-Ghazawi, P. R. Prucnal, and
V. J. Sorger, "Electrooptic Nonlinear Activation Functions for Vector Matrix Multiplications
in Optical Neural Networks," in Advanced Photonics 2018 (BGPP, IPR, NP, NOMA,
Sensors, Networks, SPPCom, SOF) (OSA, 2018), p. SpW4G.3.
10.
J. George, A. Mehrabian, R. Amin, J. Meng, T. F. de Lima, A. N. Tait, B. J. Shastri,
T. El-Ghazawi, P. R. Prucnal, and V. J. Sorger, "Neuromorphic photonics with electro-
absorption modulators," arXiv:1809.03545 [physics] (2018).
M. A. Nahmias, B. J. Shastri, A. N. Tait, T. F. de Lima, and P. R. Prucnal,
X. Wu, S. K. Gray, and M. Pelton, "Quantum-dot-induced transparency in a
11.
H. Leng, B. Szychowski, M.-C. Daniel, and M. Pelton, "Dramatic Modification of
Coupled-Plasmon Resonances Following Exposure to Electron Beams," J Phys Chem Lett 8,
3607 -- 3612 (2017).
12.
K. Santhosh, O. Bitton, L. Chuntonov, and G. Haran, "Vacuum Rabi splitting in a
plasmonic cavity at the single quantum emitter limit," Nat. Commun. 7, ncomms11823
(2016).
13.
A. Hatef, S. M. Sadeghi, and M. R. Singh, "Plasmonic electromagnetically induced
transparency in metallic nanoparticle -- quantum dot hybrid systems," Nanotechnology 23,
065701 (2012).
14.
nanoscale plasmonic resonator," Optics Express 18, 23633 (2010).
15.
Wiley & Sons, 2013).
16.
resonance in a plasmon-exciton system," Phys. Rev. B 88, 075411 (2013).
17.
Rev. B 6, 4370 -- 4379 (1972).
M. P. Lisitsa, L. F. Gudymenko, V. N. Malinko, and S. F. Terekhova, "Dispersion
18.
of the Refractive Indices and Birefringence of CdSxSe1−x Single Crystals," physica status
solidi (b) 31, 389 -- 399 (1969).
19.
J. A. Schuller, E. S. Barnard, W. Cai, Y. C. Jun, J. S. White, and M. L. Brongersma,
"Plasmonics for extreme light concentration and manipulation," Nature Materials 9, 193
(2010).
20.
V. Yannopapas, E. Paspalakis, and N. V. Vitanov, "Plasmon-Induced Enhancement
of Quantum Interference near Metallic Nanostructures," Phys. Rev. Lett. 103, 063602 (2009).
P. B. Johnson and R. W. Christy, "Optical Constants of the Noble Metals," Phys.
M. Pelton and G. W. Bryant, Introduction to Metal-Nanoparticle Plasmonics (John
R. A. Shah, N. F. Scherer, M. Pelton, and S. K. Gray, "Ultrafast reversal of a Fano
C. Li, L. Zhang, R. Wang, Y. Song, and Y. Wang, "Dynamics of reverse saturable
W. Su, T. M. Cooper, and M. C. Brant, "Investigation of Reverse-Saturable
21.
absorption and all-optical switching in C60," J. Opt. Soc. Am. B, JOSAB 11, 1356 -- 1360
(1994).
22.
Absorption in Brominated Porphyrins," Chem. Mater. 10, 1212 -- 1213 (1998).
23.
J. W. Perry, K. Mansour, S. R. Marder, K. J. Perry, D. Alvarez, and I. Choong,
"Enhanced reverse saturable absorption and optical limiting in heavy-atom-substituted
phthalocyanines," Opt. Lett., OL 19, 625 -- 627 (1994).
24.
Y. Gao, X. Zhang, Y. Li, H. Liu, Y. Wang, Q. Chang, W. Jiao, and Y. Song,
"Saturable absorption and reverse saturable absorption in platinum nanoparticles," Optics
Communications 251, 429 -- 433 (2005).
25.
absorption in time-domain," arXiv:1808.02436 [physics] (2018).
26.
Photonic Convolutional Neural Network Accelerator," arXiv:1807.08792 [cs, eess] (2018).
S. I. Azzam and A. V. Kildishev, "Full-wave analysis of reverse saturable
A. Mehrabian, Y. Al-Kabani, V. J. Sorger, and T. El-Ghazawi, "PCNNA: A
27.
Functions," arXiv:1710.05941 [cs] (2017).
P. Ramachandran, B. Zoph, and Q. V. Le, "Searching for Activation
Bhavin J. Shastri, Paul R. Prucnal, "Principles of Neuromorphic Photonics",
Bhavin J. Shastri, Paul R.Prucnal, "Principles of Neuromorphic Photonics",
M. A. Nahmias, B. J. Shastri, A. N. Tait, T. Ferreira de Lima, and P. R. Prucnal,
28.
Encyclopedia of Complexity and systems Science (2017).
29.
"Neuromorphic Photonics," Optics and Photonics News 29, 34 (2018).
30.
Encyclopedia of Complexity and systems Science (2017).
8. Article thumbnail upload
Preview of thumbnail image display on the author submission page.
|
1902.10806 | 1 | 1902 | 2019-01-31T17:25:18 | Analysis and control of geometrically nonlinear responses of piezoelectric FG porous plates with graphene platelets reinforcement using B\'ezier extraction | [
"physics.app-ph",
"math.NA"
] | In this study, we propose an effective numerical approach to analyse and control geometrically nonlinear responses for the functionally graded (FG) porous plates reinforced by graphene platelets (GPLs) integrated with piezoelectric layers. The basis idea is to use isogeometric analysis (IGA) based on the B\'ezier extraction and the $C^0$-type higher-order shear deformation theory ($C^0$-HSDT). By applying the B\'ezier extraction, the original Non-Uniform Rational B-Spline (NURBS) control meshes can be transformed into the B\'ezier elements which allow us to inherit the standard numerical procedure like the finite element method (FEM). The mechanical displacement field is approximated based on the $C^0$-HSDT whilst the electric potential is assumed to be a linear function through the thickness of each piezoelectric sublayer. The FG plate contains the internal pores and GPLs dispersed in the metal matrix either uniformly or non-uniformly according to various different patterns along the thickness of plate. In addition, to control dynamic responses, two piezoelectric layers are perfectly bonded on the top and bottom surfaces of the FG plate. The geometrically nonlinear equations are solved by the Newton-Raphson iterative procedure and the Newmark's time integration scheme. The influences of the porosity coefficients, weight fractions of GPLs as well as the external electrical voltage on the geometrically nonlinear behaviours of the plates with different porosity distributions and GPL dispersion patterns are evidently investigated through numerical examples. Then, a constant displacement and velocity feedback control approaches are adopted to active control the geometrically nonlinear static as well as the dynamic responses of the FG porous plates, where the effect of the structural damping is considered, based on a closed-loop control with piezoelectric sensors and actuators. | physics.app-ph | physics | Highlights
• B´ezier extraction based IGA approach is successfully implemented for the nonlinear static
and dynamic analyses of the FG plate reinforced by GPLs and integrated with piezoelectric
layers.
• The combination of two porosity distribution types and three GPL dispersion patterns along
the thickness direction of the FG plate is presented.
• The influences of the porosity coefficients, weight fractions of GPLs and the external elec-
trical voltage are investigated.
• A constant displacement and velocity feedback control approaches are adopted to active
control the responses of the plate structures with piezoelectric sensors and actuators.
• Numerical results demonstrate the efficiency and reliability of the proposed approach.
9
1
0
2
n
a
J
1
3
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
1
v
6
0
8
0
1
.
2
0
9
1
:
v
i
X
r
a
1
Analysis and control of geometrically nonlinear responses of
piezoelectric FG porous plates with graphene platelets
reinforcement using B´ezier extraction
Nam V. Nguyena, Lieu B. Nguyenb, Jaehong Leec, H. Nguyen-Xuand,∗
aFaculty of Mechanical Technology, Industrial University of Ho Chi Minh City, Ho Chi Minh City, Vietnam
bFaculty of Civil Engineering, University of Technology and Education Ho Chi Minh City, Vietnam
cDepartment of Architectural Engineering, Sejong University, 98 Gunja-dong, Gwangjin-gu, Seoul 143-747, South
dCIRTech Institute, Ho Chi Minh City University of Technology (HUTECH), Ho Chi Minh City, Vietnam
Korea
Abstract
In this study, we propose an effective numerical approach to analyse and control geometrically
nonlinear responses for the functionally graded (FG) porous plates reinforced by graphene platelets
(GPLs) integrated with piezoelectric layers. The basis idea is to use isogeometric analysis (IGA)
based on the B´ezier extraction and the C 0-type higher-order shear deformation theory (C 0-HSDT).
By applying the B´ezier extraction, the original Non-Uniform Rational B-Spline (NURBS) control
meshes can be transformed into the B´ezier elements which allow us to inherit the standard nu-
merical procedure like the finite element method (FEM). The mechanical displacement field is
approximated based on the C 0-HSDT whilst the electric potential is assumed to be a linear func-
tion through the thickness of each piezoelectric sublayer. The FG plate contains the internal pores
and GPLs dispersed in the metal matrix either uniformly or non-uniformly according to various
different patterns along the thickness of plate.
In addition, to control dynamic responses, two
piezoelectric layers are perfectly bonded on the top and bottom surfaces of the FG plate. The
geometrically nonlinear equations are solved by the Newton-Raphson iterative procedure and the
Newmark's time integration scheme. The influences of the porosity coefficients, weight fractions
of GPLs as well as the external electrical voltage on the geometrically nonlinear behaviours of
the plates with different porosity distributions and GPL dispersion patterns are evidently investi-
gated through numerical examples. Then, a constant displacement and velocity feedback control
approaches are adopted to active control the geometrically nonlinear static as well as the dynamic
responses of the FG porous plates, where the effect of the structural damping is considered, based
on a closed-loop control with piezoelectric sensors and actuators.
Keywords: Piezoelectric materials, FG porous plate, Graphene platelets reinforcement, B´ezier
extraction, Nonlinear dynamic, Active control
∗Corresponding author
Email address: [email protected] (H. Nguyen-Xuan )
Preprint submitted to Elsevier
March 1, 2019
1. Introduction
Nowadays, the demand for high performance structures with superior mechanical properties
and chemical stability in engineering applications has been significantly increased. With these cel-
lular structures, the porous materials whose the excellent properties such as lightweight, excellent
energy absorption, heat resistance has been extensively employed in various fields of engineering
including aerospace, automotive, biomedical and other areas [1 -- 5]. However, the existence of in-
ternal pores leads to significant reduction in the structural stiffness [6]. In order to overcome this
shortcoming, the reinforcement with carbonaceous nanofillers such as carbon nanotubes (CNTs)
[7, 8] and graphene platelets (GPLs) [9, 10] into the porous materials is an excellent and practical
choice to strengthen their mechanical properties. More importantly, this reinforcement aims also
to maintain their potential for lightweight structures [11, 12]. In comparison with CNTs, GPLs
have demonstrated great potentials to become a good candidate for reinforcement [13, 14] since
GPLs have superior mechanical properties, a lower manufacturing cost, a larger specific surface
area and two-dimensional geometry. In order to increase the performance of structure, the func-
tionally graded (FG) porous structures reinforced by GPLs have been proposed in the literature to
obtain the desired mechanical properties by modifying the sizes, the density of the internal pores
in different directions as well as the dispersion patterns of GPLs [15 -- 17]. In terms of numerical
analysis, large number of investigations have been conducted to study the influences of the internal
pores and GPLs on the behaviours of structures under various different conditions. Kitipornchai
et al. [18] and Chen et al. [19] examined the free vibration, elastic buckling and the nonlinear free
vibration, postbuckling behaviours of the FG porous beams reinforced with GPLs, respectively
based on the Timoshenko's beam theory and Ritz method. Yang et al. [20] utilized the first-order
shear deformation plate theory (FSDT) and Chebyshev-Ritz method to study the uniaxial, biax-
ial, shear buckling and free vibration of the FG porous plates reinforced with GPLs uniformly or
non-uniformly distributed in the metal matrix. Based on the isogeometric analysis (IGA), Li et al.
[21] analysed the static, free vibration and buckling of the FG porous plates reinforced by GPLs
using both first- and third-order shear deformation plate theories. Based on combination of the
Galerkin method and the fourth-order Runge-Kutta approach, Li et al. [22] studied the nonlinear
vibration and dynamic buckling of the sandwich FG porous plate reinforced by GPL resting on
Winkler-Pasternak elastic foundation.
On the other hand, the piezoelectric materials have also been extensively applied to build up
advanced smart structures for modern industrial products. One of the excellent and essential fea-
tures of these materials is the ability of transformation between the electrical and mechanical
energy which is known as the piezoelectric effect and the converse phenomenon [23]. Regard-
ing the analysis for the plate structures integrated with piezoelectric layers, a lot of studies have
been conducted to predict their behaviours in the literature [24 -- 29]. In addition, the piezoelectric
FG carbon nanotubes reinforced composite plates (FG-CNTRC) also attracted remarkable atten-
tion of researchers. Alibeigloo [30], [31] investigated the static and the free vibration analyses
of FG-CNTRC plate as well as cylindrical panel embedded in thin piezoelectric layers using the
three-dimensional theory of elasticity. Using FSDT and von K´arm´an strain assumptions, Rafiee
et al. [32] investigated the nonlinear parametric instability of initially imperfect the piezoelectric
3
FG-CNTRC plates under a combination of the electrical and thermal loadings. Then, Sharma et
al. [33] investigated the active vibration control of FG-CNTRC rectangular plates with piezo-
electric sensor and actuator layers using FEM based on FSDT. Selim et al. [34] studied the free
vibration behaviour and active vibration control of FG-CNTRC plates with piezoelectric layers
using element-free IMLS-Ritz model based on Reddy's higher-order shear deformation theory.
Nguyen-Quang et al. [35] studied the dynamic response of laminated CNTRC plates integrated
with piezoelectric layers using IGA and HSDT. Recently, Malekzade et al. [36] employed the
transformed differential quadrature method for the free vibration analysis of FG eccentric annular
plates reinforced with GPLs and integrated piezoelectric layers.
It is known that the different basis functions are applied for approximation of the geometries
and solutions in the framework of traditional FEM which leads to errors in the computational
process. To improve the accuracy of solutions as well as reduce computational costs, the IGA
[37] which employs non-uniform rational B-splines (NURBS) basis as shape functions had been
discovered. The main idea of IGA is fulfilled by using the same basis functions to describe the
geometry model and to approximate the solution field. The IGA has successfully been applied
to various fields of engineering and science. In comparison with the standard FEM, the NURBS
based IGA provides better accuracy and reliability for various engineering problems, especially for
ones with complex geometries which are reported in the literature. The basic and review of IGA
are presented in the established literature [38, 39]. Nevertheless, the implementation of NURBS
based IGA approach is not often easy as their basis functions are not confined to a unique element
but span over the entire domain instead. To overcome these obstacles, Borden et al. [40] proposed
the B´ezier extraction which represents the NURBS basis function in the form of Bernstein polyno-
mials basis defined over C 0 continuous isogeometric B´ezier element. By incorporating Bernstein
polynomials which are similar to the Lagrangian basis functions as basis function in B´ezier ex-
traction, the implementation of IGA becomes analogous to the traditional FEM. As a result, the
IGA approach can easily be embedded in most existing FEM codes while its advantages are still
kept naturally.
In the context of plate theories, there is a great deal of theories that have been introduced and
developed to estimate the responses of plate structures under different conditions. While CPT or
Kirchhoff-Love shows its drawbacks in the analysis of thick plates, FSDT which is capable of both
thin and thick plates requires an appropriate shear correction factor. To overcome these shortcom-
ings, several higher-order plate theories (HSDTs) have been proposed in the literature [41 -- 45].
Nevertheless, these HSDTs require the C 1-continuity of the generalized displacement field which
leads to the second-order derivatives of deflection in the stiffness formulation. Therefore, several
C 0-continuous elements were proposed [46 -- 48].
As previously mentioned, most of the studies mainly focused on studying the plates integrated
with piezoelectric layers which address only the core layer composed of FGM or FG-CNTRC. Fur-
thermore, the geometrically nonlinear static and dynamic analyses of the piezoelectric FG plates
under various loading types are still somewhat limited. In this study, in order to fill the existing
gap in the literature, the geometrically nonlinear static and transient responses of piezoelectric
4
FG plates which have the core layer composed of FG porous materials reinforced by GPLs using
B´ezier extraction via IGA and C 0-type HSDT. More importantly, the active control of the geo-
metrically nonlinear static and dynamic responses of the FG porous plates with the effect of the
structural damping based on a closed-loop control with piezoelectric sensors and actuators is inves-
tigated. By using the B´ezier extraction, the IGA preserves the element structure, which allows the
IGA approach to integrate conveniently into the existing FEM routine. For material distribution,
the core layer is constituted by the combination of two porosity distributions and three dispersion
patterns of GPLs along the thickness plate, while the piezoelectric layer is perfectly bonded on
the both top and bottom surfaces of plate. The Newmark's integration scheme incorporation with
Newton-Raphson iterative procedure is utilized for the geometrically nonlinear static and dynamic
analyses. Then, some verification investigations are also conducted to prove the accuracy and sta-
bility of the present method. The influence of some specific parameters such as different porosity
distributions, porosity volume fractions and GPL dispersion patterns, input voltages on the non-
linear behaviours of plate is addressed and discussed in detail through various numerical examples.
The outline of this paper is as follows. Section 2 provides the material models, the variational
and approximate formulations of the piezoelectric FG porous plates reinforced by GPLs based on
C 0-type HSDT. Meanwhile, Section 3 describes the active control algorithm. Section 4 presents
the numerical examples for the geometrically nonlinear static and transient analyses as well as the
active control of the piezoelectric FG porous plates reinforced with GPLs before some concluding
remarks are given in Section 5.
2. Theoretical formulations
2.1. Material models of the FG porous plate reinforced with GPLs
We consider a FG plate model whose core layer is made of metal foams reinforced by GPLs
with piezoelectric layers as depicted in Fig. 1. The length, width and total the thickness of piezo-
electric FG porous plate are defined as a, b and h = hc + 2hp, respectively, in which hc and hp are
the thicknesses of the porous core and the piezoelectric layers, respectively. The porous core layer
of plate is constituted by combining of two different porosity distribution types and three GPL
dispersion patterns along the thickness direction of plates which are depicted in Fig. 2, respec-
tively. The material properties including the Young's modulus, shear modulus and mass density
through the thickness of the porous core layer corresponding to two porosity distribution types can
be expressed as
where
λ (z) =
in which E1 and ρ1 denote the maximum values of Young's modulus and mass density in the
thickness direction of the porous core layer, respectively. Meanwhile, the coefficient of porosity
5
E (z) = E1 [1 − e0λ (z)] ,
(cid:26) cos (πz/hc),
G (z) = E (z) / [2 (1 + ν (z))] ,
ρ (z) = ρ1 [1 − emλ (z)] ,
Porosity distribution 1
cos (πz/2hc + π/4), Porosity distribution 2
(1)
(2)
e0 is determined by
(3)
where E(cid:48)
2 stand for the maximum and minimum values of Young's modulus for the porous
core layer without GPLs, as shown in Fig. 2. Based on Gaussian Random Field (GRF) scheme
[49], the mechanical properties of closed-cell cellular solids can be given as
1 and E(cid:48)
e0 = 1 − E(cid:48)
2/E(cid:48)
1.
Then, the mass density coefficient em in Eq. 1 can be determined as
(cid:19)
< 1
.
E (z)
E1
=
(cid:18) ρ (z) /ρ1 + 0.121
(cid:16)
1.121
1.121
em =
ν (z) = 0.221p(cid:48) + ν1
for
(cid:17)
(cid:18)
0.15 <
ρ (z)
ρ1
(cid:19)2.3
1 − 2.3(cid:112)1 − e0λ (z)
(cid:0)0.342p(cid:48)2 − 1.21p(cid:48) + 1(cid:1) ,
1 − 2.3(cid:112)1 − e0λ (z)
(cid:17)
λ (z)
.
.
(cid:16)
Also according to the closed-cell GRF scheme [50], Poisson's ratio ν(z) is determined by
where ν1 represents the Poisson's ratio of metal without internal pores with p(cid:48) is given as
(4)
(5)
(6)
(7)
(8)
(10)
(11)
p(cid:48) = 1.121
Si1 [1 − cos (πz/hc)],
(cid:90) hc/2
The volume fraction of GPLs varies along the thickness direction of plate for three dispersion
patterns which are illustrated in Fig. 2 can be given as
VGP L =
Pattern A
Si2 [1 − cos (πz/2hc + π/4)], Pattern B
Pattern C
Si3,
(cid:90) hc/2
where Si1,Si2 and Si3 are the maximum values of GPL volume fraction, in which i = 1, 2 corre-
spond to two porosity distributions. The weight fraction of GPLs is related to its volume content
which are given as follows
ΛGP Lρm
ΛGP Lρm + ρGP L − ΛGP LρGP L
The effective Young's modulus of porous core layer reinforced with GPLs without internal
−hc/2
−hc/2
VGP L [1 − emλ (z)]dz.
(9)
×
[1 − emλ (z)]dz =
pores is determined by the Halpin-Tsai micromechanics model [51, 52] as
(cid:18)1 + ξLηLVGP L
1 − ηLVGP L
(cid:19)
Em +
5
8
(cid:18)1 + ξW ηW VGP L
1 − ηW VGP L
(cid:19)
Em,
E1 =
3
8
in which
ξL = 2lGP L
tGP L
, ξW = 2wGP L
tGP L
, ηL = (EGP L/Em)−1
(EGP L/Em)+ξL
6
, ηW = (EGP L/Em)−1
(EGP L/Em)+ξW
,
where wGP L, lGP L and tGP L are dimensions of GPLs including the average width, length and
thickness, respectively; Meanwhile, EGP L and Em are the Young's modulus of GPLs and metal
matrix, respectively. Finally, ρ1 and ν1 denote the mass density and Poisson's ratio of the GPLs
reinforced porous metal matrix can be determined based on the rule of mixture [53]
ρ1 = ρGP LVGP L + ρmVm,
(12)
(13)
where the mechanical properties for GPLs and metal matrix are denoted with subscript symbols
GP L and m, respectively. Meanwhile, VGP L and Vm = 1 − VGP L denote the volume fraction of
GPLs and metal matrix, respectively.
ν1 = νGP LVGP L + νmVm,
2.2. Weak form of the governing equations
The governing equations of motion for the piezoelectric FG plate can be obtained by applying
the Hamilton's variational principle [54] which can be expressed as follows
Ldt = 0,
(14)
in which t1 and t2 denote the starting and finish time values, respectively. Meanwhile, L is the
general energy functional which contains the summation of the kinetic energy, the strain energy,
the dielectric energy and the external work is expressed as follows
uT fsdΓs−
φqsdΓφ +
(cid:88)
uT Fp−(cid:88)
φQp, (15)
(cid:90)
(cid:90)
L =
1
2
(cid:0)ρ uT u − σT ε + DT E(cid:1)dΩ +
Ω
Γs
Γφ
where ρ denotes the mass density; u and u represent the mechanical displacement and velocity
field vectors; φ represents the electric potential; Meanwhile, fs and Fp denote the external me-
chanical surface and concentrated load vectors; qs and Qp indicate the external surface and point
charges, respectively; Γs and Γφ denote the external mechanical and the electrical loading surface,
respectively.
(cid:90) t2
t1
δ
(cid:90)
Then, the variational form for the equations of motion can be expressed as follows
fsδuT dΓsdt−(cid:82) t2
(cid:82)
t1
(cid:82)
(cid:82) t2
+(cid:82) t2
t1
t1
Ω
(cid:82)
(cid:0)ρuδuT − σT δε + DT δE(cid:1)dΩdt+(cid:82) t2
(cid:80) FpδuT dt −(cid:82) t2
(cid:80) Qpδφdt = 0.
(cid:20) σ
(cid:21)
(cid:20) c −eT
t1
t1
Γs
(cid:21)(cid:20) ¯ε
(cid:21)
=
D
e
g
E
7
In this study, the linear constitutive relationships of the FG porous plate reinforced by GPLs
with the piezoelectric layers can be presented as follows [55]
qsδφdΓφdt
Γφ
(16)
,
(17)
in which ¯ε = [ε, γ]T and σ represent the strain and stress vectors, respectively; D denotes the
dielectric displacement vector and e represents the stress piezoelectric constant matrix; g indicates
the dielectric constant matrix; Meanwhile, the electric field vector E which is calculated following
the electric potential field φ can be defined as [56]
And the material constant matrix c is defined as
E = −gradφ = −∇φ.
,
A B L 0
B G F 0
L F H 0
0
0
0
0
0
0 As Bs
0 Bs Ds
0
0
c =
(A, B, G, L, F, H) =(cid:82) hc/2
(As, Bs, Ds) =(cid:82) hc/2
1
Qb =
Ee
1 − ν2
e
νe
1
0
νe
0
in which
in which
−hc/2 (1, z, z2, f (z), zf (z), f 2(z))Qb
ijdz,
−hc/2 (1, f(cid:48)(z), f(cid:48)2(z))Qs
ijdz,
, Qs =
0
0
1−νe
2
(cid:20) 1 0
0 1
(cid:21)
.
Ee
2(1 + νe)
(18)
(19)
(20)
(21)
where Ee and νe are the effective Young's modulus and Poisson's ratio, respectively.
2.3. Approximation of the mechanical displacement field
2.3.1. C 0-type higher-order shear deformation theory
Considering a plate carrying a domain V = Ω × (−h
2 ), in which Ω ∈ R2. Based on the
higher-order shear deformation theory [57], the displacement field at an arbitrary point in the plate
can be presented as follows
2 , h
where
u =
u(x, y, z) = u0(x, y) + zu1(x, y) + f (z)u2(x, y),
u
v
w
, u0 =
u0
v0
w0
, u1 = −
w0,x
w0,y
0
, u2 =
θx
θy
0
,
(22)
(23)
in which u0, v0, w0, θx and θy are the displacement components in the x, y, z directions and the
rotation components in the y- and the x- axes, respectively. Meanwhile, the subscript symbols x
and y denote the derivatives of any function with respect to x and y directions, respectively; and
f (z) is a function of the z−coordinate which is defined to describe the shear strains and stresses
along the thickness of plate, as is listed in [58]. In this work, the famous third-order function
proposed by Reddy is utilized as f (z) = z − 4z3
3h2 [59].
8
In order to avoid the high order derivations in approximate formulations and conveniently
impose the boundary conditions, additional assumptions are formulated as follows
Then, substituting Eq. 24 into Eq. 23, one obtains
w0,x = βx, w0,y = βy.
u0
v0
w0
, u1 = −
βx
βy
0
, u2 =
θx
θy
0
.
u0 =
It can be seen that the compatible strain fields which are presented in Eq. 25 only require the
C 0-continuity of the generalized displacements. Therefore, this theory is called as the C 0-type
higher-order shear deformation theory (C 0-HSDT).
The Green strain vector of a bending plate can be expressed in compact form as follows
(24)
(25)
(26)
Employing the von K´arm´an assumptions, the strain-displacement relations can be rewritten as
where
ε0 =
v0,y
2
∂xj
εij =
1
2
(cid:19)
.
+
∂uj
∂xi
+
∂uk
∂xi
∂uk
∂xj
γ = {γxz, γyz}T = εs + f(cid:48)(z)κs,
(cid:18) ∂ui
ε =(cid:8)εxx, εyy, γxy
(cid:9)T = ε0 + zκ1 + f (z)κ2,
= εL
w2
+ 1
u0,x
θx,x
, κ2 =
(cid:27)
(cid:27)
(cid:26) w0,x − βx
(cid:26) θx
w,x
(cid:26) w,x
(cid:27)
u0,y + v0,x
βx,x
βy,y
w0,y − βy
,x
w2
,y
2w,xy
βx,y + βy,x
θx,y + θy,x
=
ΘΛ.
εN L
0 =
, κs =
θy,y
θy
1
2
0
0 w,y
w,y w,x
w,y
1
2
0 + εN L
0
,
κ1 = −
ε0 =
,
(27a)
(27b)
(28)
(29)
where the nonlinear strain component is expressed as follows
2.3.2. Isogeometric analysis based on B´ezier extraction of NURBS
2.3.2.1. B-spline and NURBS basis functions. In one dimensional (1D) space, the B-spline basis
functions can be expressed by a set of knot vector in the parametric space which is defined by
Ξ = {ξ1, ξ2, ..., ξn+p+1}, where (i = 1, ...n + p) denotes the knot index. Meanwhile n and p are
9
the number of basis functions and the polynomial order, respectively. For a given knot vector Ξ,
the B-spline basis functions are defined according to recursive form
(cid:26) 1,
Ni,0 (ξ) =
if ξi ≤ ξ ≤ ξi+1
0, otherwise
, for p = 0,
Ni,p (ξ) =
ξ − ξi
ξi+p − ξi
Ni,p−1 (ξ) +
ξi+p+1 − ξ
ξi+p+1 − ξi+1
Ni+1,p−1 (ξ) , for p > 0.
Then, B-spline curves can be determined by taking a linear association of B-spline basis func-
tions and the control points Pi (i = 1, 2, ..., n) as
(30)
(31)
T (ξ) =
PiNi,p (ξ).
(32)
n(cid:88)
i=1
n(cid:80)
m(cid:80)
i=1
j=1
n(cid:88)
m(cid:88)
n(cid:88)
m(cid:88)
In two dimensional (2D) space, the B-splines basis functions can be also obtained by taking
a tensor product of two basis functions in 1D space. Similarly, the B-splines surfaces are also
expressed by
S (ξ, η) =
Pi,jNi,p (ξ) Mj,q (η) = PT N (ξ, η) .
(33)
i=1
j=1
in which Ni,p and Mj,q represent the basis functions with orders p and q in the ξ and η directions
corresponding with the knot vectors Ξ = {ξ1, ξ2, ..., ξn+p+1} and H = {η1, η2, ..., ηm+q+1}, re-
spectively.
Due to B-splines basis functions are limited the ability to exactly description some conic shapes
such as circles, cylinders, ellipsoids and spheres, the NURBS have been introduced based on the
B-spline and a set of weights. Accordingly, the NURBS basis functions can be expressed as
Ri,j (ξ, η) =
Ni,p (ξ) Mj,q (η) wi,j
Ni,p (ξ) Mj,q (η) wi,j
,
(34)
where wi,j represents the weight values. Then, the NURBS surfaces can be determined as follows
S (ξ, η) =
Ri,j (ξ, η)Pi,j.
(35)
i=1
j=1
2.3.2.2. B´ezier extraction of NURBS. The major purpose of B´ezier extraction is to instead the
NURBS basis functions by the C 0-continuous Bernstein polynomial basis functions defined over
B´ezier elements which have the similar element structure with standard FEM. By using the Bern-
stein polynomial as the basis function in B´ezier extraction, the IGA approach is straightforwardly
performed as well as can be integrated in most available FEM structures. It is well known that
10
the B-spline basis function of pth order has C p−k continuity across each element, in which k rep-
resents the multiplicity of knots in the knot value. Therefore, the C 0-continuity can be obtained
by inserting the new knots into the B-spline basis function until k = p. Accordingly, a new knot
¯ξ ∈ [ξk, ξk+1] with (k > p) is inserted into the original knot vector Ξ = {ξ1, ξ2, ..., ξn+p+1}. As a
result, a new set of control points are obtained and expressed as follows [37]
¯Pi =
where
P1,
1
αi =
αiPi+(1−αi)Pi−1
Pn
i = 1,
1 < i < n + 1,
i = n + 1,
1 ≤ i ≤ k − p,
k − p + 1 ≤ i ≤ k,
i ≥ k + 1,
¯ξ−ξi
ξi+p−ξi
0
in which Pi and ¯Pi are the original and new control points, respectively.
Then, the B´ezier extraction operator can be determined by using the new set of knots(cid:8) ¯ξ1, ¯ξ2, ..., ¯ξn+1
(cid:9)
as follows [40, 60]
Cj =
α1 1 − α2
0
α2
0
0
...
0
. . .
0
1 − α3
α3
. . .
0
1 − α4
. . .
0
. . .
0
0
0
0 αn+j−1 1 − αn+j
.
(36)
(37)
(38)
(39)
(40)
Applying the B´ezier extraction operator Cj, a new B´ezier control points Pb associated with Bern-
stein polynomial basis can be determined as follows [61]
where the whole B´ezier extraction operator C is defined as
Pb = CT P,
(cid:89)n
C =
Cj.
j=1
It should be noted that the geometries will not change after inserting a new knot into the origi-
nal knot vector. As a result, the B-spline surface is also obtained based on Bernstein polynomials
and B´ezier control points as
n(cid:88)
m(cid:88)
i,j =(cid:0)Pb(cid:1)T
S (ξ, η) =
Bi,j (ξ, η) Pb
B (ξ, η) ,
(41)
in which the 2D Bernstein polynomials B (ξ, η) in terms of parametric coordinates ξ and η are
defined recursively as
i
j
Bi,j,p (ξ, η) = 1
4 (1 + ξ) (1 − η) Bi−1,j,p−1 (ξ, η) + 1
1
4 (1 − ξ) (1 + η) Bi,j−1,p−1 (ξ, η) + 1
4 (1 + ξ) (1 + η) Bi−1,j−1,p−1 (ξ, η) ,
4 (1 − ξ) (1 − η) Bi,j,p−1 (ξ, η) +
11
(42)
where
B1,1,0 (ξ, η) = 1, Bi,j,p (ξ, η) = 0 (i, j < 1 or i,j> p + 1) .
From Eq. 33 and Eq. 41, yields the following relation
(cid:0)Pb(cid:1)T
B (ξ, η) = PT N (ξ, η) .
(43)
(44)
According to Eq. 39 for 2D, the B-spline basis functions in Eq. 44 can be rewritten based on
Bernstein polynomials as follows
Based on Eq. 45, the NURBS basis functions can be presetned by Bernstein polynomials as
N (ξ, η) = CB (ξ, η) ,
(45)
follows
R (ξ, η) =
W
W (ξ, η)
N (ξ, η) =
W
W (ξ, η)
CB (ξ, η) ,
(46)
where W denotes the diagonal matrix of the local NURBS weights. Meanwhile, the weight func-
tions W (ξ, η) are expressed with the Bernstein basis functions as follows
W (ξ, η) =(cid:0)CT w(cid:1)T
B (ξ, η) =(cid:0)wb(cid:1)T
B (ξ, η) ,
(47)
where w and wb are the weights for the NURBS and B´ezier, respectively. The relation of B´ezier
control points and NUBRS ones is described by
Pb =(cid:0)Wb(cid:1)−1
CT WP.
(48)
2.3.2.3. B´ezier extraction of NURBS for FG porous plate formulations. Based on the B´ezier ex-
traction of NURBS, the mechanical displacement field u(ξ, η) of the FG porous plate can be
approximated as follows
u (ξ, η) =
Re
A (ξ, η)dA,
in which n× m represents the number of basis functions. Meanwhile, Re
basis function which is presented in Eq. 46; dA =(cid:8) u0A v0A w0A βxA βyA θxA θyA
A (ξ, η) denotes a NURBS
A
the vector of the nodal degrees of freedom associated with the control point A.
By substituting Eq. 49 into Eq. 28, the in-plane and shear strains can be expressed as
(cid:19)
[ε, γ]T =
BL
A +
1
2
BN L
A
dA,
m×n(cid:88)
(cid:18)
m×n(cid:88)
A=1
12
(49)
(cid:9)T is
(50)
B1 =
B3 =
Bs1 =
(cid:3)T , where
, B2 = −
,
0 0 0 0 0
RA,y 0 0 0 0 0
RA,y RA,x 0 0 0 0 0
0
0
0 0 0 0 0
RA,y
0 0 0 0 0 RA,y RA,x
0
0
A B3
A B2
A =(cid:2) B1
RA,x
0 0 0 0 0 RA,x
(cid:20) 0 0 RA,x −RA
wA,x
0 0 RA,y
Θ =
0
0
0 0
−RA 0 0
, Λ =
0
0
wA,y
wA,y wA,x
0 0 0 RA,x
0
0
0 0
0 0 0
RA,y 0 0
0 0 0 RA,y RA,x 0 0
(cid:21)
, Bs2 =
(cid:20) 0 0 0 0 0 RA
(cid:21)
(cid:20) 0 0 RA,x 0 0 0 0
0 0 0 0 0
0 0 RA,y 0 0 0 0
.
0
0 RA
,
(cid:21)
,
(51)
(52)
in which BL
A Bs1
A Bs2
A
and BN L
A = ΘΛ, in which
2.4. Approximation of the electric potential field
By discretizing the piezoelectric layer into finite sublayers along the thickness, the electric
potential field on each layer is then approximated. Accordingly, in each sublayer, the electric
potential variation is considered to be linear and is approximated through the thickness as follows
[62]
(53)
φ is the shape function of the electric potential function which is determined in Eq. 46
with p = 1. Meanwhile, φi =(cid:2) φi−1, φi (cid:3) with (i = 1, 2, ..., nsub) denotes the electric potentials
φi(z) = Ri
where Ri
at the top and bottom surfaces of the sublayer, where nsub represents the number of piezoelectric
sublayers.
φφi,
In each sublayer element, the values of the electric potentials are estimated to equal at the same
height according to z− direction [55]. Therefore, the electric potential field E for each sublayer
element which is presented in Eq. 18 can be expressed as follows
where
(cid:110)
Bφ =
E = −∇Ri
φφi = −Bφφi,
(cid:111)T
.
0 0
1
hp
(54)
(55)
13
Finally, the stress piezoelectric constant matrix e, the strain piezoelectric constant matrix k
and the dielectric constant matrix g can be determined by [62]
, k =
0
0
0
0
0
0
e13 e31 e33
0
e15
0
e15
0
0
0
0
0
0
0
k31 k32 k33
0
k15
0
k15
0
0
, g =
p11
0
0
,
(56)
0
p22
0
0
0
p33
2.5. Governing equation of motion
By substituting Eqs. 49 and 54 into Eq. 16, the final form of the elementary governing equation
can be obtained and expressed as follows [62]
(cid:20) Muu 0
(cid:21)(cid:20) d
(cid:21)
0
0
+
φ
(cid:20) Kuu Kuφ
Kφu −Kφφ
(cid:21)(cid:20) d
(cid:21)
=
φ
(cid:21)
(cid:20) f
Q
,
(57)
2BN L)dΩ, Kφφ =(cid:82)
uφ, Muu=(cid:82)
Ω BT
Ω (BL + BN L)T c(BL + 1
Ω (BL)T eT BφdΩ, Kφu = KT
φ gBφdΩ,
NT m NdΩ, f =(cid:82)
N = [ N0 N1 N2 ]T , ¯N =(cid:2) 0 0 RA 0 0 0 0 (cid:3) ,
m f (z) eT
f(cid:48) (z) eT
e = [ eT
m zeT
m eT
s
s ],
Ω
Ω ¯q0 ¯NdΩ,
0
e =
where
Kuu =(cid:82)
Kuφ =(cid:82)
in which
where
em =
N1 = −
and
0
e15
0
0
0
0
0
0
e31 e32 e33
0
, es =
0 0 0 RA
0
0 0 0
0 0 0
0 0
0 RA 0 0
0 0
0
0
0 0 0 0
0 0 0 0
0 RA 0 0 0 0
0
0
e15
0
0
0
0 RA
0
, N0 =
RA
0 0 0 0 0 RA
, N2 =
I1
(cid:90) h/2
ρ(z)(cid:0)1, z, z2, f (z), zf (z), f 2(z)(cid:1)dz.
0 0 0 0 0
0 0 0 0 0
0
0 RA
0
0
I4
I5
I6
I2
I3
I5
I2
I4
m =
in which the mass inertia terms Ii with (i = 1 : 6) are given as
(I1, I2, I3, I4, I5, I6) =
−h/2
14
,
(58)
(59)
(60)
(61)
(62)
Since the electric field E exists only according to the z direction, Kuφ in Eq. 58 can be rewritten
a
Kuφ =
(cid:90)
Ω
Muu
(cid:16)(cid:0)B1(cid:1)T eT
(cid:16)
mBφ + f (z)(cid:0)B3(cid:1)T eT
mBφ + z(cid:0)B2(cid:1)T eT
(cid:17)
d = F + KuφK−1
Kφu
Kuu + KuφK−1
d +
φφ
Q.
φφ
(cid:17)
mBφ
Now, substituting the second equation into the first one of Eq. 57, one obtains
dΩ.
(63)
(64)
3. Active control analysis
In this section, a piezoelectric FG porous plate, as depicted in Fig. 3, is considered for the
active control the static and dynamic responses of the FG plates. Whereas the bottom layer is a
piezoelectric sensor labeled with the subscript s, the top layer represents a piezoelectric actuator
denoted with the subscript a. The combination between the displacement feedback control [55],
which helps the piezoelectric actuator to generate the charge, and the velocity feedback control
[54][63][64], which can provide a velocity component based on an appropriate electronic circuit,
is utilized in this study. Furthermore, a consistent method [64][65] which can predict the dynamic
responses of piezoelectric FG plate is also applied. Two constant gains Gd and Gv of the displace-
ment and velocity feedback control, respectively, are adopted in order to couple the input actuator
voltage vector φa and the output sensor voltage vector φs as follows [64]
φa = Gdφs + Gv
φs.
(65)
Assuming without any the external charge Q, the generated potential from the piezoelectric sensor
layer can be obtained from the second equation of Eq. 57
φs =(cid:2)K−1
φφ
(cid:3)
(cid:2)Kφu
(cid:3)
s
ds,
s
and the sensor charge resulted due to the deformation is determined by
Qs = [Kφu]sds.
(66)
(67)
which can be understood that when the FG plate structures deform, the electric charges are gener-
ated and gathered in the sensor layer because of the piezoelectric effect. After that, these electric
charges are amplified based on a closed loop control in order to convert into the voltage signal
before being sent and applied to the actuator layer. Due to the converse piezoelectric effect, the
strains and stresses of structures are formed which can be applied to actively control the dynamic
response of the FG porous plate.
By substituting Eqs. 65 and 66 into the second equation in Eq. 57, one obtains
(cid:2)Kφφ
(cid:3)
(cid:2)K−1
φφ
(cid:3)
(cid:2)Kφu
(cid:3)
s
a
ds − Gv
s
(cid:2)Kφφ
(cid:3)
a
(cid:2)K−1
φφ
(cid:3)
(cid:2)Kφu
(cid:3)
s
ds.
(68)
s
da − Gd
a
Qa =(cid:2)Kφu
(cid:3)
15
Then, substituting Eq. 68 into Eq. 64 yields
in which
φφ
and C is the active damping matrix which is expressed as
a
s
K∗ = Kuu + Gd
Md + C d + K∗d = F,
(cid:3)
s
,
(cid:2)Kφu
(cid:2)K−1
(cid:2)Kuφ
(cid:3)
(cid:3)
(cid:2)Kφu
(cid:2)K−1
(cid:3)
(cid:3)
φφ
.
s
C = Gv[Kuφ]a
s
Considering the effect of the structural damping, Eq. 69 can be rewritten as
Md + (C + CR) d + K∗d = F,
(69)
(70)
(71)
(72)
in which CR denotes the Rayleigh damping matrix which is defined based on a linear association
between M and Kuu as follows
CR = αRM + βRKuu,
(73)
where αR and βR are Rayleigh damping coefficients that can be defined from experiments. In this
study, the procedure in order to define the Rayleigh damping coefficients was reported in [66].
4. Numerical examples
In this study, the Newton-Rapshon iterative procedure [67] is employed to obtain the solutions
of the nonlinear problems. Accordingly, the iterations, where the solutions of current time step can
be obtained based on the solutions of previous time step, are repeated until the solutions converge.
For the geometrically nonlinear dynamic analysis of the FG plate under various dynamic loadings,
which the equations of dynamic problem depend on both the time domain and unknown displace-
ment vector, the Newmark's integration scheme [68] is selected. In all numerical examples, the
PZT-G1195N piezoelectric is employed and perfectly bonded on the top and bottom surfaces of
the FG plate structure as well as ignored the adhesive layers.
4.1. Validation analysis
In this section, various numerical studies regarding the geometrically nonlinear static and dy-
namic analyses of the isotropic as well as the piezoelectric FG square plates are carried out in
order to demonstrate the accuracy and stability of the present approach. Firstly, a fully clamped
(CCCC) isotropic square plate is considered to show the validity of the present formulation for
the geometrically nonlinear analysis. The plate is subjected to uniformly distributed load while
the width-to-thickness ratio (a/h) is taken equal to 100. The material properties of plate are
E = 3 × 107 psi and ν = 0.316. In this example, the normalized central deflection and load
parameter can be defined as w = w/h and P = q0a4/(Eh2), respectively. Table 1 presents the
normalized central deflections of the isotropic square plate which are compared with those of the
16
Levy's analytical solution [69], Urthaler and Reddy's mixed FEM using FSDT [70] and Nguyen
et al. based on IGA and refined plate theory (RPT) [71]. As can be observed that the proposed
results are in good agreement with the existing analytical solution as well as other approximate
results.
Next, in order to verify the accuracy of the proposed approach for the geometrically nonlinear
transient analysis, a fully simply supported (SSSS) orthotropic square plate subjected to uniform
load with q0 = 1.0 MPa is conducted in this example. The material properties and the geometry
of plate are considered as follows: Young's modulus E1 = 525 GPa, E2 = 21 GPa, shear modulus
G12 = G23 = G13 = 10.5 GPa, Poisson's ratio ν = 0.25, mass density ρ = 800 kg/m3, length
of the plate L = 0.25 m and thickness h = 5 mm. Fig. 4 depicts the geometrically nonlinear
transient response of the square plate subjected to uniform load. It can be seen that the present
results are in an excellent agreement with those obtained from the finite strip method, which was
reported by Chen at al. [72].
Last but not least, a cantilever piezoelectric FG square plate is exhaustively presented to
demonstrate the accuracy and validity of the present method for the static analysis of the FG
plates integrated with piezoelectric layers. The FG plate which is bonded by two piezoelectric lay-
ers on both the upper and the lower surfaces is made of aluminum oxide and Ti-6A1-4V materials
whose material properties are given in Table 2. In this study, the rule of mixture [53] is utilized
to describe the distribution of the ceramic and metal phases in core layer. The plate has a side
length a = b = 0.4 m while the thickness of the FG core layer and each piezoelectric layer are
hc = 5 mm and hp = 0.1 mm, respectively. The cantilever piezoelectric FG plate is subjected to
simultaneously a uniformly distributed load with q0=100 N/m2 and various input voltage values.
The centerline linear deflections of the piezoelectric FG square plate are plotted in Fig. 5 while the
tip node deflections are also listed in Table 3 with various material index n. The results which are
generated from the proposed method are compared with those reported in [73] using a cell-based
smoothed discrete shear gap method (CS-DSG3) based on FSDT. It can be observed that the re-
sults obtained by the present formulation generally agree well with the reference solutions.
In the next part, investigations into the geometrically nonlinear static and dynamic responses
of the piezoelectric FG porous square plate reinforced by GPLs will be presented.
4.2. Geometrically nonlinear static analysis
Firstly, the geometrically nonlinear static analysis of a piezoelectric FG plate subjected to uni-
form load with parameter load q = q0 × 103 is addressed. A SSSS piezoelectric FG square plate
which is made of aluminum oxide and Ti-6A1-4V has a side length a = b = 0.2 m, thickness of
FG core layer hc = 2 mm and thickness of each piezoelectric layer hp = 0.1 mm. Fig. 6 illustrates
the influence of the material index n on the normalized linear and nonlinear central deflections of
the piezoelectric FG plates under mechanical load. As can be seen that, by increasing the material
index n, the deflection of the piezoelectric FG plate decreases gradually. The largest deflection
is obtained when material index n = 0, where the plate consists only of Ti-6A1-4V leads to the
17
decrease in the bending stiffness. Furthermore, the values of the central deflection of the geomet-
rically nonlinear analysis are always smaller than that of the linear one and this difference reduces
with the increase of the material index.
Next, a SSSS piezoelectric FG plate with porous core layer which is constituted by combining
of two porosity distribution types and three GPL dispersion patterns, respectively, is considered
in this example. The piezoelectric FG plate is subjected to sinusoidally distributed load which
is defined as q = q0sin(πx/a)sin(πy/b) in which q0 = 1.0 MPa. The plate has a side length
a = b = 0.4 m, thickness of the FG porous core layer hc = 20 mm and thickness of each piezoelec-
tric layer hp = 1 mm. In this study, the copper is chosen as the metal matrix whose material prop-
erties are given in Table 2 while the dimensions of GPLs are lGP L = 2.5 µm, wGP L = 1.5 µm,
tGP L = 1.5 nm. Fig. 7 examines the influence of the porosity coefficients on the nonlinear de-
flection of the piezoelectric FG porous plate with GPL dispersion pattern A (ΛGP L = 1.0 wt. %)
for two porosity distribution types, respectively. It can be observed that an increase of the porosity
coefficients leads to the increase of the nonlinear deflection of the FG porous plate since the higher
density of internal pores in material yields the reduction stiffness of plate structures. In addition,
Fig. 8 depicts the effect of the weight fraction and the GPL dispersion patterns on the nonlinear
deflection of the piezoelectric FG porous plate with e0 = 0.2 and two porosity distribution types,
respectively. It can be observed that the effective stiffness of the FG porous core layer is greatly
strengthened when adding a small amount of GPLs (ΛGP L = 1.0 wt. %) into metal matrix as ev-
idenced by decreasing the nonlinear deflection of the FG plate. More importantly, the reinforcing
effect of GPLs also depends significantly on the dispersion of GPLs in material matrix. Accord-
ingly, with the same weight fraction of GPLs, the dispersion pattern A, where GPLs are dispersed
symmetric through the midplane of porous core layer, achieves the smallest nonlinear deflection
while the asymmetric dispersion pattern B provides the largest one. For further illustration, Fig. 9
depicts the variation of the nonlinear deflection of the piezoelectric FG porous plate reinforced by
GPLs which is constituted by porosity distribution 1 and three different GPL dispersion patterns
corresponding to parameter load 10, respectively.
The combination influences of two porosity distribution types and three GPL dispersion pat-
terns on the nonlinear deflection of the piezoelectric FG porous plate with ΛGP L = 1.0 wt. % and
e0 = 0.4 is also investigated. As evidently depicted in Fig. 10, for all the considered associations,
the combination between the porosity distribution 1 and the GPL dispersion pattern A obtains the
best reinforcing performance in the geometrically nonlinear static analysis of the piezoelectric FG
porous plate. This indicated that the plate structures, where the internal pores are distributed on the
midplane and GPLs are dispersed around the top and bottom surfaces, can provide the optimum
reinforcement.
4.3. Geometrically nonlinear dynamic analysis
In this part, the geometrically nonlinear dynamic responses of a CCCC piezoelectric FG porous
plate reinforced by GPLs are studied. The dimensions and the material properties of the FG plate
18
are the same previous example. The plate is assumed to be subjected to time-dependent sinu-
soidally distributed transverse loads which are expressed as follows q = q0sin(πx/a)sin(πy/b)F (t),
where F (t) is defined as
(cid:26) 1
(cid:26) 1 − t/t1
(cid:26) sin (πt/t1)
0
0
F (t) =
0
e−γt,
0 ≤ t ≤ t1,
t > t1,
0 ≤ t ≤ t1,
t > t1,
0 ≤ t ≤ t1,
t > t1,
Step load
Triangular load
Sinusoidal load
(74)
Explosive blast load
in which q0 = 100 MPa, γ = 330s−1 and the time history F (t) is plotted in Fig. 11.
Fig. 12 illustrates the influence of the porosity coefficient on the nonlinear transient re-
sponse of the piezoelectric FG porous plate with porosity distribution 1 and dispersion pattern
A (ΛGP L = 1.0 wt. %) under step and sinusoidal loads, respectively. It can be seen that by
increasing the porosity coefficients, the amplitude of the transverse deflection of the FG porous
plate can be increased while the period of motion does not seem to affect. It can be concluded
that the presence of porosities in core layer of the FG plate reduces the capacity of itself against
external excitation. Furthermore, Fig. 13 demonstrates the influence of the weight fraction and
the dispersion pattern of GPLs on the nonlinear transient response of the piezoelectric FG porous
plate with e0 = 0.2 and porosity distribution 2 corresponding to triangular and explosive blast
loads, respectively. As expected, smaller magnitude of the deflection can be obtained when the
weight fraction of GPLs in metal matrix increase. Again, the dispersion of GPLs into the metal
matrix also affects the reinforcing performance of structure that dispersion pattern A provides the
smallest magnitude of the deflection.
Next, the combination influences of various porosity distribution types and the GPL disper-
sion patterns on the nonlinear dynamic response of the piezoelectric FG plate is also examined
and indicated in Fig. 14. For this specific example, the porous core layer of the piezoelectric
plate has the porosity coefficient e0 = 0.4 and the GPL weight fraction ΛGP L = 1.0 wt. %.
As clearly demonstrated in Fig. 14, the combination between the porosity distribution 1 and the
GPL dispersion pattern A always provides the best reinforcement as evidenced by obtaining the
smallest amplitude of the deflection. Moreover, the dynamic responses of the linear and nonlinear
of the FG porous plate with porosity distribution 2 (e0 = 0.3) and the GPL dispersion pattern
C (ΛGP L = 1.0 wt. %) under triangular and sinusoidal loads are also considered and depicted
in Fig. 15. As can be observed, the geometrically nonlinear responses generally obtain smaller
magnitudes of the deflection and periods of motion.
4.4. Static and dynamic responses active control
In this section, the active control for the static and dynamic responses of the FG porous plate
reinforced by GPLs using integrated sensors and actuators is investigated. Firstly, the active con-
19
trol for the linear static responses of a SSSS FG plate which is subjected to a uniformly distributed
load with q0=100 N/m2 is investigated to perform the accuracy of the proposed approach. The
FG plate composed of Ti-6A1-4V and aluminum oxide materials with material index n = 2 has
the side length a = b = 0.2 m while thickness of core FG layer and each piezoelectric layer
are taken to be 1 mm and 0.1 mm, respectively. Fig. 16 illustrates the linear static deflections
of the FG plate with various the displacement feedback control gains Gd. As can be observed
that the present results agree well with the reference solution which is reported in [73] who em-
ployed the CS-DSG3 based on FSDT. As expected, when the displacement feedback control gain
Gd increases, the linear static deflection of the FG plate decreases. Furthermore, the active control
for the linear dynamic responses of the FG plate is also investigated based on a constant velocity
feedback control algorithm Gv and a closed loop control. In this specific example, the FG plate
is initially subjected to a uniform load q0=100 N/m2 and then the load is suddenly removed. In
this study, the modal superposition is adopted in order to reduce the computational cost and the
first six modes are considered in the modal space analysis, while the initial modal damping ratio
for each mode is assumed to be 0.8 %. Fig. 17 shows the linear dynamic responses of the central
deflection of the FG plate. The results which are generated from present method agree well with
the reference solution [73].
Next, the active control for the nonlinear static responses of the SSSS FG porous plate rein-
forced with GPLs is further investigated in this part. The FG plate consisting of combined the
porosity distribution 1 and GPL dispersion pattern A, which provides the best structural perfor-
mance, is selected to study. The material properties of the FG porous plate are the same in Section
4.2. The plate has a side length a = b = 0.4 m, thickness of the FG porous core layer hc = 20 mm
and thickness of each piezoelectric layer hp = 1 mm under sinusoidally distributed load which is
defined as q = q0sin(πx/a)sin(πy/b) with q0 = 1.0 MPa. Fig. 18 depicts the nonlinear static
deflection of the FG porous reinforced by GPLs with the porosity coefficient e0 = 0.4 and the
GPL weight fraction ΛGP L = 1.0 wt. % corresponding to various displacement feedback control
gains. As can be observed that the deflection of the FG porous plate decreases significantly when
the displacement feedback control gain increase.
In the last example, the active control for the geometrically nonlinear dynamic responses of
the CCCC FG porous plate reinforced by GPLs is conducted. The plate has the both length and
width set the same at 0.2 m with the thickness of core layer hc = 10 mm and each piezoelectric
layer hp = 0.1 mm. The FG plate with the porosity distribution 1 (e0 = 0.4) and dispersion
pattern A (ΛGP L = 1.0 wt. %) is subjected to sinusoidally distributed transverse loads which are
the same as those in Section 4.3. Fig. 19 illustrates the nonlinear dynamic responses of the central
deflection of the FG plate corresponding to various the velocity feedback control gains Gv. It can
be observed that when the control gain Gv is equal to zero corresponding to without control case,
the nonlinear dynamic response of the FG porous plate still attenuates with respect to time since
the effect of the structural damping is considered in this study. More importantly, the geometrically
nonlinear dynamic response can be suppressed more faster in the case controlled by higher velocity
feedback control gain values. As a result, depending on the specific cases, the responses of the
FG porous plate structures including deflection, oscillation time or even both can be controlled to
20
satisfy an expectation by designing an appropriate value for the velocity feedback control gain. It
should be noted that the feedback control gain values could not be increased without limit since
piezoelectric materials have their own breakdown voltage values. In addition, Fig. 20 depicts the
influence of the velocity feedback control gain Gv on the linear and nonlinear responses of the
CCCC FG porous square plate subjected to step load. As expected, the geometrically nonlinear
dynamic responses provide smaller magnitudes of the deflection and periods of motion.
5. Conclusions
In this study, the IGA based on the B´ezier extraction and the C 0-HSDT was successfully
was presented for the geometrically nonlinear static and dynamic responses for FG porous plates
with GPLs reinforcement and integration with piezoelectric layers. The equations of motion were
derived based on the C 0-HSDT in conjunction with von K´arm´am strain assumptions. Whereas the
mechanical displacement field is approximated using the C 0-HSDT based on the B´ezier extraction
of NURBS, the electric potential field was considered as a linear function through the thickness of
each piezoelectric sublayer. Two porosity distributions and three dispersion patterns of GPLs with
various related parameters were exhaustively carried out through numerical examples. The control
algorithms based on the constant displacement and velocity feedbacks were utilized to control
the geometrically nonlinear static and dynamic responses of the FG porous plate reinforced with
GPLs. Through the present numerical results, several major remarks can be drawn:
• By applying Bernstein polynomials as basis functions in the B´ezier extraction, the IGA
approach can easily be integrated into most existing FEM structures while its advantages
are maintained effectively.
• After adding a small amount of GPLs into the metal matrix, the stiffness of the structures
is significantly improved while an increase of the porosity coefficients leads to the decrease
of the reinforcing effect. Furthermore, the distribution of porosities and GPLs in metal
matrix also affect significantly the reinforcing performance of the structures. For all the
combinations, the association between the porosity distribution type 1 with internal pores
distributed on the midplane and the GPL dispersion pattern A, where GPLs are dispersed
around the top and bottom surfaces, obtained the best reinforcing performance.
• For geometrically nonlinear static responses control of the FG porous plates, two effective
algorithms are considered including the input voltage control with opposite signs applied
across the thickness of two piezoelectric layers and the displacement feedback control al-
gorithm. In addition, the dynamic response of the FG porous plate can be expectantly sup-
pressed based on the effectiveness of the velocity feedback control algorithm.
• Finally, the combination advantages of both the porous architecture and GPL reinforcement
into material matrices is a good choice to provide the advanced ultra-light high-strength
structures in engineering.
21
Acknowledgement
The support provided by RISE-project BESTOFRAC (734370)H2020 is gratefully acknowl-
edged.
References
References
[1] A. Tampieri, G. Celotti, S. Sprio, A. Delcogliano, S. Franzese, Porosity-graded hydroxyapatite ceramics to
replace natural bone, Biomaterials 22 (11) (2001) 1365 -- 1370.
[2] W. Pompe, H. Worch, M. Epple, W. Friess, M. Gelinsky, P. Greil, U. Hempel, D. Scharnweber, K. Schulte,
Functionally graded materials for biomedical applications, Materials Science and Engineering: A 362 (1-2)
(2003) 40 -- 60.
[3] L.-P. Lefebvre, J. Banhart, D. C. Dunand, Porous metals and metallic foams: current status and recent develop-
ments, Advanced Engineering Materials 10 (9) (2008) 775 -- 787.
[4] C. Betts, Benefits of metal foams and developments in modelling techniques to assess their materials behaviour:
a review, Materials Science and Technology 28 (2) (2012) 129 -- 143.
[5] B. Smith, S. Szyniszewski, J. Hajjar, B. Schafer, S. Arwade, Steel foam for structures: A review of applications,
manufacturing and material properties, Journal of Constructional Steel Research 71 (2012) 1 -- 10.
[6] X. Xia, X. Chen, Z. Zhang, X. Chen, W. Zhao, B. Liao, B. Hur, Effects of porosity and pore size on the
compressive properties of closed-cell mg alloy foam, Journal of Magnesium and Alloys 1 (4) (2013) 330 -- 335.
[7] S. Iijima, Helical microtubules of graphitic carbon, nature 354 (6348) (1991) 56.
[8] K. Liew, Z. Lei, L. Zhang, Mechanical analysis of functionally graded carbon nanotube reinforced composites:
a review, Composite Structures 120 (2015) 90 -- 97.
[9] G. Mittal, V. Dhand, K. Y. Rhee, S.-J. Park, W. R. Lee, A review on carbon nanotubes and graphene as fillers in
reinforced polymer nanocomposites, Journal of Industrial and Engineering Chemistry 21 (2015) 11 -- 25.
[10] D. G. Papageorgiou, I. A. Kinloch, R. J. Young, Mechanical properties of graphene and graphene-based
nanocomposites, Progress in Materials Science 90 (2017) 75 -- 127.
[11] L. J. Groven, J. A. Puszynski, Solution combustion synthesis of carbon nanotube loaded nickel foams, Materials
Letters 73 (2012) 126 -- 128.
[12] I. Duarte, E. Ventura, S. Olhero, J. M. Ferreira, An effective approach to reinforced closed-cell al-alloy foams
with multiwalled carbon nanotubes, Carbon 95 (2015) 589 -- 600.
[13] M. A. Rafiee, J. Rafiee, Z. Wang, H. Song, Z.-Z. Yu, N. Koratkar, Enhanced mechanical properties of nanocom-
posites at low graphene content, ACS nano 3 (12) (2009) 3884 -- 3890.
[14] I. Zaman, H.-C. Kuan, J. Dai, N. Kawashima, A. Michelmore, A. Sovi, S. Dong, L. Luong, J. Ma, From carbon
nanotubes and silicate layers to graphene platelets for polymer nanocomposites, Nanoscale 4 (15) (2012) 4578 --
4586.
[15] A. Hassani, A. Habibolahzadeh, H. Bafti, Production of graded aluminum foams via powder space holder tech-
nique, Materials & Design 40 (2012) 510 -- 515.
[16] Y. Hangai, K. Saito, T. Utsunomiya, S. Kitahara, O. Kuwazuru, N. Yoshikawa, Compression properties of al/al --
si -- cu alloy functionally graded aluminum foam fabricated by friction stir processing route, Materials Transac-
tions 54 (3) (2013) 405 -- 408.
[17] S.-Y. He, Y. Zhang, G. Dai, J.-Q. Jiang, Preparation of density-graded aluminum foam, Materials Science and
Engineering: A 618 (2014) 496 -- 499.
[18] S. Kitipornchai, D. Chen, J. Yang, Free vibration and elastic buckling of functionally graded porous beams
reinforced by graphene platelets, Materials & Design 116 (2017) 656 -- 665.
[19] D. Chen, J. Yang, S. Kitipornchai, Nonlinear vibration and postbuckling of functionally graded graphene rein-
forced porous nanocomposite beams, Composites Science and Technology 142 (2017) 235 -- 245.
[20] J. Yang, D. Chen, S. Kitipornchai, Buckling and free vibration analyses of functionally graded graphene rein-
forced porous nanocomposite plates based on chebyshev-ritz method, Composite Structures 193 (2018) 281 --
294.
22
[21] K. Li, D. Wu, X. Chen, J. Cheng, Z. Liu, W. Gao, M. Liu, Isogeometric analysis of functionally graded porous
plates reinforced by graphene platelets, Composite Structures 204 (2018) 114 -- 130.
[22] Q. Li, D. Wu, X. Chen, L. Liu, Y. Yu, W. Gao, Nonlinear vibration and dynamic buckling analyses of sand-
wich functionally graded porous plate with graphene platelet reinforcement resting on winkler -- pasternak elastic
foundation, International Journal of Mechanical Sciences 148 (2018) 596 -- 610.
[23] Z. Wang, S.-h. Chen, W. Han, The static shape control for intelligent structures, Finite elements in analysis and
design 26 (4) (1997) 303 -- 314.
[24] X. He, T. Ng, S. Sivashanker, K. Liew, Active control of fgm plates with integrated piezoelectric sensors and
actuators, International journal of Solids and Structures 38 (9) (2001) 1641 -- 1655.
[25] K. Liew, S. Sivashanker, X. He, T. Ng, The modelling and design of smart structures using functionally graded
materials and piezoelectrical sensor/actuator patches, Smart Materials and Structures 12 (4) (2003) 647.
[26] F. Ebrahimi, A. Rastgoo, Free vibration analysis of smart annular fgm plates integrated with piezoelectric layers,
Smart Materials and Structures 17 (1) (2008) 015044.
[27] R. Talebitooti, K. Daneshjoo, S. Jafari, Optimal control of laminated plate integrated with piezoelectric sensor
and actuator considering tsdt and meshfree method, European Journal of Mechanics-A/Solids 55 (2016) 199 --
211.
[28] B. Selim, L. Zhang, K. Liew, Active vibration control of fgm plates with piezoelectric layers based on reddys
higher-order shear deformation theory, Composite Structures 155 (2016) 118 -- 134.
[29] P. Phung-Van, L. V. Tran, A. Ferreira, H. Nguyen-Xuan, M. Abdel-Wahab, Nonlinear transient isogeometric
analysis of smart piezoelectric functionally graded material plates based on generalized shear deformation theory
under thermo-electro-mechanical loads, Nonlinear Dynamics 87 (2) (2017) 879 -- 894.
[30] A. Alibeigloo, Static analysis of functionally graded carbon nanotube-reinforced composite plate embedded in
piezoelectric layers by using theory of elasticity, Composite Structures 95 (2013) 612 -- 622.
[31] A. Alibeigloo, Free vibration analysis of functionally graded carbon nanotube-reinforced composite cylindrical
panel embedded in piezoelectric layers by using theory of elasticity, European Journal of Mechanics-A/Solids
44 (2014) 104 -- 115.
[32] M. Rafiee, X. He, K. Liew, Non-linear dynamic stability of piezoelectric functionally graded carbon nanotube-
reinforced composite plates with initial geometric imperfection, International Journal of Non-Linear Mechanics
59 (2014) 37 -- 51.
[33] A. Sharma, A. Kumar, C. Susheel, R. Kumar, Smart damping of functionally graded nanotube reinforced com-
posite rectangular plates, Composite Structures 155 (2016) 29 -- 44.
[34] B. Selim, L. Zhang, K. Liew, Active vibration control of cnt-reinforced composite plates with piezoelectric
layers based on reddys higher-order shear deformation theory, Composite Structures 163 (2017) 350 -- 364.
[35] K. Nguyen-Quang, T. Vo-Duy, H. Dang-Trung, T. Nguyen-Thoi, An isogeometric approach for dynamic re-
sponse of laminated fg-cnt reinforced composite plates integrated with piezoelectric layers, Computer Methods
in Applied Mechanics and Engineering 332 (2018) 25 -- 46.
[36] P. Malekzadeh, A. Setoodeh, M. Shojaee, Vibration of fg-gpls eccentric annular plates embedded in piezoelec-
tric layers using a transformed differential quadrature method, Computer Methods in Applied Mechanics and
Engineering 340 (2018) 451 -- 479.
[37] T. J. Hughes, J. A. Cottrell, Y. Bazilevs, Isogeometric analysis: Cad, finite elements, nurbs, exact geometry and
mesh refinement, Computer methods in applied mechanics and engineering 194 (39-41) (2005) 4135 -- 4195.
[38] J. A. Cottrell, T. J. Hughes, Y. Bazilevs, Isogeometric analysis: toward integration of CAD and FEA, John Wiley
[39] V. P. Nguyen, C. Anitescu, S. P. Bordas, T. Rabczuk, Isogeometric analysis: an overview and computer imple-
mentation aspects, Mathematics and Computers in Simulation 117 (2015) 89 -- 116.
[40] M. J. Borden, M. A. Scott, J. A. Evans, T. J. Hughes, Isogeometric finite element data structures based on b´ezier
extraction of nurbs, International Journal for Numerical Methods in Engineering 87 (1-5) (2011) 15 -- 47.
[41] J. N. Reddy, A simple higher-order theory for laminated composite plates, Journal of applied mechanics 51 (4)
[42] N. Senthilnathan, S. Lim, K. Lee, S. Chow, Buckling of shear-deformable plates, AIAA journal 25 (9) (1987)
& Sons, 2009.
(1984) 745 -- 752.
1268 -- 1271.
23
[43] M. Karama, K. Afaq, S. Mistou, Mechanical behaviour of laminated composite beam by the new multi-layered
laminated composite structures model with transverse shear stress continuity, International Journal of solids and
structures 40 (6) (2003) 1525 -- 1546.
[44] C. H. Thai, A. Ferreira, S. P. A. Bordas, T. Rabczuk, H. Nguyen-Xuan, Isogeometric analysis of laminated
composite and sandwich plates using a new inverse trigonometric shear deformation theory, European Journal
of Mechanics-A/Solids 43 (2014) 89 -- 108.
[45] T.-V. Vu, A. Khosravifard, M. Hematiyan, T. Q. Bui, Enhanced meshfree method with new correlation functions
for functionally graded plates using a refined inverse sin shear deformation plate theory, European Journal of
Mechanics-A/Solids.
[46] C. Shankara, N. Iyengar, A c0element for the free vibration analysis of laminated composite plates, Journal of
Sound and Vibration 191 (5) (1996) 721 -- 738.
[47] T. Kant, K. Swaminathan, Analytical solutions for the static analysis of laminated composite and sandwich
plates based on a higher order refined theory, Composite structures 56 (4) (2002) 329 -- 344.
[48] N. V. Nguyen, H. X. Nguyen, D.-H. Phan, H. Nguyen-Xuan, A polygonal finite element method for laminated
composite plates, International Journal of Mechanical Sciences 133 (2017) 863 -- 882.
[49] A. P. Roberts, E. J. Garboczi, Elastic moduli of model random three-dimensional closed-cell cellular solids,
Acta materialia 49 (2) (2001) 189 -- 197.
[50] A. Roberts, E. J. Garboczi, Computation of the linear elastic properties of random porous materials with a
wide variety of microstructure, in: Proceedings of the Royal Society of London A: Mathematical, Physical and
Engineering Sciences, Vol. 458, The Royal Society, 2002, pp. 1033 -- 1054.
[51] J. H. Affdl, J. Kardos, The halpin-tsai equations: a review, Polymer Engineering & Science 16 (5) (1976) 344 --
352.
[52] S. C. Tjong, Recent progress in the development and properties of novel metal matrix nanocomposites reinforced
with carbon nanotubes and graphene nanosheets, Materials Science and Engineering: R: Reports 74 (10) (2013)
281 -- 350.
[53] T. Nakamura, T. Wang, S. Sampath, Determination of properties of graded materials by inverse analysis and
instrumented indentation, Acta Materialia 48 (17) (2000) 4293 -- 4306.
[54] W.-S. Hwang, H. C. Park, Finite element modeling of piezoelectric sensors and actuators, AIAA journal 31 (5)
(1993) 930 -- 937.
Structures 10 (4) (2001) 637.
[55] S. Wang, S. Quek, K. Ang, Vibration control of smart piezoelectric composite plates, Smart materials and
[56] H. Tzou, C. Tseng, Distributed piezoelectric sensor/actuator design for dynamic measurement/control of dis-
tributed parameter systems: a piezoelectric finite element approach, Journal of sound and vibration 138 (1)
(1990) 17 -- 34.
[57] M. Aydogdu, A new shear deformation theory for laminated composite plates, Composite structures 89 (1)
(2009) 94 -- 101.
[58] T. N. Nguyen, C. H. Thai, H. Nguyen-Xuan, On the general framework of high order shear deformation theories
for laminated composite plate structures: a novel unified approach, International Journal of Mechanical Sciences
110 (2016) 242 -- 255.
[59] J. Reddy, Analysis of functionally graded plates, International Journal for numerical methods in engineering
47 (1-3) (2000) 663 -- 684.
[60] H. V. Do, H. Nguyen-Xuan, Limit and shakedown isogeometric analysis of structures based on b´ezier extraction,
European Journal of Mechanics-A/Solids 63 (2017) 149 -- 164.
[61] D. C. Thomas, M. A. Scott, J. A. Evans, K. Tew, E. J. Evans, B´ezier projection: a unified approach for local
projection and quadrature-free refinement and coarsening of nurbs and t-splines with particular application to
isogeometric design and analysis, Computer Methods in Applied Mechanics and Engineering 284 (2015) 55 --
105.
[62] S. Wang, A finite element model for the static and dynamic analysis of a piezoelectric bimorph, International
Journal of Solids and Structures 41 (15) (2004) 4075 -- 4096.
[63] K. Lam, X. Peng, G. Liu, J. Reddy, A finite-element model for piezoelectric composite laminates, Smart Mate-
rials and Structures 6 (5) (1997) 583.
24
[64] G. Liu, K. Dai, K. Lim, Static and vibration control of composite laminates integrated with piezoelectric sensors
and actuators using the radial point interpolation method, Smart materials and structures 13 (6) (2004) 1438.
[65] C.-H. Hong, I. Chopra, Modeling and validation of induced strain actuation of composite coupled plates, AIAA
journal 37 (3) (1999) 372 -- 377.
[66] I. Chowdhury, S. P. Dasgupta, Computation of rayleigh damping coefficients for large systems, The Electronic
[67] J. N. Reddy, An Introduction to Nonlinear Finite Element Analysis: with applications to heat transfer, fluid
Journal of Geotechnical Engineering 8 (0) (2003) 1 -- 11.
mechanics, and solid mechanics, OUP Oxford, 2014.
[68] N. M. Newmark, A method of computation for structural dynamics, Journal of the engineering mechanics divi-
sion 85 (3) (1959) 67 -- 94.
[69] S. Levy, Square plate with clamped edges under normal pressure producing large deflections.
[70] Y. Urthaler, J. Reddy, A mixed finite element for the nonlinear bending analysis of laminated composite plates
based on fsdt, Mechanics of Advanced Materials and structures 15 (5) (2008) 335 -- 354.
[71] H. X. Nguyen, E. Atroshchenko, H. Nguyen-Xuan, T. P. Vo, Geometrically nonlinear isogeometric analysis of
functionally graded microplates with the modified couple stress theory, Computers & Structures 193 (2017)
110 -- 127.
[72] J. Chen, D. Dawe, S. Wang, Nonlinear transient analysis of rectangular composite laminated plates, Composite
structures 49 (2) (2000) 129 -- 139.
[73] K. Nguyen-Quang, H. Dang-Trung, V. Ho-Huu, H. Luong-Van, T. Nguyen-Thoi, Analysis and control of fgm
plates integrated with piezoelectric sensors and actuators using cell-based smoothed discrete shear gap method
(cs-dsg3), Composite Structures 165 (2017) 115 -- 129.
Figure 1: Configuration of a piezoelectric FG porous plate reinforced by GPLs.
25
ba0qPiezoelectric layerPiezoelectric layerFG porous with GPLschphphhxyz (a) Porosity distribution types
(b) Dispersion patterns of GPLs.
Figure 2: Porosity distribution types and dispersion patterns of GPLs [18].
Figure 3: A schematic diagram of a FG porous plate with integrated piezoelectric sensors and actuators.
26
xz'1E'1E'2EPorosity distribution 12ch2ch−xz'1E'2EPorosity distribution 22ch2ch−1iS1iS0zGPLV2ch2ch−Pattern A2iS00zGPLV2ch2ch−Pattern B3iS0zGPLV2ch2ch−Pattern CSensor outputActuator inputControllerFG porous with GPLsActuator layerSensor layer Figure 4: Normalized nonlinear transient central deflection of a square orthotropic plate under the uniform load.
27
00.511.52Time(second)#10-3-0.500.511.522.53Normalizedde.ection,wFSMPresent (a) n = 0
(b) n = 0.5
Figure 5: Centerline linear deflections of the cantilever piezoelectric FG plate under the uniform loading and various
actuator input voltages with n = 0 and n = 0.5.
Figure 6: Effect of the material index n on the linear and nonlinear central deflections of the piezoelectric FG plate
under the mechanical load.
28
00.050.10.150.20.250.30.350.4xaxisdistance(m)-2.5-2-1.5-1-0.50Centerlinede.ection,w(m)#10-4CS-DSG3(0V)Present(0V)CS-DSG3(20V)Present(20V)CS-DSG3(40V)Present(40V)00.050.10.150.20.250.30.350.4xaxisdistance(m)-2.5-2-1.5-1-0.50Centerlinede.ection,w(m)#10-4CS-DSG3(0V)Present(0V)CS-DSG3(20V)Present(20V)CS-DSG3(40V)Present(40V)-20-15-10-50Loadparameter-0.5-0.4-0.3-0.2-0.10Normalizedde.ection,wn=0(Linear)n=0(Nonlinear)n=0:5(Linear)n=0:5(Nonlinear)n=2:0(Linear)n=2:0(Nonlinear)n=10(Linear)n=10(Nonlinear) (a) Porosity distribution 1
(b) Porosity distribution 2
Figure 7: Effect of the porosity coefficients on the nonlinear deflection of the piezoelectric FG porous square plate
with GPL dispersion pattern A and ΛGP L = 1.0 wt. %.
29
-20-15-10-50Loadparameter-0.35-0.3-0.25-0.2-0.15-0.1-0.050Normalizedde.ection,we0=0:0e0=0:2e0=0:4e0=0:6-20-15-10-50Loadparameter-0.4-0.35-0.3-0.25-0.2-0.15-0.1-0.050Normalizedde.ection,we0=0:0e0=0:2e0=0:4e0=0:6 (a) Porosity distribution 1
(b) Porosity distribution 2
Figure 8: Effect of the weight fractions and dispersion patterns of GPLs on the nonlinear deflection of the piezoelectric
FG porous square plate with e0 = 0.2.
30
-20-15-10-50Loadparameter-0.45-0.4-0.35-0.3-0.25-0.2-0.15-0.1-0.050Normalizedde.ection,w$GPL=0wt%PatternA($GPL=1wt%)PatternB($GPL=1wt%)PatternC($GPL=1wt%)-20-15-10-50Loadparameter-0.45-0.4-0.35-0.3-0.25-0.2-0.15-0.1-0.050Normalizedde.ection,w$GPL=0wt%PatternA($GPL=1wt%)PatternB($GPL=1wt%)PatternC($GPL=1wt%) (a) Pattern A
(b) Pattern B
Figure 9: Effect of the porosity coefficients and weight fractions of GPLs on the nonlinear deflection of piezoelectric
FG porous square plate for porosity distribution 1 and different GPL dispersion patterns.
(c) Pattern C
31
e0-0.1283$(%)-0.2418-0.2034-0.1597-0.2510.60.40.5-0.20.2Normalizedde.ection,w00-0.15-0.12-0.24-0.22-0.2-0.18-0.16-0.14e0-0.1871-0.1538-0.2418-0.2034$(%)-0.250.610.40.50.2-0.2Normalizedde.ection,w00-0.15-0.24-0.23-0.22-0.21-0.2-0.19-0.18-0.17-0.16-0.1838-0.1515-0.2418e0-0.2034$(%)-0.250.610.40.50.2-0.2Normalizedde.ection,w00-0.15-0.24-0.23-0.22-0.21-0.2-0.19-0.18-0.17-0.16 Figure 10: Effect of the porosity distributions and GPL dispersion patterns on the nonlinear deflection of the piezo-
electric FG porous square plate with e0 = 0.4 and ΛGP L = 1.0 wt. %.
Figure 11: Time history of load factor.
32
-20-15-10-50Loadparameter-0.4-0.35-0.3-0.25-0.2-0.15-0.1-0.050Normalizedde.ection,wDist.1andPatternADist.1andPatternBDist.1andPatternCDist.2andPatternADist.2andPatternBDist.2andPatternC0123456Time(second)#10-300.20.40.60.81LoadfactorSteploadTriangularloadSinusoidalloadExplosiveblastload (a) Step load
(b) Sinusoidal load
Figure 12: Effect of the porosity coefficients on the nonlinear dynamic responses of the CCCC piezoelectric FG
porous plate with GPL dispersion pattern A and ΛGP L = 1.0 wt. %.
(a) Triangular load
(b) Explosive blast load
Figure 13: Effect of the weight fractions and dispersion patterns of GPLs on the nonlinear dynamic responses of the
CCCC piezoelectric FG porous square plate with porosity distribution 2 and e0 = 0.2.
33
0123456Time(second)#10-3-1-0.8-0.6-0.4-0.200.20.40.60.8Normalizedde.ection,we0=0:0e0=0:2e0=0:4e0=0:6ForcedvibrationFreevibration0123456Time(second)#10-3-0.6-0.5-0.4-0.3-0.2-0.100.10.2Normalizedde.ection,we0=0:0e0=0:2e0=0:4e0=0:6ForcedvibrationFreevibration0123456Time(second)#10-3-1.2-1-0.8-0.6-0.4-0.200.20.40.6Normalizedde.ection,w$GPL=0:0wt:%PatternA($GPL=1:0wt:%)PatternB($GPL=1:0wt:%)PatternC($GPL=1:0wt:%)ForcedvibrationFreevibration0123456Time(second)#10-3-1-0.500.5Normalizedde.ection,w$GPL=0:0wt:%PatternA($GPL=1:0wt:%)PatternB($GPL=1:0wt:%)PatternC($GPL=1:0wt:%) (a) Step load
(b) Sinusoidal load
Figure 14: Effect of the porosity distributions and GPL dispersion patterns on the nonlinear dynamic responses of the
CCCC piezoelectric FG porous square plate with e0 = 0.4 and ΛGP L = 1.0 wt. %.
(a) Triangular load
(b) Sinusoidal load
Figure 15: Linear and nonlinear dynamic responses of the CCCC piezoelectric FG porous square plate with porosity
distribution 2 (e0 = 0.3) and dispersion pattern C (ΛGP L = 1.0 wt. %).
34
0123456Time(second)#10-3-1.3-1-0.500.50.8Normalizedde.ection,wLinearNonlinearForcedvibrationFreevibration0123456Time(second)#10-3-0.8-0.6-0.4-0.200.2Normalizedde.ection,wLinearNonlinearForcedvibrationFreevibration Figure 16: Effect of the displacement feedback control gain Gd on the linear static responses of the SSSS plate
subjected to uniformly distributed load.
Figure 17: Effect of the velocity feedback control gain Gv on the linear dynamic response of the SSSS FG square
plate.
35
00.050.10.150.2xaxisdistance(m)-3-2.5-2-1.5-1-0.50Centerlinede.ection,w(m)#10-5PresentRef.Gd=30Gd=10Gd=5Gd=000.10.20.30.40.5Time (second)-3-2-10123De.ection,w(m)#10-5Present(withoutcontrolGv=0)CS-DSG3(withoutcontrolGv=0)Present(controlgainGv=2e!4)CS-DSG3(controlgainGv=2e!4) Figure 18: Effect of the displacement feedback control gain Gd on the nonlinear static responses of the SSSS FG
porous plate with porosity distribution 1 (e0 = 0.2) and GPL dispersion pattern A (ΛGP L = 1.0 wt. %).
36
-20-15-10-50Loadparameter-0.35-0.3-0.25-0.2-0.15-0.1-0.050Normalizedde.ection,wGd=0Gd=5Gd=10Gd=30 (a) Step load
(b) Triangular load
(c) Sinusoidal load
(d) Explosive blast load
Figure 19: Effect of the velocity feedback control gain Gv on the nonlinear dynamic responses of the CCCC FG
porous square plate subjected to dynamic loadings.
37
0123456Time(second)#10-3-1.5-1-0.500.51Normalizedde.ection,wWithoutcontrolGv=0ControlgainGv=2e!3ControlgainGv=5e!3ForcedvibrationFreevibration0123456Time(second)#10-3-1.2-1-0.8-0.6-0.4-0.200.20.40.6Normalizedde.ection,wWithoutcontrolGv=0ControlgainGv=2e!3ControlgainGv=5e!3ForcedvibrationFreevibration0123456Time(second)#10-3-0.7-0.6-0.5-0.4-0.3-0.2-0.100.1Normalizedde.ection,wWithoutcontrolGv=0ControlgainGv=2e!3ControlgainGv=1e!1ForcedvibrationFreevibration0123456Time(second)#10-3-1.2-1-0.8-0.6-0.4-0.200.20.4Normalizedde.ection,wWithoutcontrolGv=0ControlgainGv=2e!3ControlgainGv=5e!3 Figure 20: Effect of the velocity feedback control gain Gv on the linear and nonlinear dynamic responses of the CCCC
FG porous square plate subjected to step load.
Table 1: Normalized central deflection w of CCCC isotropic square plate under the uniform load with a/h = 100.
P
17.79
38.3
63.4
95.0
134.9
184.0
245.0
318.0
402.0
Present Analytical [69] MXFEM [70]
0.2348
0.4663
0.6873
0.8983
1.1016
1.2960
1.4875
1.6728
1.8492
0.2328
0.4738
0.6965
0.9087
1.1130
1.3080
1.5010
1.6880
1.8660
0.237
0.471
0.695
0.912
1.121
1.323
1.521
1.714
1.902
IGA-RPT [71]
0.2365
0.4692
0.6908
0.9024
1.1060
1.3008
1.4926
1.6784
1.8552
38
0123456Time(second)#10-3-2-1.5-1-0.500.511.5Normalizedde.ection,wLinear(Gv=0)Nonlinear(Gv=0)Linear(Gv=3e!3)Nonlinear(Gv=3e!3)ForcedvibrationFreevibration Table 2: Material properties of the core and piezoelectric layers.
Properties
Elastic properties
E11 (GPa)
E22 (GPa)
E33 (GPa)
G12 (GPa)
G13 (GPa)
G23 (GPa)
ν12
ν13
ν23
Mass density
ρ (kg/m3)
Piezoelectric coefficients
k31 (m/V)
k32 (m/V)
Electric permittivity
p11 (F/m)
p22 (F/m)
p33 (F/m)
Core layer
Ti-6A1-4V Aluminum oxide Copper GPLs
Piezoelectric layer
PZT-G1195N
105.70
105.70
105.70
−
−
−
0.2981
0.2981
0.2981
4429
−
−
−
−
−
320.24
320.24
320.24
−
−
−
0.26
0.26
0.26
3750
−
−
−
−
−
130
130
130
−
−
−
0.34
0.34
0.34
8960
−
−
−
−
−
1010
1010
1010
−
−
−
0.186
0.186
0.186
1062.5
−
−
−
−
−
63.0
63.0
63.0
24.2
24.2
24.2
0.30
0.30
0.30
7600
254 × 10−12
254 × 10−12
15.3 × 10−9
15.3 × 10−9
15.3 × 10−9
Table 3: Tip node deflection of the cantilever piezoelectric FGM plate subjected to the uniform load and various input
voltages (×10−4 m).
n
Method
n = 0
Present
CS-DSG3 [73]
n = 0.5 Present
CS-DSG3 [73]
Present
n = 5
CS-DSG3 [73]
n = ∞ Present
CS-DSG3 [73]
40
Input voltages (V)
0
-2.5437
-2.5460
-1.6169
-1.6199
-1.1233
-1.1266
-0.8946
-0.8947
20
-1.3328 -0.1229
-1.3346 -0.1232
-0.8418 -0.0667
-0.8440 -0.0681
-0.5808 -0.0382
-0.5820 -0.0375
-0.4608 -0.0271
-0.4609 -0.0271
39
|
1909.06304 | 1 | 1909 | 2019-09-13T15:55:23 | Narrow optical linewidths in erbium implanted in TiO$_2$ | [
"physics.app-ph",
"cond-mat.mtrl-sci",
"quant-ph"
] | Atomic and atom-like defects in the solid-state are widely explored for quantum computers, networks and sensors. Rare earth ions are an attractive class of atomic defects that feature narrow spin and optical transitions that are isolated from the host crystal, allowing incorporation into a wide range of materials. However, the realization of long electronic spin coherence times is hampered by magnetic noise from abundant nuclear spins in the most widely studied host crystals. Here, we demonstrate that Er$^{3+}$ ions can be introduced via ion implantation into TiO$_2$, a host crystal that has not been studied extensively for rare earth ions and has a low natural abundance of nuclear spins. We observe efficient incorporation of the implanted Er$^{3+}$ into the Ti$^{4+}$ site (40% yield), and measure narrow inhomogeneous spin and optical linewidths (20 and 460 MHz, respectively) that are comparable to bulk-doped crystalline hosts for Er$^{3+}$. This work demonstrates that ion implantation is a viable path to studying rare earth ions in new hosts, and is a significant step towards realizing individually addressed rare earth ions with long spin coherence times for quantum technologies. | physics.app-ph | physics | Narrow optical linewidths in erbium implanted in TiO2
Christopher M. Phenicie1,3, Paul Stevenson1,3, Sacha Welinski1,3, Brendon C. Rose1, Abraham
T. Asfaw1, Robert J. Cava2, Stephen A. Lyon1, Nathalie P. de Leon1, and Jeff D. Thompson∗1
1Department of Electrical Engineering, Princeton University, Princeton, NJ 08544, USA
2Department of Chemistry, Princeton University, Princeton, NJ 08544, USA
3Contributed equally to this work
September 16, 2019
Abstract
Atomic and atom-like defects in the solid-state are widely explored for quantum computers, networks and
sensors. Rare earth ions are an attractive class of atomic defects that feature narrow spin and optical transitions
that are isolated from the host crystal, allowing incorporation into a wide range of materials. However, the
realization of long electronic spin coherence times is hampered by magnetic noise from abundant nuclear spins in
the most widely studied host crystals. Here, we demonstrate that Er3+ ions can be introduced via ion implantation
into TiO2, a host crystal that has not been studied extensively for rare earth ions and has a low natural abundance
of nuclear spins. We observe efficient incorporation of the implanted Er3+into the Ti4+site (40% yield), and
measure narrow inhomogeneous spin and optical linewidths (20 and 460 MHz, respectively) that are comparable
to bulk-doped crystalline hosts for Er3+. This work demonstrates that ion implantation is a viable path to
studying rare earth ions in new hosts, and is a significant step towards realizing individually addressed rare earth
ions with long spin coherence times for quantum technologies.
Rare earth ion impurities in crystalline hosts are a promising platform for quantum technologies, combining
narrow, stable optical transitions with isolated electronic or nuclear spins. Crystals doped with ensembles of rare
earth ions (REIs) have been used to demonstrate a variety of quantum memory protocols for quantum networks [1],
demonstrating coherence times of hours [2], light-matter entanglement [3] and quantum state teleportation [4].
Recent work has focused on individually addressed REIs [5, 6, 7], and has made significant steps towards spin-
photon entanglement with single ions, including the demonstration of radiatively broadened optical transitions [8]
and single-shot spin readout [9, 10] in optical nanocavities. Other efforts have focused on quantum transduction
from microwave to optical frequencies [11, 12, 13]. Among several widely studied REIs, erbium is particularly well
suited to many of these tasks, as its telecom-wavelength optical transition enables low-loss propagation in optical
fibers and integration with silicon nanophotonics [14].
There are several materials challenges to future development of REI-based quantum technologies. First, abundant
nuclear spins in the host crystal limit REI electronic spin coherence times, despite work on several mitigation strategies
including using clock states in hyperfine isotopes [15, 16] or electronic states with small magnetic moments [17]. The
vast majority of studied hosts for REIs involve at least one element with no stable spin-0 nuclear isotopes, including
all yttrium-based hosts [e.g. Y2SiO5 (YSO), Y3Al5O12 (YAG), YLiF4 and Y2O3], as well as alkali halides and
lithium niobate (a notable exception is CaWO4 [18]). Second, it is difficult to isolate single REIs in the well-studied
yttrium-based materials, because they are typically contaminated with ppm-level concentrations of all rare earth
elements [19, 14]. Finally, the incorporation of REIs into crystalline hosts typically involves bulk doping during
growth, which precludes controllable positioning and the isolation of individual defects.
Ion implantation is an established route to controllably introducing small numbers of defects into a wide range of
host materials, and in the context of quantum emitters has found recent application creating isolated color centers
in diamond [20, 21, 22, 23]. However, ion implantation has not been extensively studied in the context of REIs in
crystalline hosts, particularly in the low-density regime. Ion-implanted Er3+:YSO [24] and Gd3+:Al2O3 [25] have
been studied in electron spin resonance (ESR), while ion-implanted Ce3+:YAG [26, 27] and Pr3+:YAG [28] have been
studied in single-center confocal microscopy. In these works, increased linewidths were observed compared to bulk-
doped crystals, presumably as a result of unrepaired lattice damage resulting from ion implantation. In contrast,
localized Er3+ doping in LiNbO3 using solid diffusion has yielded bulk-like properties [29].
∗[email protected]
1
arXiv:1909.06304v1 [physics.app-ph] 13 Sep 2019
Figure 1: (a) PLE spectrum at 11 K in a high-dose sample. Stars indicate lines belonging to Z1 → Y1 − Y4 transitions of
the same Er3+site, as determined by PL (see Figure 2). (b) Spectrum at 5 K of the Z1 − Y1 transition in a low-dose sample,
demonstrating the narrowest observed inhomogeneous linewidth of 460 MHz (FWHM). (c) Fluorescence lifetime measurements
under excitation at the wavelengths indicated by arrows in (a). The long (blue dots) lifetime is 5.25 ± 0.03 ms, while the
short (orange squares) lifetime is 2.51 ± 0.04 ms. The other starred peaks in (a) exhibit the same lifetime as the blue curve.
In this work, we study the properties of implanted Er3+ in single crystal rutile TiO2. Based on its elemental
composition, this host material is largely free of trace REI impurities and has a low abundance of nuclear spins (Ti:
87% I=0, O: 99.96% I=0), which could be reduced further using isotopically enriched Ti precursors. Bulk-doped
Er3+:TiO2 has previously been observed in electron spin resonance (ESR) in the context of maser development [30, 31].
Here, we use ESR and photoluminescence (PL) spectroscopy to demonstrate that implanted Er3+ ions in TiO2
have inhomogeneous spin and optical linewidths (20 MHz and 460 MHz, respectively) that are comparable to or
smaller than typical values for bulk-doped Er3+ in most host crystals [32]. The Er3+ ions that we observe occupy
substitutional Ti4+ sites with D2h symmetry, and we hypothesize that the narrow linewidths are a consequence of
the non-polar symmetry of this site. We determine the energy of the lowest few ground and excited state crystal field
levels. For the sample processing conditions that lead to the highest yield, we find that 40 ± 16 % of the implanted
Er3+ ions are situated in Ti4+ sites.
Samples of rutile TiO2 (MTI) with (001) orientation were implanted with erbium ions with a range of energies
up to 350 keV chosen to achieve a uniform Er3+ concentration in the first 100 nm of the crystal, and total fluences
of 9× 1011 cm−2 (hereafter, "low dose") and 9× 1013 cm−2 (high dose). After implantation, some samples were
annealed in air at temperatures up to 1000oC (Table 2 and Ref. [33]). We look for evidence of implanted erbium
using photoluminescence excitation (PLE) spectroscopy in a helium cryostat, performed by scanning a chopped laser
and collecting delayed fluorescence onto a low-noise photodiode [14]. The resulting spectrum (Figure 1a) shows
several sharp absorption resonances near 1.5 µm, with the narrowest having a total linewidth of 460 MHz (Figure
1b). During the scan, we measure the fluorescence lifetime at each excitation wavelength (Figure 1c) and find that
the four starred peaks have the same lifetime (5.25± 0.03 ms), while some of the smaller peaks have a shorter lifetime
(2.51 ± 0.04 ms).
In the site symmetries possible in the rutile space group P 42/mnm, the ground (4I15/2) and excited (4I13/2)
electronic states of Er3+ split into 8 and 7 Kramers' doublets labeled Z1−8 and Y1−7, respectively (Figure 2a).
Absorption lines arise from transitions from Z1 → Yn, while fluorescence occurs primarily from Y1 → Zn, regardless
of which Y level is excited, because of rapid nonradiative relaxation to Y1 [19, 34]. We conjecture that the four starred
peaks in Figure 1a correspond to transitions from Z1 → Y1 − Y4 in the same Er3+ site. To check this hypothesis,
we filter the fluorescence through a grating spectrometer to resolve the emission wavelength in a PL measurement
(Figure 2b) while exciting at each of these peaks. The emission spectra are qualitatively similar, with a principal
2
Figure 2: (a) Level structure of Er3+:TiO2. Solid red lines indicate crystal field levels in the 4I15/2 and 4I13/2 manifolds
measured in this work (also listed in Table 1). Unobserved levels are depicted with dashed lines. Each line is a Kramers'
doublet, which splits into two sublevels when a magnetic field is applied. (b) Emission spectrum observed when exciting on
the Z1 → Y1 − Y4 transitions at the indicated wavelength λex. The linewidths are instrument-limited, with a wider slit width
for the 1498.40 nm measurement, since the fluorescence was very weak. (c) Emission spectrum with Z1 → Y1 excitation
at several temperatures, normalized to the Y1 → Z1 peak. The magenta, cyan, and yellow arrows indicate groupings of
lines corresponding to decay to Z1 − Z5 from Y1, Y2 and Y3, respectively. Lines without arrowheads are not observed. (d)
Temperature dependence of the PL intensity for different lines in (c). Dashed lines (cyan, yellow, corresponding to arrows of
the same color in (c)) are the expected PL intensities as a function of temperature assuming thermal equilibrium within the
Y manifold.
peak at 1520 nm and a series of smaller peaks at longer wavelengths, confirming this hypothesis. Furthermore, the
absence of any shorter-wavelength emission confirms that 1520 nm is the Z1 → Y1 transition.
The emission spectrum also allows several of the Z energy levels to be determined. These measurements are
performed at T = 11 K to increase the excitation efficiency by homogeneously broadening the optical transition, but
at these temperatures, several excited Y states are appreciably thermally populated and contribute extra lines to the
emission spectrum. These transitions can be separated using temperature-dependent PL spectroscopy (Figure 2c),
where lines with the same temperature dependence are interpreted to originate from the same Yi level (Figure 2d).
This allows us to assign energies to the first five ground states Z1 − Z5, and confirms the first three Y assignments
determined from Figure 1a. We additionally note that the PL line intensities show that the dominant decay pathways
for the first few excited states are Y1 to Z1, Y2 to Z2, and Y3 to Z3 and Z4.
Next, we probe the spin properties of the Z1 doublet using ESR in an X-band continuous wave (CW) spectrometer.
We identify the Er3+ peak through its characteristic high g-factor (gzz = 14.30 ± 0.42) and hyperfine spectrum
(167Er3+ with I = 7/2 and 23% abundance, Azz = 1503± 11 MHz, Figure 3a). The magnetic moment varies strongly
with the magnetic field orientation, with its maximal value along the [001] direction (c-axis). The effective g-factor in
the [100] direction is 1.63± 0.3, as determined from a fit to the orientation dependence (Figure 3b). We are unable to
directly measure the line in this orientation because it is very weak. We note that these g-factors are not consistent
with those previously reported for Er3+:TiO2 (gzz = 15.1, gxx, gyy < 0.1) [30], although the hyperfine constant is
similar to the previously reported value (A = 1484 MHz).
To probe whether the Er3+ site identified in ESR is the same as the one in the optical measurements above, we
illuminate the sample with light near the Z1 → Y1 line and observe the resulting change in the ESR signal. The
experimental apparatus for this measurement is depicted in Figure 3c. The ESR signal changes when the laser is
resonant with the Z1 → Y1 optical transition (Figure 3d) because some fraction of the Er3+ ions are transferred to
the Y1 state and are no longer resonant with the microwaves. This confirms that the optical and ESR measurements
probe the same Er3+ site. In a magnetic field, the Z1 → Y1 transition splits into four lines. Here, only two are
observed, presumably because the other two are weak or orthogonal to the light polarization direction. From the
splitting, we can extract the Y1 magnetic moment along the c-axis, ge
zz: the spacing between the peaks corresponds
3
Term Level
Z1
Z2
Z3
Z4
Z5
Z6-Z8
4I15/2
4I13/2
THz
0
1.06
3.61
5.10
8.05
Energy
nm (vac.)
-
-
-
-
-
not observed
cm−1
0
35.3
120
170
268
Y1
Y2
Y3
Y4
197.1651
197.6204
198.2888
200.0745
1520.515 6576.719
1517.012 6591.905
1511.898 6614.203
1498.404 6673.767
Y5-Y7
not observed
Table 1: Energy levels for Er3+:TiO2. The uncertainty in the Y energies is 200 MHz, determined from a wavemeter calibrated
to an acetylene absorption cell. The uncertainty in the Z levels is ∼ 10 GHz, based on fits to the data in Figure 2b.
to a difference in g-factors gg
bigger than the maximum possible value of 14.4 for 4I13/2.
zz = 2.105 ± 0.013, implying ge
zz − ge
zz = 12.19 ± 0.42 < gg
zz, since the larger value is
Figure 3: (a) CW ESR spectrum with the magnetic field nearly parallel to the c-axis of the crystal. Arrows point to hyperfine
lines in the spectrum belonging to the 167Er3+ ions. (b) Angular variation of the effective g-factor of Er3+ in the ac- (red
crosses) and bc- (blue circles) planes. (c) Schematic of the optical-ESR setup. The sample sits in a CW ESR spectrometer, and
is illuminated by a laser delivered through an optical fiber. When the laser is resonant with the Z1 → Y1 optical transition,
we observe a change in the ESR signal from population shelving in the excited state. To increase sensitivity, we measure this
optical ESR response by chopping the excitation light and measuring the modulated ESR signal via lock-in detection. (d)
The optical ESR response is resonant at the same wavelength as the Z1-Y1 spectral line in Figure 1, indicating that the spin
and optical transitions arise from the same Er3+ site. (e) Unit cell of TiO2 (rutile). We propose that Er3+ occupies a Ti site,
as shown in green.
The rotation dependence of g is essentially identical in the ac and bc crystallographic planes; however, we observe
that the single line splits into two lines when the field is rotated slightly into the ab plane. These observations
suggest that Er3+ is incorporated into a well-defined crystallographic site with two orientations differing by a 90
degree rotation around the c-axis, which is consistent with substitutional incorporation on the Ti4+ site with D2h
symmetry. This incorporation site has been suggested using similar measurement techniques for a range of transition
4
# Total fluence
1
2
3
4
5
6
9 × 1013 cm−2 As implanted
"
"
9 × 1011 cm−2 As implanted
"
"
800oC, 2 hours in air
1000oC, 2 hours in air
Annealing conditions
800oC, 2 hours in air
1000oC, 2 hours in air
Implantation yield Data shown in Figure
2 ± 1 %
5 ± 2 %
11 ± 4 %
40 ± 16 %
34 ± 14 %
40 ± 16 %
3a),b),4b)
4b)
1a), 2b),c),3c),4b),d)
1b),4c)
Table 2: List of samples used in this work.
metal impurities in TiO2, including iron [35], manganese [36] and others [30]; in contrast, nickel incorporates in an
interstitial site and gives a qualitatively different ESR spectrum [37].
Lastly, we study the dependence of the ion properties on implantation and annealing conditions. First, the
implantation yield into the Ti4+ site (measured using quantitative ESR [33]) depends strongly on the implantation
dose (Figure 4a). Without post-implantation annealing, the yield for low-dose implantation is 40 ± 16 %, while for
high-dose implantation it is only 2 ± 1 %. After annealing in air for 2 h at 1000oC, the yield for the high-dose sample
increases to 11 ± 4 %, but is unchanged in the low-dose sample. The optical lineshapes also change with implantation
and annealing conditions. Figure 4b shows the PLE spectrum of the Z1 → Y1 transition in the high-dose sample as
a function of annealing temperature, demonstrating a significant reduction in the overall peak width and confirming
the change in yield measured with ESR. Figure 4c shows the PLE spectrum of the Z1 → Y1 transition in the low-dose
and high-dose samples after 1000oC annealing, demonstrating that the inhomogeneous linewidth is much smaller in
the low-dose sample (FWHM of central feature is 0.5 vs 1.6 GHz), and that the broad, long-wavelength tail observed
in all of the PLE features in the high-dose sample (Figure 4b) is absent in the low-dose sample. We note that the
linewidth observed in the optical-ESR spectrum in the high-dose sample (Figure 3d) does not show a long-wavelength
tail. This may result from probing only a subset of un-disturbed ions selected by the ESR transition.
Figure 4: (a) Implantation yield, measured using quantitative ESR, as a function of implantation and annealing conditions.
(b) PLE spectrum at T = 11 K of the Z1 → Y1 transition in the high-dose sample after different annealing conditions. (c)
PLE spectrum at T = 5 K of the Z1 → Y1 transition in the low-dose and high-dose samples after a 1000◦C anneal. The data
are scaled to have the same peak height to facilitate comparison of the linewidths.
We now turn to a discussion of these results. Several important properties follow from the symmetry of the D2h
incorporation site. First, the optical transition should be nearly purely magnetic dipole (MD) in nature, since the
inversion-symmetric crystal field cannot mix 4f and 5d states to introduce a forced electric dipole transition. The
observed excited state lifetime of 5.25 ms is consistent with the predicted MD decay rate of 5-6.7 ms [38], depending
5
on the TiO2 refractive index for the polarization direction of the dipole (no = 2.45, ne = 2.70) [39]. Second, an
expected consequence of the non-polar nature of this site is that the wavefunctions should not have a permanent
electric dipole moment, eliminating first-order sensitivity of the optical transitions to electric fields. We hypothesize
that the observed narrow inhomogeneous linewidths are partially a consequence of this insensitivity, and that the
asymmetric inhomogeneous linewidths observed in the high-dose sample (Figure 4b,c) arise from a quadratic DC
Stark shift. In future work, it will be interesting to probe whether the lack of a permanent dipole moment leads to
long coherence of the optical transitions.
Previous studies of ion implantation in TiO2 using Rutherford backscattering have reported nearly unity substi-
tutional fractions of Sn and Hf implanted in TiO2 at fluences up to 3× 1015 cm−2 without significant lattice damage,
and that significant lattice damage from implanting La (which does not substitute Ti4+) can be reversed by annealing
above 900oC [40]. The Er3+ properties that we measure under different implantation and annealing conditions are
consistent with a picture that lattice damage caused by ion implantation adversely affects the defect properties, and
the amount of damage is proportional to the implantation dose and can be reduced by post-implantation thermal
annealing. The dependence of the implantation yield on dose is similar to reported values for Ce3+ and Pr3+ in
YAG [27, 28].
An important question about the incorporation of trivalent Er3+ into the Ti4+ site is how charge compensation is
achieved. In many materials, including alkali halides and CaF2, local charge compensation in nearest or next-nearest
neighbor sites around aliovalent defects is observed through a lowering of the symmetry and increase in the number
of spectral lines [41], while in other materials, such as CaWO4 [42], SrTiO3 [43] and TiO2 [30], this effect is weak or
absent, and the charge compensation is believed to be long-range. Since our ESR spectra are consistent with a single
crystallographic site with two orientations, we conclude that the charge compensation is remote for Er3+:TiO2. Based
on a broad ESR scan, we believe our sample has non-negligible concentrations of several transition metal impurities,
and changes in their valence state may also play a role. We cannot rule out that a small minority of Er3+sites have
a local charge-compensating defect.
We conclude with a discussion of prospects for building quantum systems out of individually addressed Er3+ ions
in TiO2 using nanophotonic circuits [14, 9]. The apparent strong branching ratio of Y1 to Z1 is favorable for Purcell
enhancement of the emission, and we note that MD Purcell enhancement can be of a similar magnitude to the more
common electric dipole enhancement [14]. Furthermore, the low implantation dose used here is already quite high
from the perspective of single ion studies, corresponding to an areal density of 104 µm−2 and an average defect spacing
of 21 nm, suggesting that high implantation yield and narrow linewidths can be expected in the relevant density
range for single-ion work. The significant reduction in the concentration of nuclear magnetic moments compared to
typical Er3+ hosts like YSO and YAG may enable extended electronic spin coherence times and open the door to
manipulating individual nuclear spins [44], and further reduction of the nuclear magnetism may be accomplished by
CVD growth [45] of isotopically enriched layers.
We note that these results, together with recent work on implanted color centers in diamond [23, 22, 21], suggest
that good defect properties can be achieved using ion implantation and thermal annealing. In contrast to doping
during growth, ion implantation allows rapid exploration of many materials and defects. This approach may be
extended to search for other hosts for rare earth ions, as well as other optically active defects, which may be
attractive for a wide range of quantum technologies including computing, communications and sensing.
Acknowledgements
We gratefully acknowledge helpful conversations with Philippe Goldner, early technical contributions to the apparatus
from Henry Ando and Zheru Qiu, and assistance with data collection from Sabrina Chern. Funding for this research
was provided by the AFOSR (contract FA9550-18-1-0334), the Eric and Wendy Schmidt Transformative Technology
Fund, the Princeton Catalysis Initiative and the Princeton Center for Complex Materials (PCCM), an NSF-funded
MRSEC (DMR-1420541). We acknowledge the use of Princeton's Imaging and Analysis Center, which is partially
supported by PCCM, as well as the Princeton Micro-Nano Fabrication Lab. C.M.P is supported by an NDSEG
graduate fellowship.
References
[1] W. Tittel, M. Afzelius, T. Chaneli´ere, R. Cone, S. Kroll, S. Moiseev, and M. Sellars, Laser & Photonics Reviews
4, 244 (2009).
6
[2] M. Zhong, M. P. Hedges, R. L. Ahlefeldt, J. G. Bartholomew, S. E. Beavan, S. M. Wittig, J. J. Longdell, and
M. J. Sellars, Nature 517, 177 (2015).
[3] C. Clausen, I. Usmani, F. Bussi´eres, N. Sangouard, M. Afzelius, H. De Riedmatten, and N. Gisin, Nature 469,
508 (2011).
[4] F. Bussi`eres, C. Clausen, A. Tiranov, B. Korzh, V. B. Verma, S. W. Nam, F. Marsili, A. Ferrier, P. Goldner,
H. Herrmann, C. Silberhorn, W. Sohler, M. Afzelius, and N. Gisin, Nature Photonics 8, 775 (2014).
[5] R. Kolesov, K. Xia, R. Reuter, R. Stohr, A. Zappe, J. Meijer, P. Hemmer, and J. Wrachtrup, Nature Commu-
nications 3, 1029 (2012).
[6] I. Nakamura, T. Yoshihiro, H. Inagawa, S. Fujiyoshi, and M. Matsushita, Scientific Reports 4, 7364 (2015).
[7] E. Eichhammer, T. Utikal, S. Gotzinger, and V. Sandoghdar, New Journal of Physics 17, 083018 (2015).
[8] T. Zhong, J. M. Kindem, J. G. Bartholomew, J. Rochman, I. Craiciu, V. Verma, S. W. Nam, F. Marsili, M. D.
Shaw, A. D. Beyer, and A. Faraon, Physical Review Letters 121, 183603 (2018).
[9] M. Raha, S. Chen, C. M. Phenicie, S. Ourari, A. M. Dibos, and J. D. Thompson, eprint arXiv:1907.09992
(2019).
[10] J. M. Kindem, A. Ruskuc, J. G. Bartholomew, J. Rochman, Y. Q. Huan, and A. Faraon, eprint arXiv:1907.12161
(2019).
[11] L. A. Williamson, Y. H. Chen, and J. J. Longdell, Physical Review Letters 113, 203601 (2014).
[12] C. O'Brien, N. Lauk, S. Blum, G. Morigi, and M. Fleischhauer, Physical Review Letters 113, 063603 (2014).
[13] X. Fernandez-Gonzalvo, Y.-H. Chen, C. Yin, S. Rogge, and J. J. Longdell, Physical Review A 92, 062313
(2015).
[14] A. M. Dibos, M. Raha, C. M. Phenicie, and J. D. Thompson, Physical Review Letters 120, 243601 (2018).
[15] D. L. McAuslan, J. G. Bartholomew, M. J. Sellars, and J. J. Longdell, Physical Review A 85, 032339 (2012).
[16] A. Ortu, A. Tiranov, S. Welinski, F. Frowis, N. Gisin, A. Ferrier, P. Goldner, and M. Afzelius, Nature Materials
17, 671 (2018).
[17] H. J. Lim, S. Welinski, A. Ferrier, P. Goldner, and J. J. Morton, Physical Review B 97, 064409 (2018).
[18] R. M. Rakhmatullin, I. N. Kurkin, G. V. Mamin, S. B. Orlinskii, M. R. Gafurov, E. I. Baibekov, B. Z. Malkin,
S. Gambarelli, S. Bertaina, and B. Barbara, Physical Review B 79, 172408 (2009).
[19] G. Liu and B. Jacquier, Spectroscopic Properties of Rare Earths in Optical Materials (Springer-Verlag Berlin
Heidelberg, 2005).
[20] D. M. Toyli, C. D. Weis, G. D. Fuchs, T. Schenkel, and D. D. Awschalom, Nano Letters 10, 3168 (2010).
[21] Y. Chu, N. P. De Leon, B. J. Shields, B. Hausmann, R. Evans, E. Togan, M. J. Burek, M. Markham, A. Stacey,
A. S. Zibrov, A. Yacoby, D. J. Twitchen, M. Loncar, H. Park, P. Maletinsky, and M. D. Lukin, Nano Letters
14, 1982 (2014).
[22] B. C. Rose, D. Huang, Z.-H. Zhang, P. Stevenson, A. M. Tyryshkin, S. Sangtawesin, S. Srinivasan, L. Loudin,
M. L. Markham, A. M. Edmonds, D. J. Twitchen, S. A. Lyon, and N. P. de Leon, Science 361, 60 (2018).
[23] A. Sipahigil, R. E. Evans, D. D. Sukachev, M. J. Burek, J. Borregaard, M. K. Bhaskar, C. T. Nguyen, J. L.
Pacheco, H. A. Atikian, C. Meuwly, R. M. Camacho, F. Jelezko, E. Bielejec, H. Park, M. Loncar, and M. D.
Lukin, Science 354, 847 (2016).
[24] S. Probst, N. Kukharchyk, H. Rotzinger, A. Tkalcec, S. Wunsch, A. D. Wieck, M. Siegel, A. V. Ustinov, and
P. A. Bushev, Applied Physics Letters 105, 162404 (2014).
[25] I. S. Wisby, S. E. De Graaf, R. Gwilliam, A. Adamyan, S. E. Kubatkin, P. J. Meeson, A. Y. Tzalenchuk, and
T. Lindstrom, Physical Review Applied 6, 024021 (2016).
7
[26] K. Xia, R. Kolesov, Y. Wang, P. Siyushev, R. Reuter, T. Kornher, N. Kukharchyk, A. D. Wieck, B. Villa,
S. Yang, and J. Wrachtrup, Physical Review Letters 115, 093602 (2015).
[27] T. Kornher, K. Xia, R. Kolesov, N. Kukharchyk, R. Reuter, P. Siyushev, R. Stohr, M. Schreck, H. W. Becker,
B. Villa, A. D. Wieck, and J. Wrachtrup, Applied Physics Letters 108, 053108 (2016).
[28] K. Groot-Berning, T. Kornher, G. Jacob, F. Stopp, S. T. Dawkins, R. Kolesov, J. Wrachtrup, K. Singer, and
F. Schmidt-Kaler, Physical Review Letters 123, 106802 (2019).
[29] C. W. Thiel, Y. Sun, R. M. Macfarlane, T. Bottger, and R. L. Cone, Journal of Physics B: Atomic, Molecular
and Optical Physics 45, 124013 (2012).
[30] Gerritsen, H.J. Paramagnetic Resonance of Transition Metal Ions in Rutile (TiO2). In Paramagnetic Resonance,
Proceedings of the First International Conference Held in Jerusalem, Jerusalem, Israel, 1962; Low, W., Ed.;
Academic Press, (1963) pp. 3-12.
[31] Sabisky, E. S.; Gerritsen, H. J. Doubly doped titanium dioxide maser element. U.S. Patent 3,275,558, September
27, 1966.
[32] C. Thiel, T. Bottger, and R. Cone, Journal of Luminescence 131, 353 (2011).
[33] See Supporting Information.
[34] T. Bottger, Y. Sun, C. W. Thiel, and R. L. Cone, Physical Review B 74, 075107 (2006).
[35] D. L. Carter and A. Okaya, Physical Review 118, 1485 (1960).
[36] H. J. Gerritsen and E. S. Sabisky, Physical Review 132, 1507 (1963).
[37] H. J. Gerritsen and E. S. Sabisky, Physical Review 125, 1853 (1962).
[38] C. M. Dodson and R. Zia, Physical Review B 86, 125102 (2012).
[39] J. Rams, A. Tejeda, and J. M. Cabrera, Journal of Applied Physics 82, 994 (1997).
[40] R. Fromknecht, I. Khubeis, and O. Meyer, Nuclear Instruments and Methods in Physics Research Section B:
Beam Interactions with Materials and Atoms 116, 109 (1996).
[41] G. D. Watkins, Physical Review 113, 79 (1959).
[42] W. B. Mims, Physical Review 168, 370 (1968).
[43] N. J. Cockroft and J. C. Wright, Physical Review B 45, 9642 (1992).
[44] L. Jiang, J. S. Hodges, J. R. Maze, P. Maurer, J. M. Taylor, D. G. Cory, P. R. Hemmer, R. L. Walsworth,
A. Yacoby, A. S. Zibrov, and M. D. Lukin, Science 326, 267 (2009).
[45] R. N. Ghoshtagore and A. J. Noreika, Journal of the Electrochemical Society 117, 1310 (1970).
8
Supporting Information for "Narrow optical linewidths in erbium
implanted in TiO2"
Christopher M. Phenicie1,3, Paul Stevenson1,3, Sacha Welinski1,3, Brendon C. Rose1, Abraham
T. Asfaw1, Robert J. Cava2, Stephen A. Lyon1, Nathalie P. de Leon1, and Jeff D. Thompson∗1
1Department of Electrical Engineering, Princeton University, Princeton, NJ 08544, USA
2Department of Chemistry, Princeton University, Princeton, NJ 08544, USA
3Contributed equally to this work
September 16, 2019
1 Experimental Details
Rutile TiO2 samples were obtained from MTI (TOb101005S1) and implanted with doubly-ionized erbium (Innovion).
The implantation was performed in multiple steps (Table S1) to give a uniform density of implanted ions over a 100
nm depth, as calculated by Stopping-Range of Ions in Matter (SRIM) simulations [1].
Optical measurements were performed using an external cavity diode laser (Toptica CTL 1500) and erbium-doped
fiber amplifier (PriTel). In PL and PLE experiments, the signal was detected by a high-gain InGaAs photodiode
(Femto OE-200); mechanical chopping (Thorlabs MC2000B) of both the excitation and detection arms was essential
to minimize scattered excitation light at the detector. For the optical-ESR measurements the excitation light was
delivered via a multimode optical fiber. The sample and the fiber were both fixed to a piece of sapphire with GE
varnish to maintain consistent illumination.
Implantation Energy / keV Fluence (High Dose) / cm−2 Fluence (Low Dose) / cm−2
10
25
50
100
150
250
350
3.8 × 1010
5 × 1010
7.5 × 1010
1 × 1011
1.3 × 1011
1.3 × 1011
3.8 × 1011
9.0 × 1011
3.8 × 1012
5 × 1012
7.5 × 1012
1 × 1013
1.3 × 1013
1.3 × 1013
3.8 × 1013
9.0 × 1013
Total
Table S1: Implantation recipe for high and low dose sample
2 Annealing conditions
The samples were annealed in air in a tube furnace (Lindberg/Blue M). The annealing temperature was reached by
ramping the temperature at a rate of ≈ 5◦C/min, after which the temperature was kept constant during 2h, before
cooling down to room temperature. This annealing process resulted in no change in the color and transparency of
the samples, while annealing at the same temperatures in vacuum darkened the sample, consistent with literature
reports of Ti4+ reduction from the creation of oxygen vacancies [2]
∗[email protected]
1
arXiv:1909.06304v1 [physics.app-ph] 13 Sep 2019
Figure S1: a) Sample arrangement in the ESR tube. b) ESR spectrum at 8 K with both the test and control samples in the
cavity. Lines corresponding to Er3+:YSO and Er3+:TiO2 are labeled. The magnetic field is oriented along the c-axis of the
TiO2 sample. c) Integrated spectrum corresponding to the ESR spectrum displayed in b).
We estimate the number of Er3+ spins on the Ti4+ site using quantitative ESR measurements. We compare the
ESR response of Er3+:TiO2 with a control sample of Er3+:YSO (10 ppm, single crystal). By measuring the control
and test samples simultaneously, we ensure that the temperature, cavity quality factor, and microwave power are
the same. For all measurements, we also checked that the ESR spectra were taken in the linear regime with respect
to microwave power and that the lines were not broadened by the magnetic field modulation. This allows us to
quantitatively relate the ESR signal to the number of spins [3].
The two samples are fixed to two different pieces of single-crystal quartz held in place by a piece of cotton (Figure
S1a)). We verified that the quartz and cotton pieces do not induce any perturbation signal at the magnetic fields used
for the measurements. The temperature was maintained at 8 K, which ensures that >99.9% of the Er3+ population
is in the lowest crystal field level, so thermal occupancy to the Z2 level is ignored.
The control sample and the orientation of the samples in the spectrometer was the same for all measurements.
The samples are oriented so that B0 points along the c-axis of the TiO2 sample and close to D1 axis of the YSO sam-
ple. This orientation makes the effective g-factors of Er3+ ions in TiO2 and in site 2 of YSO similar (gT iO2, ≈ 14.3
and gY SO, ≈ 15.1). B0 is slightly misaligned (<5◦ in the D1b plane) from D1 axis of Y2SiO5 in order to split the
two magnetically nonequivalent subsites in Y2SiO5 [4]. Figure S1b) shows an ESR spectrum where the signal from
both samples is visible. The two strongest lines correspond to the I = 0 Er3+ isotopes in site 2 in YSO. The other
four lines labeled with black arrows correspond to the −1/2,−1/2i ↔ 1/2,−1/2i and −1/2, 1/2i ↔ 1/2, 1/2i
(mS, mIi states) transitions of 167Er3+ ions in the same site. The line labeled with a red arrow corresponds to the
I = 0 Er3+ isotopes present in TiO2. The oscillating magnetic field B1 in the cavity points along a direction close to
[100] for the TiO2 sample and close to D2 for the YSO, for which the effective g-factors are gT iO2,⊥ = 1.63 (+0.4/−0)
and gY SO,⊥ = 1.86 (+0.8/ − 0.2) for the two samples, including the 5◦ misorientation.
The mass of the control sample was measured to be m = 1.2 mg using a 0.1 mg precision balance. The density
of YSO is d = 4.44 g.cm−3. The total concentration of Er3+ ions into the YSO piece is c = 1.85 × 1017 cm−3. The
number of Er3+ ions with I = 0 present in each subsite in YSO is therefore :
(1)
m d
× (1 − η) × c ×
1 2
×
1 2
nY SO,I=0 =
with η = 0.23 the isotopic abundance of 167Er3+. The two 1
subsite is equally populated in YSO [5].
2 factors come from the fact that each site and each
3
Implantation yield measurement
2
Parameter
Mass of Y2SiO5
Er3+concentration in Y2SiO5
gT iO2,⊥
gY SO,⊥
AY SO,I=0
AT iO2,I=0
Er3+ implantation dose
Sample area
Conversion efficiency
Value
1.2 mg
10 ppm
1.63
1.86
Variable
Variable
Variable
Variable
Variable
Total error Uncorrelated error
0.08
0
0.20
0.20
0.10
0.10
0.10
0.10
0.39
0
0
0.20
0
0.10
0.10
0
0.10
0.28
Table S2: Error estimations for the different parameters taken into account for conversion efficiency. Here, total error denotes
the fractional uncertainty in each parameter affecting the absolute accuracy of all the measurements, while uncorrelated error
denotes the fractional uncertainty contributing to relative errors between different samples.
The double integral of the Er3+ peak corresponding to the I = 0 isotopes of the measured sample (TiO2) was
compared to both of the Er3+ I = 0 isotopes peaks of the control sample (YSO). An example of an integrated
spectrum shown on Figure S1c). The area of the I = 0 peaks of the YSO and TiO2 samples are determined and
respectively called AY SO,I=0 and AT iO2,I=0. The number of spins in the TiO2 samples can be calculated using the
expression [6]:
gY SO,⊥
gT iO2,⊥
The conversion efficiency is calculated as the ratio nT iO2 ,I=0,measured
.
nT iO2,I=0,implanted
AT iO2,I=0
AY SO,I=0 × (
nT iO2,I=0,measured =
)2 × nY SO,I=0
(2)
The number of implanted spins nT iO2,I=0,implanted is estimated using the implantation dose multiplied by the
surface of the sample and by the isotopic abundance (1− η). The conversion efficiencies for the different samples are
displayed in the table 2 in the main text.
The error sources in measuring conversion efficiency are listed in table S2. The same reference sample was used in
all cases, so uncertainty in the mass, Er3+ concentration and gT iO2,⊥ can be neglected when considering the relative
uncertainty between samples.
References
[1] J. F. Ziegler, M. Ziegler, and J. Biersack, Nuclear Instruments and Methods in Physics Research Section B:
Beam Interactions with Materials and Atoms 268, 1818 (2010).
[2] V. M. Khomenko, K. Langer, H. Rager, and A. Fett, Physics and Chemistry of Minerals 25, 338 (1998).
[3] D. Barr, S. S. Eaton, and G. R. Eaton, in 31st Annual International EPR Symposium, Breckenridge, Colorado
(2008) pp. 1 -- 125.
[4] Y. Sun, T. Bottger, C. W. Thiel, and R. L. Cone, Physical Review B 77, 085124 (2008).
[5] T. Bottger, Y. Sun, C. W. Thiel, and R. L. Cone, Physical Review B 74, 075107 (2006).
[6] A. Abragam and B. Bleaney, Electron Paramagnetic Resonance of Transition Ions (Oxford University Press,
1970).
3
|
1806.03071 | 2 | 1806 | 2018-06-22T02:17:55 | Single Photon Source Driver Designed in ASIC | [
"physics.app-ph",
"quant-ph"
] | The single photon source is an important part of the quantum key distribution (QKD) system. At present, the single photon source is large in size and complex in structure for a lot of discrete components which are used. The miniaturization of the photon source is the tendency of the QKD system. We integrate all laser driver electronic module into one single ASIC chip, which can be used to drive the 1550nm DFB laser in random pulse mode and it can greatly reduce the volume of the single photon source. We present the design of the chip named LSD2018 and simulation results before the tape-out. The LSD2018 is fabricated with a 130 nm CMOS process and consists of a discriminator, an adjustable pulse generator, a bandgap reference, an SPI bus, and an amplitude-adjustable current pulse driver. The electronic random pulse from the driver can go 20mA to 120mA in amplitude and 400ps to 4ns in pulse width. The parameters can be set by an SPI bus. | physics.app-ph | physics |
Single Photon Source Driver Designed in ASIC
Bo Feng, Futian Liang, Xinzhe Wang, Chenxi Zhu, Yulong Zhu, and Ge Jin
Abstract -- The single photon source is an important part of the
quantum key distribution (QKD) system. At present, the single
photon source is large in size and complex in structure for a
lot of discrete components which are used. The miniaturization
of the photon source is the tendency of the QKD system. We
integrate all laser driver electronic module into one single ASIC
chip, which can be used to drive the 1550nm DFB laser in
random pulse mode and it can greatly reduce the volume of
the single photon source. We present the design of the chip
named LSD2018 and simulation results before the tape-out. The
LSD2018 is fabricated with a 130 nm CMOS process and consists
of a discriminator, an adjustable pulse generator, a bandgap
reference, an SPI bus, and an amplitude-adjustable current pulse
driver. The electronic random pulse from the driver can go 20mA
to 120mA in amplitude and 400ps to 4ns in pulse width. The
parameters can be set by an SPI bus.
Index Terms -- Semiconductor lasers, Application specific inte-
grated Circuits.
I. INTRODUCTION
T HE QKD system has been proved to be unconditionally
secure by the uncertainty principle and the no-cloning
theorem in quantum mechanics. [1][2] The ideal choice in
the QKD system is the true single photon source. However,
the suitable deterministic single photon source is still not
available. [3] We use the phase-random weak-coherent light
emitted by the distributed feedback (DFB) laser as a source.
To meet the needs of the QKD system for the source, the DFB
laser should be precisely modulated. Therefore, a precise laser
driver circuit is necessary. The 1550nm DFB laser needs a
drive current pulse signal with the frequency up to 625MHz,
the pulse width from 400ps to 800ps, and the amplitude from
20mA to 100mA. Because of the fast speed and the high
current of the drive signal, most of the single photon source
drivers we use consist of a lot of discrete components. For
the better performance, the design of the circuit board must
be very compact. To make the driver more integrated, we
designed an ASIC chip named LSD2018, a laser source driver
Manuscript received June 2, 2018. This work was supported by the National
Natural Science Foundation of China under Grants No. 61401422.
Bo Feng, Xinzhe Wang, Yulong Zhu, and Ge Jin are with State
and Electronics, University
and Technology of China, Hefei, Anhui 230026, P.R.
(e-mail:
[email protected], [email protected],
Key Laboratory
of Science
of China
[email protected], [email protected]).
of Particle Detection
Futian Liang is with Hefei National Laboratory for Physical Sciences at
the Microscale and Department of Modern Physics, University of Science
and Technology of China, Hefei, Anhui 230026, P.R. of China, and Chinese
Academy of Sciences (CAS) Center for Excellence and Synergetic Innova-
tion Center in Quantum Information and Quantum Physics, University of
Science and Technology of China, Shanghai 201315, P.R. of China (email:
[email protected]).
Chenxi Zhu is with School of Microelectronics, University of Science
and Technology of China, Hefei, Anhui 230026, P.R. of China (e-mail:
[email protected])
First author: Bo Feng, Corresponding author: Futian Liang.
Fig. 1. The structure of the LSD2018
chip used to drive the 1550nm DFB laser in random pulse
mode.
II. DESIGN SCHEME
The structure of the LSD2018 is shown on Fig. 1. The
LSD2018 consists of a discriminator, an adjustable pulse
generator, a bandgap reference, an SPI bus, and an amplitude-
adjustable current pulse driver. The discriminator compares the
input signal and the threshold and outputs a signal with the
amplitude from 0 to 1.2V. The bandgap reference provides the
reference current for the entire circuit. The SPI bus delivers the
control signals which are used to configure the adjustable pulse
generator and the amplitude-adjustable current pulse driver.
A. Adjustable Pulse Generator
The adjustable pulse generator consists of a delay line and
an AND gate. The generator receives the signals from the
discriminator and generates two time-delay signals by the
delay line. The two time-delay signals are different in delay
time and phase. So, when the two signals are added by the
AND gate, we can get a narrow pulse signal. The pulse width
of the narrow pulse signal is determined by the relative delay
of the two time-delay signals. The delay of the two time-delay
signals is configured by the SPI bus.
B. Amplitude-adjustable Current Pulse Driver
The amplitude-adjustable current pulse driver consists of
a single-ended to differential module, a step-by-step current
pulse amplification module, and a 4-bit DAC. The single-
ended to differential module receives the output signal from
DiscriminatorDelay lineSPISPIANDSingle-EndedToDifferentialStep-by-step CurrentPulseamplificationBandgapReferenceDACSPIThreholdInputLSD2018SPI busI.RefAdjustable pulse generatorAmplitude-adjustable current pulse driver Fig. 2. The layout of the LSD2018
Fig. 4.
the "eye diagram" of the simulation which the output current
frequency is 625MHz, the pulse width is 400ps, and the amplitude is 60mA
at the process corner tt and 27◦C (Because the output current signal is a pulse
signal which only consists of long 0 and short 1, the "eye diagram" looks
incomplete.)
THE DESIGN INDICATORS AND THE SIMULATION RESULTS OF THE
TABLE I
LSD2018
Fig. 3. The layout of the simulation
the adjustable pulse generator and converts it from a single-
ended signal to a differential signal. The step-by-step current
pulse amplification module converts the differential signal into
a current pulse signal and amplifies the current amplitude
step by step. Finally, it generates a signal with high current
amplitude up to 120mA to drive the laser. [4] The amplitude
of the output current can be configured by the DAC. The DAC
is configured by the SPI bus.
The LSD2018 is designed in a 130nm CMOS process. The
die size is 2.4mm × 1.2mm and consists of 4 different drivers.
They are different from each other in parameters and integra-
tion. So that, we can test more easily and comprehensively.
The layout of the LSD2018 is shown in Fig. 2.
III. SIMULATION
The LSD2018 tape-outed on May 15, 2018. The schedule
delivery date is August 2018. In the paper, we provide simula-
tion results only. We simulated across different process corners
(tt, ss, ff) and at different temperatures (0◦C, 27◦C, 85◦C). In
the simulation on layout in different conditions, the LSD2018
is fully functional. The structure of simulation is shown in
Fig. 3. Considering the influences of the parasitic parameters,
some resistances, capacitances and inductances are connected
in the rail of power and the line of input and output. We place
a pull-up resistor at the output as the 1550nm DFB laser.
Project indicators
simulation results
Maximum output frequency
Minimum output pulse width
625MHz
400ps
625MHz
400ps
Output current amplitude
20mA-100mA
20mA-120mA
The Fig. 4 is the "eye diagram" of the simulation which the
output current frequency is 625MHz, the pulse width is 400ps,
and the amplitude is 60mA at the process corner tt and 27◦C.
The off current is less than 0.6mA. The time jitter is about
20ps on the rising edge, and about 30ps on the falling edge.
The range of the amplitude change is about 2mA at 60mA.
The design indicators and the simulation results are shown in
Table I. The LSD2018 can generate a current pulse signal of
which the frequency is up to 625MHz, the amplitude is from
20mA to 120mA and the pulse width is from 400ps to 4ns. It
can satisfy the requirements of the 1550nm DFB laser.
IV. CONCLUSION
In the primary simulation, the LSD2018 is fully functional.
It can generate a drive current pulse signal with the high cur-
rent from 20mA to 120mA, the narrow pulse width from 400ps
to 4ns, and the fast speed up to 625MHz. The drive signal
can ideally drive the 1550nm DFB laser. By integrating laser
driver electronic module into one ASIC chip, the LSD2018
greatly reduces the volume of the single photon source driver.
It is the beginning of the miniaturization of the QKD system.
The full performance tests will be done when the LSD2018 is
delivered.
REFERENCES
[1] N. Gisin, G. Ribordy, W. Tittel, and H. Zbinden, "Quantum cryptography,"
Reviews of modern physics, vol. 74, no. 1, p. 145, 2002.
Laser DriverPowerDinSPI configLaser [2] H. Takesue, S. W. Nam, Q. Zhang, R. H. Hadfield, T. Honjo, K. Tamaki,
and Y. Yamamoto, "Quantum key distribution over a 40-db channel loss
using superconducting single-photon detectors," Nature photonics, vol. 1,
no. 6, p. 343, 2007.
[3] D. Rusca, A. Boaron, F. Grunefelder, A. Martin, and H. Zbinden,
"Finite-key analysis on the 1-decoy state qkd protocol," arXiv preprint
arXiv:1801.03443, 2018.
[4] R. Tao, M. Berroth, and Z. G. Wang, "Low power 10 gbit/s vcsel driver
for optical interconnect," Electronics Letters, vol. 39, no. 24, pp. 1743 --
1744, 2003.
|
1711.10321 | 1 | 1711 | 2017-11-28T14:54:15 | Deepening subwavelength acoustic resonance via metamaterials with universal broadband elliptical microstructure | [
"physics.app-ph"
] | Slow sound is a frequently exploited phenomenon that metamaterials can induce in order to permit wave energy compression, redirection, imaging, sound absorption and other special functionalities. Generally however such slow sound structures have a poor impedance match to air, particularly at low frequencies, and consequently exhibit strong transmission only in narrow frequency ranges. This therefore strongly restricts their application in wave manipulation devices. In this work we design a slow sound medium that halves the effective speed of sound in air over a wide range of low frequencies, whilst simultaneously maintaining a near impedance match to air. This is achieved with a rectangular array of cylinders of elliptical cross section, a microstructure that is motivated by combining transformation acoustics with homogenization. Microstructural parameters are optimised in order to provide the required anisotropic material properties as well as near impedance matching. We then employ this microstructure in order to halve the size of a quarter-wavelength resonator (QWR), or equivalently to halve the resonant frequency of a QWR of a given size. This provides significant space savings in the context of low-frequency tonal noise attenuation in confined environments where the absorbing material is adjacent to the region in which sound propagates, such as in a duct. We term the elliptical microstructure `universal' since it may be employed in a number of diverse applications. | physics.app-ph | physics | Deepening subwavelength acoustic resonance via
metamaterials with universal broadband elliptical
microstructure
William D. Rowley1, William J. Parnell1, I. David Abrahams2, S. Ruth Voisey3, John
Lamb3, and Nicolas Etaix3
1School of Mathematics, University of Manchester, Oxford Road, Manchester, M13 9PL, UK
2Isaac Newton Institute, University of Cambridge, 20 Clarkson Road, Cambridge CB3 0EH, UK
3Dyson Technology Limited, Tetbury Hill, Malmesbury SN16 0RP, UK
Abstract
Slow sound is a frequently exploited phenomenon that metamaterials can induce in order
to permit wave energy compression, redirection, imaging, sound absorption and other spe-
cial functionalities. Generally however such slow sound structures have a poor impedance
match to air, particularly at low frequencies, and consequently exhibit strong transmission
only in narrow frequency ranges. This therefore strongly restricts their application in wave
manipulation devices. In this work we design a slow sound medium that halves the effective
speed of sound in air over a wide range of low frequencies, whilst simultaneously maintain-
ing a near impedance match to air. This is achieved with a rectangular array of cylinders
of elliptical cross section, a microstructure that is motivated by combining transformation
acoustics with homogenization. Microstructural parameters are optimised in order to pro-
vide the required anisotropic material properties as well as near impedance matching. We
then employ this microstructure in order to halve the size of a quarter-wavelength resonator
(QWR), or equivalently to halve the resonant frequency of a QWR of a given size. This
provides significant space savings in the context of low-frequency tonal noise attenuation
in confined environments where the absorbing material is adjacent to the region in which
sound propagates, such as in a duct. We term the elliptical microstructure 'universal' since
it may be employed in a number of diverse applications.
7
1
0
2
v
o
N
8
2
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
1
v
1
2
3
0
1
.
1
1
7
1
:
v
i
X
r
a
1
The ability to shape, redirect and manipulate sound has been of interest for many decades
and the recent emergence of new composite materials and metamaterials has driven this area
forward with a multitude of exciting results including cloaking, negative refraction and lensing
[1, 2, 3, 4, 5, 6, 7]. The application of slow sound via space-coiling, labyrinthe type structures
[8, 9, 10, 11] and helical devices [12] is of significant interest since, as with slow light, the concept
has great promise in a number of scenarios. In many applications however, when the surrounding
material is air, one requires full, broadband transmission through slow sound devices in order
to enable complete manipulation of the sound field via a metamaterial. This is not the case in
general for previously developed materials [9, 12, 13].
In this Letter we use the method of transformation acoustics [14, 15] to design a medium
in which space is apparently stretched or alternatively sound is effectively slowed, whilst also
ensuring that the medium is almost-impedance matched to air. We employ a recently developed
homogenization method in order to realise this material via a microstructure consisting of rigid
elliptical cylinders arranged in a rectangular array. The explicit form of anisotropic density
provided via the homogenization scheme means that microstructures can be optimised to best
effect. This microstructure may be contrasted with perforated plates, which have been the focus
of the majority of previous studies, and have been employed to realise cloaks [16], ground cloaks
[17, 18], lensing [5] and right angle bend transformations [19]. The newly proposed elliptical
microstructure achieves a near impedance match to air at a broad range of low frequencies
and furthermore induces an effective sound speed in the heterogeneous medium that can be
reduced to one half that of air. These achievements and generalizations of the microstructure
can be exploited in a number of important applications and therefore we term the microstructure
'universal'. Here we focus on the application of low-frequency resonance and exploit our findings
to halve the size of the well known quarter-wavelength resonator (QWR) [20]. A QWR is a
side branch of an acoustic duct used to attenuate tonal noise of wavelength approximately four
times the length of the side branch; this attenuation is achieved through destructive interference.
Theoretical predictions are validated experimentally; the resonance peak is shifted close to that
predicted by the theory, associated with the reduction in sound speed in the resonator. The
amplitude of the resonance is reduced somewhat due to inherent viscous and thermal effects in
the heterogeneous structure, but the ability to significantly reduce the resonant frequency (or
equivalently reduce the size of the resonator) is undoubtedly of significant practical importance.
More broadly speaking, along with Helmholtz resonators and Herschel-Quincke tubes, QWRs
are frequently used in applications where low-frequency tonal noise needs to be attenuated in
a confined environment such as a duct. Manipulating acoustic waves in such contexts where
sound propagates orthogonal to the walls of the confined region is traditionally very difficult.
Although high-frequency sound can easily be attenuated via standard sound-absorbing foams,
low-frequency sound is more difficult to attenuate since it requires prohibitively large amounts
of such materials. The standard approach for low-frequency sound attenuation in ducts is there-
fore to employ resonators and side-branches [21, 22] and although such side branch resonators
can be useful, their size frequently prohibits practical use, although some work has been done
using standard resonators in sequence [21, 22, 23]. Generally there is a lack of work in the area
of novel metamaterial approaches applied to sound attenuation in confined spaces; furthermore
the resonators employed would often impede flow e.g. [24]. This is in contrast to sound ab-
sorption in free-space where sound is incident onto a surface: the concept of metasurfaces has
received significant attention, with many works describing structures capable of perfect sound
absorption at low-frequency e.g. [25, 26]. Unfortunately, this concept cannot be of use in the
same manner in the context of propagation in confined regions, since the metasurface would ei-
ther sit across the duct, orthogonal to the direction of the sound field, perform well but obstruct
flow, or sit parallel to the direction of sound, and be ineffective, whilst permitting flow. The
2
size restriction associated with side-branch resonators means therefore that the present finding
could be of broad significance in low-frequency noise control.
As described above, a range of necessary material properties derived through transformation
acoustics have been realised in a variety of metamaterial applications and have been experimen-
tally verified to good effect. Here, an effective elongation of space is provided by employing
one of the simplest mappings in the theory of transformation acoustics. The transforma-
tion from the virtual primed coordinate system (x(cid:48), y(cid:48), z(cid:48)) onto the physical system (x, y, z)
is x = x(cid:48), y = αy(cid:48), z = βz(cid:48). We are interested in sound propagation in the two-dimensional
xy-plane, with α < 1, which creates an apparent elongation of space via a reduction in the
sound speed (see Fig. 1(a)). Microstructure will subsequently be chosen in order to realise this
scenario. The parameter β is employed here to optimise impedance matching to air whilst still
permitting realistic microstructures.
With ρ(cid:48) and K(cid:48) denoting the acoustic density and bulk modulus of air, transformation
acoustics dictates the required material properties associated with the mapping. These required
properties take the form of an anisotropic density with components ρ(cid:63)
y and a scalar bulk
modulus K(cid:63) (see Supplemental Material). Note that we do not concern ourselves with matching
the required density in the z-direction, and so disregard any stretch or confinement in the z-
direction; this will lead to an impedance mismatch that can be quantified through the value of
y is required to be less than ρ(cid:48). This would provide a perfect impedance match
β. With β = 1, ρ(cid:63)
y, above ρ(cid:48),
but naturally raises fabrication issues. Employing β > 1 allows an increase in ρ(cid:63)
and therefore ensures more straightforward fabrication but with a slight impedance mismatch.
This approach was suggested by Popa et al. [17] in the context of the two-dimensional acoustic
ground cloak.
x and ρ(cid:63)
In order to realise the required properties we employ an array of rigid cylinders with elliptical
cross-section arranged on a rectangular lattice. The ellipse has semi-axes ax and ay and the array
has lattice spacings of Ax and Ay along the x and y directions respectively as shown in Fig. 1(b).
Parameters need to be chosen such that the effective material properties of this inhomogeneous
y and Keff , match those of the required material
x , ρeff
medium, which we shall denote as ρeff
arising from transformation acoustics, i.e. ρ(cid:63)
y and K(cid:63) = Keff, but in general
this is not possible for the application considered here. We therefore follow a similar approach
to Popa et al [17] and minimise the difference between the material properties of a structure
we can readily fabricate and those of the required material with β as an additional parameter
associated with impedance mismatch. The problem therefore possesses five parameters, two
belonging to the transformation, α and β, and three belonging to the elliptical-microstructure,
ax, ay, the semi-axes, and A = Ax/Ay the relative spacing. The lengthscale Ay is the size of
the microstructure, which is required to be much smaller than the propagating wavelengths of
interest for homogenization theory to be valid.
x = ρeff
x , ρ(cid:63)
y = ρeff
The effective properties of the array are determined by the integral equation method of
homogenization [27, 28], the benefit of which is an explicit form for the effective properties in
terms of the parameters of the problem. For fast optimisation we use the simplest version of
the method to determine the domain in which the optimal properties will reside. We then use
Comsol Multiphysics to further explore this local region of parameter space identified by the
homogenization method, including viscous and thermal losses to provide an optimal parameter
choice (see Supplemental Material).
Using this approach, we minimise the difference between the effective properties of a medium
with a given ellipse of fixed semi-axes, and those required by a general transformation over the
transformation parameters and relative lattice spacing. We find an optimal microstructure
yielding α = 0.5, with ax = 0.47Ax, ay = 0.05Ay, Ax = 2.67Ay. This gives effective densities
of approximately 1.1 and 5.0 times the density of air in the x and y directions respectively;
3
Figure 1: a) Visualisation of the transformation on the xy-plane from the uniform virtual space
onto physical space used to obtain the required material properties. b) The physical 'stretched'
space is realised by the employment of rigid elliptical cylinders placed on a rectanglar lattice as
is illustrated here together with the unit cell employed. c) Illustrating that the microstructure
realised in (b) can be employed in the quarter wavelength resonator to halve its length, whilst
ensuring that it operates at the same frequency.
(cid:113)
Keff ρeff
the bulk modulus is approximately 1.1 times that of air also. Hence in the y direction the
speed of sound (given by
y ) is reduced to approximately half that of air, while in the
x direction the speed of sound is approximately matched to that of air. It remains to choose a
value for Ay, which should be significantly smaller than the incident wavelength in order for the
homogenization theory to be valid. However, if we take Ay too small we will greatly increase
the surface area of the embedded microstructure and promote viscous and thermal losses within
our effective material, which may not be ideal (see Supplemental Material). The selection of
this final constant is application dependent. With the parameters as chosen above, we term
this a 'near-miss' microstructure, referring to its ability to closely match the required material
properties from transformation acoustics.
Having optimised the microstructure to reduce sound speed inside the inhomogeneous medium
to half that of air, we now employ this medium in the application to a QWR of length L, as
illustrated in Fig. 1(c). Without microstructure the QWR will naturally attenuate waves of
wavelength approximately λ = 4L corresponding to a frequency of attenuation f = c/λ where
λ is the wavelength of the incident wave in air and c the speed of sound in air. Here however,
by employing the slow sound medium with near impedance matching, for the same length L we
achieve f = c/2λ, hence deepening the resonant frequency significantly and effectively creating
an eighth wavelength resonator. An alternative way of viewing this is to say that with the
microstructured resonator one can attenuate waves of the same frequency with a resonator that
is half the length of a standard QWR, thus creating a space-saving resonator device.
Employing a 3D printer, we fabricated a microstructured resonator of total length L =
40mm, with cell height Ay = 10mm and with the other parameters as detailed above, for
experimental testing in a standard two port experiment [29, 30]. The associated transmission
loss, defined in the standard manner as T L = 20 log10 (pin/pout) (where pin is the total pressure
field prior to the side branch and pout is the total pressure field after the side branch) is plotted
4
a)b)c)y0x0yxαFyxaxayAyAx as a function of frequency in Fig. 2. The data shown has been post-processed using a Fourier
transform to remove noise for clarity. The blue curve in the figure shows the transmission loss
due to an ordinary air filled side branch, a QWR; we see a large amount of transmission loss
at the predicted frequency, that corresponding to a wavelength of approximately four times the
QWR's length. For the case of a single ellipse across the width of a side-branch of the same
length (red curve) we clearly see the decrease in resonant frequency by a factor of a half as
predicted, although as may be observed, the magnitude of the loss is decreased somewhat. This
reduction is almost certainly associated with viscous and thermal losses in boundary layers
in the narrow air gaps of the microstructure between the walls of the side branch and the
ellipses. This loss mechanism results in less energy returning to the resonator neck in order to
destructively interfere with the incoming wave at the resonant frequency. With this in mind a
further resonator was printed with the microstructure shifted by a half cell width in order to
have one wider air gap along the centre of the resonator rather than two narrower air gaps at
the resonator walls. This offset microstructure does indeed give significant gain in transmission
loss (green curve).
In conclusion, we have shown how to design and construct a slow sound material with near
impedance match to air. Transformation acoustics is employed to derive the required material
properties due to our prescribed coordinate stretch. Following this, the integral equation method
of homogenization was pivotal in yielding approximate analytical formulae for the material
properties of an array of elliptical cylinders on a rectangular lattice. These formulae allowed for
fast and accurate optimisation of the microstructure in order to best fit the required slow sound
properties. This approximation was then used as a starting point for numerical simulations,
vastly narrowing down the search space in which to find an optimal configuration and including
the effects of viscous and thermal losses. The developed slow sound material was used inside
a quarter wavelength resonator in order to halve its resonant frequency, or alternatively to
halve the length of a resonator required to attenuate a given tone. This specific application
has the potential for significant industrial impact, specifically improved sound quality with a
reduced form factor (patent pending [31]); while the theoretical ideas and associated universal
broadband microstructure has widespread potential for exploitation in a breadth of acoustic
devices in the future.
References
[1] Steven A Cummer, Johan Christensen, and Andrea Al`u. Controlling sound with acoustic metamaterials.
Nature Reviews Materials, 1:16001, 2016.
[2] Steven A Cummer and David Schurig. One path to acoustic cloaking. New Journal of Physics, 9(3):45,
2007.
[3] Daniel Torrent and Jos´e S´anchez-Dehesa. Acoustic cloaking in two dimensions: a feasible approach. New
Journal of Physics, 10:063015, 2008.
[4] Shu Zhang, Leilei Yin, and Nicholas Fang. Focusing ultrasound with an acoustic metamaterial network.
Phys. Rev. Letters, 102(19):194301, 2009.
[5] Johan Christensen and F Javier Garc´ıa de Abajo. Anisotropic metamaterials for full control of acoustic
waves. Phys. Rev. Letters, 108(12):124301, 2012.
[6] Nad`ege Kaina, Fabrice Lemoult, Mathias Fink, and Geoffroy Lerosey. Negative refractive index and acoustic
superlens from multiple scattering in single negative metamaterials. Nature, 525(7567):77–81, 2015.
[7] J-P Groby, W Huang, A Lardeau, and Y Aur´egan. The use of slow waves to design simple sound absorbing
materials. Journal of Applied Physics, 117(12):124903, 2015.
[8] Zixian Liang and Jensen Li. Extreme acoustic metamaterial by coiling up space. Phys. Rev. Letters,
108(11):114301, 2012.
5
Figure 2: Experimental results illustrating the transmission loss across three resonant side
branches of equal length. Branch (a), corresponding to the red curve, has a microstructure of
one ellipse wide with the ellipse centred on the branch centre. Each unit cell is 26.7mm wide
and 10mm tall; the ellipse has a major semi-axis of 12.5 mm and a minor semi-axis of only
0.5mm. There are four unit cells within the side branch making the side branch 40mm deep
with an opening of 26.7mm. Branch (b), corresponding to the green curve, once again has a
microstructure one ellipse wide, however now the ellipses are centred on the branch walls. All
dimensions are as in the previous case, the unit cell is simply shifted. Branch (c), corresponding
to the blue curve, is a standard air filled resonator for comparison. The data shown here has
had noise removed using a fast Fourier transform. The halving of the resonant frequency in the
case of a microstructured resonator is clearly visible.
[9] Zixian Liang, Tianhua Feng, Shukin Lok, Fu Liu, Kung Bo Ng, Chi Hou Chan, Jinjin Wang, Seunghoon Han,
Sangyoon Lee, and Jensen Li. Space-coiling metamaterials with double negativity and conical dispersion.
Scientific Reports, 3:1614, 2013.
[10] Yangbo Xie, Bogdan-Ioan Popa, Lucian Zigoneanu, and Steven A Cummer. Measurement of a broadband
negative index with space-coiling acoustic metamaterials. Phys. Rev. Letters, 110(17):175501, 2013.
[11] Tobias Frenzel, Jan David Brehm, Tiemo Buckmann, Robert Schittny, Muamer Kadic, and Martin Wegener.
Three-dimensional labyrinthine acoustic metamaterials. Applied Physics Letters, 103(6):061907, 2013.
[12] Xuefeng Zhu, Kun Li, Peng Zhang, Jie Zhu, Jintao Zhang, Chao Tian, and Shengchun Liu. Implementation of
dispersion-free slow acoustic wave propagation and phase engineering with helical-structured metamaterials.
6
5001000150020002500Frequency (Hz)0510Transmission Loss (dB)a)b)c) Nature Communications, 7, 2016.
[13] Alfonso Climente, Daniel Torrent, and Jos´e S´anchez-Dehesa. Sound focusing by gradient index sonic lenses.
Applied Physics Letters, 97(10):104103, 2010.
[14] Steven A. Cummer. Transformation acoustics.
In Richard V Craster and S´ebastien Guenneau, editors,
Acoustic Metamaterials: Negative Refraction, Imaging, Lensing and Cloaking, chapter 8, pages 196–218.
Springer Science & Business Media, 2012.
[15] Andrew N Norris. Acoustic cloaking theory. Proc. Roy. Soc. A, 464(2097):2411–2434, 2008.
[16] Shu Zhang, Chunguang Xia, and Nicholas Fang. Broadband acoustic cloak for ultrasound waves. Phys. Rev.
Letters, 106(2):024301, 2011.
[17] Bogdan-Ioan Popa, Lucian Zigoneanu, and Steven A Cummer. Experimental acoustic ground cloak in air.
Physical review letters, 106(25):253901, 2011.
[18] Lucian Zigoneanu, Bogdan-Ioan Popa, and Steven A Cummer. Three-dimensional broadband omnidirec-
tional acoustic ground cloak. Nature materials, 13(4):352–355, 2014.
[19] Wenjia Lu, Han Jia, Yafeng Bi, Yuzhen Yang, and Jun Yang. Design and demonstration of an acoustic
right-angle bend. The Journal of the Acoustical Society of America, 142(1):84–89, 2017.
[20] Lawrence E Kinsler, Austin R Frey, Alan B Coppens, and James V Sanders. Fundamentals of acoustics.
Fundamentals of Acoustics, 4th Edition, by Lawrence E. Kinsler, Austin R. Frey, Alan B. Coppens, James
V. Sanders, pp. 560. ISBN 0-471-84789-5. Wiley-VCH, December 1999., page 560, 1999.
[21] Sam Hyeon Lee, Choon Mahn Park, Yong Mun Seo, Zhi Guo Wang, and Chul Koo Kim. Acoustic metama-
terial with negative modulus. Journal of Physics: Condensed Matter, 21(17):175704, 2009.
[22] Xu Wang and Cheuk-Ming Mak. Wave propagation in a duct with a periodic helmholtz resonators array.
The Journal of the Acoustical Society of America, 131(2):1172–1182, 2012.
[23] Nicholas Fang, Dongjuan Xi, Jianyi Xu, Muralidhar Ambati, Werayut Srituravanich, Cheng Sun, and Xiang
Zhang. Ultrasonic metamaterials with negative modulus. Nature Materials, 5(6):452–456, 2006.
[24] Sam Hyeon Lee, Choon Mahn Park, Yong Mun Seo, Zhi Guo Wang, and Chul Koo Kim. Composite acoustic
medium with simultaneously negative density and modulus. Phys. Rev. Letters, 104(5):054301, 2010.
[25] Jun Mei, Guancong Ma, Min Yang, Zhiyu Yang, Weijia Wen, and Ping Sheng. Dark acoustic metamaterials
as super absorbers for low-frequency sound. Nature Communications, 3:756, 2012.
[26] No´e Jim´enez, Weichun Huang, Vicent Romero-Garc´ıa, Vincent Pagneux, and J-P Groby. Ultra-thin meta-
material for perfect and quasi-omnidirectional sound absorption. Applied Physics Letters, 109(12):121902,
2016.
[27] William J Parnell and I. David Abrahams. A new integral equation approach to elastodynamic homoge-
nization. Proc. Roy. Soc. A, 464(2094):1461–1482, 2008.
[28] Duncan Joyce, William J Parnell, Raphael C Assier, and I David Abrahams. An integral equation method
for the homogenization of unidirectional fibre-reinforced media; antiplane elasticity and other potential
problems. Proc. Roy. Soc. A, 473(2201):20170080, 2017.
[29] Standard test method for measurement of normal incidence sound transmission of acoustical materials based
on the transfer matrix method. (ASTM E2611 - 09), 2009.
[30] Manchar Lal Munjal. Acoustics of ducts and mufflers with application to exhaust and ventilation system
design. John Wiley & Sons, 1987.
[31] The University of Manchester. Apparatus for modifying acoustic transmission, 06 2017.
Acknowledgements WDR acknowledges the EPSRC and Dyson for funding via a KTN
Industrial CASE PhD Studentship, WJP is grateful to the EPSRC for his fellowship grant
EP/L018039/1, IDA contributed whilst in receipt of a Royal Society Wolfson Research Merit
Award, and latterly under EPSRC grant EP/K033208/I.
Author Contributions WDR carried out initial studies into transformation acoustics. WJP
and WDR worked on the associated homogenisation tools. WDR carried out analytical opti-
misation between required and effective properties and further numerical optimisation. WJP
and IDA supervised the project. The QWR application was conceived by JL. SRV acted as
the industrial supervisor to the project and facilitated all interactions between the University
of Manchester and Dyson. Initial experimental testing was carried out by WDR and SRV. Op-
timisation in Comsol Multiphysics and final experiments were carried out by NE. All authors
have discussed and contributed towards the final writing of this manuscript.
7
Supplemental material
Transformation acoustics
The simple transformation used in order to derive the necessary material properties for a 'slow
sound' or alternatively an 'effectively elongated', material is given by
x = x
(cid:48)
,
y = αy
(cid:48)
,
z = βz
(cid:48)
,
(1)
where we use the convention that primed coordinates belong to the virtual space and unprimed
to the physical. The resulting Jacobian of transformation, F , and associated determinant, J
are
∂x/∂x(cid:48) ∂x/∂y(cid:48) ∂x/∂z(cid:48)
∂y/∂x(cid:48) ∂y/∂y(cid:48) ∂y/∂z(cid:48)
∂z/∂x(cid:48) ∂z/∂y(cid:48) ∂z/∂z(cid:48)
F =
=
1 0
0
0 α 0
0 0 β
,
(2)
(3)
J = det F = αβ.
Transformation acoustics then gives the required material properties according to the formulae,
in our case
(ρ(cid:63))
−1 =
−1 =
(ρ(cid:63))
1/αβ
0
0
1
Jρ(cid:48) F F T , K(cid:63) = JK
(cid:48)
,
1
0
α/β
0
0
0
β/α
ρ(cid:48) , K(cid:63) = αβK
(4)
(5)
(cid:48)
.
We define the various directional components of the required density matrix, and restrict our
study to the first two dimensions, so that the properties to be met in order to achieve the slow
sound material are
,
(cid:48)
ρ(cid:63)
x = αβρ
(cid:48)
ρ
ρ(cid:63)
y =
,
β
α
K(cid:63) = αβK
(cid:48)
.
(6)
(7)
(8)
We will attempt to realise these two-dimensional properties by considering an inhomogeneous
medium and corresponding expressions for its effective properties derived via homogenization
theory. We consider an array of cylindrical columns in order to fabricate the medium in three
dimensions. This leads to an impedance mismatch since the explicit z density is not realised.
However this 'error' is quantifiable through the value of β, and we anticipate that provided β
remains relatively close to unity, the device should perform as required.
Homogenization
Recently a new method of homogenization called the 'Integral Equation method' (IEM) was
devised for the determination of the effective (acoustic) properties of a medium consisting of
cylindrical fibres/rods on a periodic lattice [28] based on the original work in [27]. When the rods
are rigid, have elliptical cross-section with semi-axes ax and ay along the x and y directions, and
are arranged on a rectangular lattice of period A in the x direction and unity in the y direction,
8
the effective densities are predicted at leading order in terms of the volume filling fraction φ
and given by
(cid:18) 1 − B + φ(1 + S1)
(cid:18) 1 − B + φ(1 + S2)
1 + B − φ(1 − S1)
1 + B − φ(1 − S2)
(cid:19)
(cid:19)
ρeff
x =
ρeff
y =
(cid:48)
ρ
,
(cid:48)
ρ
,
(9)
(10)
where B = (1 − )/(1 + ), = ax/ay and S1 and S2 are lattice-sums, details of which are
provided below. By a leading order approximation in terms of filling fraction, we mean that all
terms of order φ2 and higher are removed from the denominator and numerator in expressions
(9)-(10); this is not a dilute approximation, see [27, 28]. Further the well-known result
Keff = K
(cid:48)
/(1 − φ)
(11)
is also derived. The explicit form of effective densities allows for extremely rapid optimisation
of material parameters, in contrast to computational homogenization schemes, where trial-and-
error often plays a dominant role in the realisation process.
The values of S1 and S2 can be calculated according to the formulae
(cid:32)
(cid:32)
S1 = 1 −
S2 = 1 −
πA
3
π
3A
1 − 6
1 − 6
(cid:33)
(cid:17)(cid:33)
csch2(πnA)
csch2(cid:16) πn
A
∞(cid:88)
∞(cid:88)
n=1
n=1
,
,
(12)
(13)
where A = Ax/Ay and S1 and S2 are illustrated in Fig. 3. Note that for A = 1 both sums are
equal to zero. This is the case of a square lattice in which the rectangular lattice sum disappears
and the analysis is somewhat simplified.
Figure 3: Plot of the lattice sums S1 and S2 as a function of the lattice width A.
Near-miss microstructures
Due to the use of the IEM of homogenization we have approximate explicit forms for the
effective material properties of the rectangular array of elliptical rods in terms of the physical
geometry (9) - (13). From transformation acoustics we also have the functional form of the
9
012345678910−10−8−6−4−20246810AS1S2 required properties in terms of the transformation parameters (7)-(8). We may then consider
the difference between these quantities in the sense of the L2 norm and attempt to minimise
this error. Provided that this error is sufficiently small, the elliptical microstructure should
give a physical effect that is similar to that predicted by transformation acoustics with the
determined values of α and β. To this end we consider ellipses with semi-axes between 0.01 and
0.49 (in increments of 0.01 due to fabrication tolerances) of the cell side length and minimise
the quantity
(14)
(cid:17)
(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)M (cid:63)(cid:16)
x, aj
ai
M (cid:63)(cid:16)
(cid:17)
y, A
− M eff (α, β)
x, aj
ai
y, A
(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)
over the parameters A, α and β. Where M is a vector of material properties, M = (ρx, ρy, K)
while ai
y are the semi axis of the ijth ellipse considered, given by
x and aj
ai
x = Ai,
aj
y = j.
(15)
(16)
and i and j run between 0 and 0.5 in increments of 0.01.
This optimisation scheme is easily automated and indeed fast to run; we can produce plots
showing the various parameters found by the scheme in order to display the sensitivity of the
transformation that can be achieved when a small change in shape occurs. Each pixel in the
plots of Figs 4-8 represents a distinct choice of ellipse. The relative semi-axes are specified via
(15)-(16) with the i and j values specified along the x and y coordinates in the figures. These
images specify the ellipse required for each scaling and how sensitive such a microstructure is
in achieving this scaling. We show the optimisation results for the transformation properties,
α in Fig. 4 and β in Fig. 6, the necessary relative spacing A is shown in Fig. 5. The remaining
plots of this nature show the percentage error between the required and effective properties,
Fig. 7, and a desirable region in Fig. 8. This desirable region highlights all microstructure that
come within 10% of a transformation where α < 0.7 and β < 3. These are desirable since the
percentage error is low, that is the effect of the microstructure should be as predicted, α is
small, and therefore the space elongation is pronounced and β is sufficiently close to unity to
give a good impedance match. We can see that this desirable region highlights a tight cluster
in the upper right hand corner.
Once a region of desirable ellipses has been found mathematically we then use this as a first
step to inform an optimisation routine in Comsol Multiphysics. This latter simulation accounts
for viscous and thermal losses in the microstructure and its full three-dimensional geometry.
Microstructure size and viscous losses
In order for the homogenisation result given above to hold, the microstructure must be suffi-
ciently smaller than the impinging wavelength. Hence the method is applicable to any wave-
length as long as one can fabricate a structure that is sufficiently small. However if we construct
our ellipses too small then this will increase the surface area of the microstructure and promote
viscous and thermal losses in gaps generated by the presence of the ellipses. We have seen in the
main body of the paper that viscous losses are undesirable in the QWR application considered
here and so must be minimised. We therefore seek an optimal size for the microstructure, where
the homogenisation theory is as accurate as possible for the broadest range of frequencies, whilst
viscous and thermal losses are not too large. We have implemented numerical simulations of
arrays of various sizes of ellipse in the same duct in order to demonstrate this effect; we again
use Comsol Multiphysics but specifically the 'narrow region' acoustics model. This model takes
10
Figure 4: Illustrating the various α values that can be achieved by employing an elliptical
microstructure.
Figure 5: Illustrating the necessary cell width A, required for a given ellipse to achieve the α
scaling in figure 4.
11
0.10.20.30.4i0.10.20.30.4j0.20.30.40.50.60.70.80.90.10.20.30.4i0.10.20.30.4j20406080100120140160180 Figure 6: Illustrating the out of plane scaling β associated with the microstructure of the given
ellipse with array spacing given by figure 5 to attain the α scaling in figure 4
Figure 7: Plot of the percentage error between a given microstructure and the transformation
that it is closest to.
12
0.10.20.30.4i0.10.20.30.4j51015202530350.10.20.30.4i0.10.20.30.4j102030405060708090100 Figure 8: Plot of the region containing desirable ellipses (in black), which are defined by deter-
mining whether the microstructure less than 10% away from their closest transformation. Their
closest transformation also has an α < 0.7 and β < 3.
into account both viscous and thermal losses in the small air gaps between microstructure.
Results shown in Fig. 9 compare a 1 × 4, 2 × 8 and a 3 × 12 array of ellipses in a side branch of
the same length. We see that we can expect a greater transmission loss for the resonator with
larger microstructure, i.e.
less viscous and thermal damping occurs here as anticipated. All
resonators have a comparable resonant frequency despite the change in microstructure size as
predicted. Notice that the transmission loss shown in simulations is greater than that achieved
experimentally; these losses are indicative of the true experimental loss since all values will
reduce proportionally in reality, even in the case of a resonator without microstructure.
Figure 9: The simulated transmission loss across a side branch containing a 1 × 4 array, a
2 × 8 array and a 3 × 12 array of elliptical microstructure, when viscous and thermal losses are
considered in narrow air gaps.
13
0.10.20.30.4i0.10.20.30.4j50010001500200025000510Frequency(Hz)TransmissionLoss(dB)1by4array2by8array3by12array |
1905.02251 | 1 | 1905 | 2019-05-06T20:04:26 | Hybrid 1D Plasmonic/Photonic Crystals are Responsive to Escherichia Coli | [
"physics.app-ph",
"cond-mat.mtrl-sci",
"physics.bio-ph",
"physics.optics"
] | Photonic crystal-based biosensors hold great promise as valid and low-cost devices for real-time monitoring of a variety of biotargets. Given the high processability and easiness of read-out even for unskilled operators, these systems can be highly appealing for the detection of bacterial contaminants in food and water. Here, we propose a novel hybrid plasmonic/photonic device that is responsive to Escherichia coli, which is one of the most hazardous pathogenic bacterium. Our system consists of a thin layer of silver, a metal that exhibits both a plasmonic behavior and a well-known biocidal activity, on top of a solution processed 1D photonic crystal. We attribute the bio-responsivity to the modification of the dielectric properties of the silver film upon bacterial contamination, an effect that likely stems from the formation of polarization charges at the Ag/bacterium interface within a sort of bio-doping mechanism. Interestingly, this triggers a blue-shift in the photonic response. This work demonstrates that our hybrid plasmonic/photonic device can be a low-cost and portable platform for the detection of common contaminants in food and water. | physics.app-ph | physics | Hybrid 1D Plasmonic/Photonic Crystals are Responsive to
Escherichia Coli
Giuseppe Maria Paternò1†*, Liliana Moscardi1,2†, Stefano Donini1, Davide Ariodanti3, Ilka Kriegel4, Maurizio Zani2,
Emilio Parisini1, Francesco Scotognella1,2 and Guglielmo Lanzani1,2*
1Center for Nano Science and Technology@PoliMi, Istituto Italiano di Tecnologia, Via Giovanni Pascoli, 70/3, 20133 Milano, Italy;
2Dipartimento di Fisica, Politecnico di Milano, Piazza Leonardo da Vinci 32, 20133 Milano, Italy.
3Dipartimento di Chimica, Materiali e Ingegneria Chimica "Giulio Natta", Piazza Leonardo da Vinci 32, 20133 Milano, Italy.
4Department of Nanochemistry, Istituto Italiano di Tecnologia (IIT), via Morego, 30, 16163 Genova, Italy
†These authors contributed equally to this work
* Corresponding authors
Abstract
Photonic crystal-based biosensors hold great promise as valid and low-cost devices for real-time
monitoring of a variety of biotargets. Given the high processability and easiness of read-out even
for unskilled operators, these systems can be highly appealing for the detection of bacterial
contaminants in food and water. Here, we propose a novel hybrid plasmonic/photonic device that
is responsive to Escherichia coli, which is one of the most hazardous pathogenic bacterium. Our
system consists of a thin layer of silver, a metal that exhibits both a plasmonic behavior and a well-
known biocidal activity, on top of a solution processed 1D photonic crystal. We attribute the bio-
responsivity to the modification of the dielectric properties of the silver film upon bacterial
contamination, an effect that likely stems from the formation of polarization charges at the
Ag/bacterium interface within a sort of "bio-doping" mechanism. Interestingly, this triggers a blue-
shift in the photonic response. This work demonstrates that our hybrid plasmonic/photonic device
can be a low-cost and portable platform for the detection of common contaminants in food and
water.
Introduction
The integration of sensing elements with photonic crystals (PhCs) allows a simple readout
of the detection event, often based on color changing, fostering applications in portable, and cheap
technologies (1 -- 5). For instance, photonic sensing might be of particular interest for the detection
of contaminants or pathogenic bacteria in food and water, as in this case the vast majority of the
existing detection systems are relatively time- and money-consuming mostly due to complexity of
the read-out (6, 7).
Briefly, the periodicity in the dielectric constant along 1, 2 or 3 spatial dimensions gives
rise to a forbidden gap for photons (stop-band) of specific wavelengths that, in turns, confers
structural reflection colors to the material (8). Focusing on the simplest case of one-dimensional
photonic crystals (also known as Bragg stacks, BSs) in which the stop-band arises from the
alternation of layers with high/low refractive index, the structural color can be easily tuned by
varying either the dielectric contrast or the periodicity of the alternated layers (or both). This can
be achieved by the introduction of a medium within the 1D structure, as enabled by porosity at the
meso/nanoscale in the BSs (9 -- 16). To further enhance selectivity and to detect large or complex
analytes (i.e. bacteria and biomolecules) it is also possible to chemically functionalize the surface
of porous BS (17 -- 19), although such a step would hamper easy scalability of the process. To this
end, the fabrication of photonic sensors from scalable and low-cost procedures allowing fast and
reliable detection of contaminants (i.e. in food and water) is highly desirable (20).
In this context, our recent work has been focused on the development of responsive BSs
made of alternating layer of dielectric materials and electro-optical responsive plasmonic
materials, which are fabricated from easy and low-cost solution-based processes. In particular, the
integration of metal plasmonic systems in PhCs provides both high sensitivity to environmental
changes as well peculiar sensing capabilities (21, 22) due to the specific metal interactions (23 --
25). Carrier density modulation results in the change of the refractive index that ultimately
determines the photonic stop-band, thus offering a handle for easy optical detection. We have
shown that such peculiar feature of plasmonic materials can be exploited to build-up electro-optical
switches based on the photonic reflection shift upon photo-electro doping of indium tin oxide
(ITO) nanoparticles (NPs) in SiO2/ITO and TiO2/ITO photonic crystals (26, 27), and electro
doping of silver NPs in TiO2/Ag crystals (28). Furthermore, the specific interactions occurring at
the metal surface in contact with the analytes can be exploited for label-free and low-cost
(bio)sensing purposes (29, 30). It is well-known that silver films and NPs exhibit antibacterial
properties (31 -- 35). Although the exact antibacterial mechanism is still under debate (36) (see
supplementary information section for a brief discussion on Ag bactericidal mechanism), many
reports agree that electrostatic attraction is crucial for the Ag adhesion to the bacterial membrane
and to the consequent bactericidal activity possibly mediated by transmembrane ion penetration
(35, 37 -- 39). Interestingly, this might lead to a modification of Ag charge carriers density and
plasmon resonance upon bacteria/Ag interaction, for example as a result of polarization (40)
charges accumulating at the bacteria/Ag interface.
Here, we show that a novel hybrid plasmonic/photonic device consisting of a thin layer of
silver deposited on top of a solution processed BS is responsive to one of the most common
bacterial contaminant, namely Escherichia coli. Our data suggest that the increase in the plasmon
charge density likely originates from the formation of polarization charges at the bacterium/Ag
interface, resulting in a blue-shift of the plasmon resonance. This eventually determines a change
in the photonic read-out (blue-shift) that translates the plasmonic effect occurring in the UV/blue
(330 - 440 nm) to the more convenient spectral region (600-530 nm). These promising results
indicate that hybrid plasmonic/photonic PhCs can represent a novel class of low-cost devices
responsive to common contaminants in food and water.
Results and discussion
Hybrid plasmonic/photonic devices
The multilayered 1D photonic structures show the expected structural color in reflection (5
× SiO2/TiO2 bilayers) as shown in figure 1a-b, while electron microscopy images are reported in
figure S1a. The BSs were fabricated via simple spin-coating deposition of the respective aqueous
colloidal dispersions. This is a key point in the view to scale the process by means of large-area
and low-cost deposition techniques, such as ink-jet printing and roll-to-roll. On top of the dielectric
BS we deposited a thin layer of silver (8 nm, Fig. S1b for the electron microscopy image), to
exploit both the plasmonic behaviour and the marked and well-documented bioactivity. The thin
silver layer is in-fact a defective cap layer of the photonic crystal that affects the optical response
of the BS through the silver free carrier density (Drude model) (41). Therefore, the main idea here
is to exploit the possible change in the silver complex dielectric function driven by Ag/bacteria
interaction to modify the dielectric properties at the BS/metal interface and, thus, the BS optical
read-out. To this end, we firstly selected the minimum Ag thickness achievable with our deposition
apparatus to localize strongly the plasmonic response in close proximity to the BS interface. To
observe both the plasmonic and the photonic contributions to the overall sample transmission and
disentangle them, we carried out measurements as a function of incidence angle (Fig. 1c). These
data show a blue-shift of the photonic band-gap (PBG, 586 nm at 0°) by increasing the angle in
agreement with the Bragg-Snell law (inset Fig. 1c) (21), while the plasmonic peak at 500 nm does
not display any angular dependence.
Figure 1. Hybrid plasmonic/photonic devices. (A) Picture of the fabricated 1D photonic crystal with a 8 nm silver capping layer
and (B) sketch of the multilayered structure. (C) Light transmission of the Ag/(SiO2/TiO2)5 photonic crystals as a function of the
light incidence angle (inset Bragg-Snell equation). PBG stands for photonic band-gap
Plasmonic response upon E. coli contamination
To evaluate the effect of bacteria on the optical properties of silver, we first exposed Ag
films to LB only (control experiment) and then to E. coli (Fig. 2a) in agar plate, as described in the
experimental section. We observe that the sample exposed to LB undergoes a substantial red-shift
(+ 60 nm) that is likely due to the infiltration of the aqueous culture medium across the silver
grains, leading to an increase in the effective refractive index (42). On the other hand, when the
Ag layer is contaminated with E. coli in LB medium we note the concomitant increase (+25%) of
the plasmonic absorption at the high energy side (330 - 440 nm) and an attenuated red-shift (+ 35
nm) with respect to LB only exposure. Taken together, these data indicate an overall blue-shift (-
25 nm) in the Ag plasmonic response upon contamination with E. coli. Furthermore, to mimic
exposure to contaminated liquid samples, we dipped them either in LB medium (control) or in an
LB/E. coli mixture with increasing bacterial loading (0.1, 0.5 and 1.2 OD600nm) and measured their
plasmonic response (Fig. S2). Here, we essentially observed an analogous effect, with an increased
plasmonic blue-shift upon exposure to bacteria that, interestingly, can be already noticed at the
lowest loading (-15 nm at 0.1 OD600nm). In this scenario, we hypothesize that the bacteria-induced
blue-shift in the plasmon resonance could stem from the formation of polarization charges at the
silver-bacterium interface, i.e. negative on bacterial membrane (38) and positive charges on Ag
surface respectively (see Fig. 2c), finally leading to an overall increase of the charge carrier density
(26, 28). By employing the Lorentz-Drude model (Fig. 2d), we estimated that the 25 nm blue-shift
observed experimentally corresponds to a 15% increase in the charge carried density of Ag in LB
medium.
Figure 2. Plasmonic response upon bacterial contamination. (A) Average optical absorption (with standard deviation) of Ag
thin films (8 nm on glass) before (green) and after exposure to either LB (orange line) or to (B) LB/E. coli (blue line) on agar plate
(12 samples, two replicas). (C) Pictorial representation of the hypothesized mechanism giving rise to the plasmonic blue-shift
(polarization accumulation charge). (D) Calculated transmission of the Ag layer for the pristine and "bio-doped" Ag layer in LB
medium obtained by using the Lorentz - Drude model.
Photonic response upon E. coli contamination
After having investigated the Ag plasmonic response of metallic film alone upon
contamination with E. coli, we proceeded to study how this is affecting the all optical read-out in
presence of the photonic crystal (Fig. 3). Qualitatively, both exposure to LB or E. coli + LB leads
to the same behavior, namely decreased transmittance, resonance broadening (20 nm) and red-
shift of the stop-band. Again we have two physical phenomena concurring to these changes: i)
infiltration of LB inside the porous BS architecture, leading to an enhanced effective refractive
index and ii) the modulation of the plasmon resonance in the top metal layer. The magnitude of
the PBG red-shift, however, results decreased in presence of E. coli in LB, featuring an average
shift of 5 nm compared to LB alone with + 15 nm. Interestingly, this yields an overall PBG blue-
shift of -10 nm upon contamination, which essentially translates the plasmonic effect observed in
the UV/blue region into the green/red part of the spectrum. The presence of a well-defined zone
of inhibition matching the shape of our samples in conformal contact with the agar medium
inoculated with E. coli confirms the anti-bacterial activity of our Ag layer (Fig. S3). Furthermore,
this effect seems to be confined within the sample area, suggesting that silver does not appreciably
detach from the surface and diffuse in all the agar plate as it occurs with silver NPs (43), which
may be of importance when considering possible applications of these devices in food packaging
(44). To study the optical response of Ag/PhCs in contact with contaminated liquid specimens, we
also immersed our samples in LB medium or E. coli/LB (10 minutes) and measured their optical
transmission (Fig. S4). Although in this case the effect might appear less evident due the massive
and unavoidable LB infiltration throughout the porous structure, we can still observe a decreased
red-shift for the contaminated samples when compared to the LB case, an effect that can be already
noted at the lowest bacterial loading (-10 nm for 0.1 OD600nm) in analogy with the plasmonic
response. Finally, as a control experiment, we repeated the same procedure in contaminated agar
plate on identical 1D photonic crystals but without the top Ag layer, in order to further prove the
role of the plasmonic material in the bio-responsivity of our device (Fig. 3c). Here, we could not
discriminate any difference between the samples exposed to LB or to LB/E. coli, with the
exception of the usual read-shift and broadening of the stop-band that are connected to liquid
condensation and infiltration throughout the porous structure. This confirms that the upper
plasmonic layer represents the responsivity element of our hybrid plasmonic/photonic device, as
the modification of the dielectric properties at the Ag/BS interface, which in turns is brought about
by Ag/bacteria interaction, and governs the total optical read-out.
Figure 3. Hybrid plasmonic/photonic response of hybrid Ag (8 nm) /1D PhCs upon contamination with E. coli. (A) Average
transmittance spectrum (with standard deviation) of Ag/1D PhC after exposure to the LB medium (red-line) and (B) E. coli (blue-
line). (C) Average shift at the stop-band maximum for the LB and E. coli contaminated PhCs. Data were averaged over two sets of
measurement (six samples per measurement). (C) Average transmittance spectrum (with standard deviation) of Silica/Titania 1D
PhC after exposure to the LB medium (red-line) and E. coli (blue-line). (D) Calculated transmission spectrum for the Ag/BS
structure in LB medium for 𝑛 = 5.76 × 1028 𝑚3 and 6.76 × 1028 𝑚3.
To obtain insights into the mechanism underpinning the photonic shift, we calculated the
transmission spectra as a function of the silver carrier density (Fig. 3d). For this, we combined the
transfer matrix method to model the alternating refractive index of the periodic structure, with the
Maxwell-Garnett effective medium approximation for the description of the effective refractive
indexes of the SiO2/TiO2 layers soaked with LB (see experimental section), In addition, we made
use of the Lorentz-Drude model to account for the plasmonic contribution to the overall dielectric
response of the device. We then proceeded to the simulation of the transmission spectra for 𝑛 =
5.76 × 1028 𝑚3 and 6.76 × 1028 𝑚3, with the former charge carrier density accounting for
pristine Ag and the latter for the "bio-doped" film in the LB infiltrated photonic structure. Indeed,
while an increased carrier density induces a large blue-shift of the plasmon resonance (25 nm) as
it has been shown in the previous section, the PBG exhibits a less obvious 5 nm blue-shift. Notably,
such a behavior corroborates our experimental data, at least from the qualitative point of view.
Such discrepancy can be probably attributed to the contribution of a different effect that intensifies
the dielectric mechanism, for instance a strong field enhancement and confinement at the
metal/PhC interface (i.e. Tamm optical modes) (45, 46) that cannot be taken into account by our
model. To preliminary assess this possible scenario, we also studied the photonic response of our
hybrid PhCs with a top Ag layer of 16 nm (Fig. 4). In this case, we observe a more pronounced
contribution of the Ag plasmon to the overall transmission of the hybrid structure due to its higher
optical density than in the previous samples (Fig. 4a). Remarkably, despite the plasmon resonance
still shows a total 10 nm blue-shift after contamination when compared with the LB-exposed
samples, the PBG shift is limited to - 5 nm. This might imply the involvement of an enhancement
mechanism when Ag thickness is kept at relatively low value, likely due to the strong confinement
of the plasmon at the dielectric interface (45 -- 47). Further experiments are needed to elucidate such
a mechanism.
Figure 4. Plasmonic/photonic response of hybrid Ag (16 nm)/1D PhCs. (A) Light transmission of the Ag/(SiO2/TiO2)5 photonic
crystals as a function of the light incidence angle. (B) Average transmittance spectrum (with standard deviation) of Ag/1D PhC
after exposure to the LB medium (red-line) and (B) E. coli (blue-line). Data were averaged over two sets of measurement (six
samples per measurement).
Conclusions
To summarize, we have shown that a novel hybrid plasmonic/photonic device consisting
of a thin layer of the plasmonic and biocidal silver on top of a 1D photonic crystal can be responsive
against one of the most hazardous bacterial contaminant in food and water, Escherichia coli. Our
data points towards a scenario in which the polarization charge at the Ag/bacterium interface
causes an increase of the metal charge carrier density that leads to a plasmon blue-shift, within a
sort of "bio-doping" mechanism. This, in turns, leads a photonic blue-shift in the visible spectral
range. It is worth adding that the photonic band-gap can be placed in any spectral region by a
judicious choice of the layer thicknesses, allowing one to translate the photonic read-out in the
most convenient part of the spectrum. This, taken together with the quick processability from
solution and easiness of the read-out, makes these devices promising for low-cost and real-time
monitoring of contaminants in food and water.
Experimental section
Photonic crystals fabrication. Porous 1D PhCs were fabricated by alternating layers of SiO2 and
TiO2 via spin-casting deposition from their colloidal aqueous dispersions following the procedure
employed in past experiments.(26, 27) Firstly, we suspended the TiO2 (Gentech Nanomaterials,
average size 5 nm) and SiO2 nanoparticles (Sigma Aldrich LUDOX SM-30, average size 8 nm) in
MilliQ distilled water to obtain a concentration of 5 wt. %. The dispersions were then sonicated
for 2 h at 45 °C (Bandelin SONOREX Digital 10 P) and filtered with a 0.45 μm PVDF filter. The
glass substrates were previously cleaned by means of ultra-sonication in isopropanol (10 min) and
acetone (10 min), and then subjected to an oxygen plasma treatment (Colibrì Gambetti, 10 min) to
increase wettability. During the fabrication of the PhCs, the dispersions were continuously kept in
sonication at 45 °C to maintain homogeneity of the dispersions during the whole process. The
fabrication of the crystals was performed by alternating the deposition of the two materials through
spin-coating (Laurell WS-400-6NNP-Lite) with a speed of 2000 rpm. After each deposition, the
samples were annealed on a hot-plate for 20 min at 350 °C. Finally, we deposited an 8 nm-thick
silver layer on top of the photonic structure (or glass substrate only) via thermal evaporation
(MBRAUN metal evaporator).
Bacterial culture. A single colony from the Escherichia coli Rosetta (DE3) strain carrying a
pET23a (+) plasmid (Novagen) was inoculated in Luria Bertani (LB) broth in the presence of
Ampicillin (50 μg/ml) and incubated overnight at 37ºC with shaking at 200 rpm until stationary
phase was reached. Then, bacterial suspension turbidity (expressed as optical density at 600nm;
O.D. 600) was diluted to O.D. 600 ~ 0.5 in LB broth (no antibiotic). The suspension (500 μL) was
spread over an LB agar plate. The Ag/PhCs were placed at the center of the Petri dish with the top
silver layer facing the contaminated surface (or LB only for the control experiment) and incubated
for 24 h at 37 °C. We also dipped the devices in either LB only (control) or an LB/E.coli mixture
to mimic exposure to contaminated liquid samples (0.1, 0.5 and 1.2 O.D.). The same protocol was
also repeated for silver thin films on glass substrates in order to understand the effect of bacteria
exposure on the silver plasmon resonance. Data were averaged over two sets of measurements (six
samples per measurement).
Optical characterization. The optical characterization was performed using a spectrophotometer
(Perkin Elmer Lambda 1050 WB), measuring the percentage loss of transmittance after exposure
to LB or LB/E. coli. To disentangle the photonic from the plasmonic contribution to the overall
transmission of the sample, we recorded the transmission as a function of the incidence angle,
showing a blue-shift of the photonic stop-band upon increase of the angle, in accordance with the
Bragg-Snell law.
Scanning electron microscopy. We used a Tescan MIRA3. The measurements were performed
at a voltage of 5 kV and backscattered electrons were detected. The sample was covered with
carbon paste to improve conductivity
Transfer matrix method. Refractive indexes of the Ag layers, and the SiO2 and TiO2 layers
composing the photonic crystal: we employ the combination of Drude model and Lorentz model
to describe the plasmonic response(48 -- 50). Thus, the frequency dependent complex dielectric
function of silver can be written as:
𝜀𝐴𝑔,𝜔 = 𝜀∞ −
2
𝜔𝐷
𝜔(𝜔+𝑖Γ𝐷)
−
2
∆𝜀 𝜔𝐿
𝜔2−𝜔𝐿
2+𝑖Γ𝐿𝜔
(1)
where 𝜔 is in rad/s. The first term 𝜀∞ is the high frequency dielectric constant (𝜀∞ = 0.1148 in
this work(48)). The second term is the Drude model part, where 𝜔𝐷 is the Drude plasma frequency:
𝜔𝐷 = √
𝑁𝑒2
𝑚∗𝜀0
(2)
with N number of charges, e the electron charge, 𝑚∗ the effective mass and 𝜀0 the vacuum
dielectric constant. For silver, we use 𝑁 = 5.76 × 1028 𝑐ℎ𝑎𝑟𝑔𝑒𝑠 𝑚3⁄
and 𝑚∗ = 0.96 𝑚0 𝑘𝑔
⁄
(51). Γ𝐷 is the damping coefficient (in this work Γ𝐷 = 7.055 × 1015 𝑟𝑎𝑑/𝑠 (48)).
The third term of Equation 1 is the Lorentz model part with a single Lorentz term. ∆𝜀 is the
oscillator strength, 𝜔𝐿 the Lorentz plasma frequency, Γ𝐿 is the damping coefficient (in this work
∆𝜀 = 3.6276, 𝜔𝐿 = 1.5812 × 1016 𝑟𝑎𝑑/𝑠, Γ𝐿 = 1.0463 × 1014 𝑟𝑎𝑑/𝑠 (48)).
The wavelength dependent refractive index of TiO2 can be written(52):
𝑛𝑇𝑖𝑂2(𝜆) = (4.99 +
1
96.6𝜆1.1 +
1
4.60𝜆1.95)
1 2⁄
(3)
where 𝜆 is the wavelength [in micrometers, and 𝜀𝑇𝑖𝑂2(𝜆) = 𝑛𝑇𝑖𝑂2
2
(𝜆)]. Instead, the wavelength
dependent refractive index of SiO2 can be described by the following Sellmeier equation (53):
2
𝑛𝑆𝑖𝑂2
(𝜆) − 1 =
0.6961663𝜆2
𝜆2−0.06840432 +
0.4079426𝜆2
𝜆2−0.11624142 +
0.8974794𝜆2
𝜆2−9.8961612
(4)
where 𝜆 is the wavelength in micrometers [also for SiO2 𝜀𝑆𝑖𝑂2(𝜆) = 𝑛𝑆𝑖𝑂2
2
(𝜆)].
Taking into account the infiltration of LB in the silver layer and in the silica and titania layers of
the photonic crystal, we determine the effective dielectric function of the SiO2:air layer (we call it
𝜀𝑒𝑓𝑓2,𝜔) by using the Maxwell Garnett effective medium approximation (54, 55):
𝜀𝑒𝑓𝑓,𝜔 = 𝜀𝐿𝐵
2(1−𝑓)𝜀𝐿𝐵+(1+2𝑓)𝜀𝜔
2(2+𝑓)𝜀𝐿𝐵+(1−𝑓)𝜀𝜔
(5)
where 𝑓 is the filling factor of the silver, SiO2, or TiO2 layers and 𝜀𝜔 is the dielectric function of
the silver, SiO2, or TiO2 layers. In this study we choose 𝑓𝐴𝑔 = 0.5 ; 𝑓𝑆𝑖𝑂2 = 0.6 ; 𝑓𝑇𝑖𝑂2 = 0.6. The
dielectric constant in the visible range for LB is approximated to the one of water, thus 𝜀𝐿𝐵 =
1.769.
Transmission of the multilayer photonic crystal: We use the two effective refractive indexes
of the Ag:LB, SiO2:LB, TiO2:LB layers to study the light transmission through the photonic
structure by employ the transfer matrix method (56, 57). For a transverse electric (TE) wave
the transfer matrix for the kth layer is given by
𝑐𝑜𝑠 (
2𝜋
𝜆
𝑛𝑘𝑑𝑘)
−
𝑖
𝑛𝑘
𝑠𝑖𝑛 (
2𝜋
𝜆
𝑛𝑘𝑑𝑘)
𝑀𝑘 = [
−𝑖𝑛𝑘𝑠𝑖𝑛 (
2𝜋
𝜆
𝑛𝑘𝑑𝑘)
𝑐𝑜𝑠 (
2𝜋
𝜆
𝑛𝑘𝑑𝑘)
]
(6)
with nk the refractive index and dk the thickness of the layer. In this study the thickness of the
Ag:LB layers is 8 nm, while the thickness of the SiO2:LB layers and TiO2:LB layers is 100 nm.
The product 𝑀 = 𝑀1 ⋅ 𝑀2 ⋅ … ⋅ 𝑀𝑘 ⋅ … ⋅ 𝑀𝑠 = [
𝑚11 𝑚12
𝑚21 𝑚22
] gives the matrix of the multilayer (of
s layers). The transmission coefficient is
𝑡 =
(𝑚11+𝑚12𝑛0)𝑛𝑠+(𝑚21+𝑚22𝑛0)
2𝑛𝑠
(7)
with ns the refractive index of the substrate (in this study 𝑛𝑠 = 1.46) and n0 the refractive
index of air. Thus, the light transmission of the multilayer photonic crystal is
𝑇 =
𝑛0
𝑛𝑠
𝑡2
(8)
Author information
Corresponding authors
E-mail: [email protected] ; [email protected]
Authors contribution
G.M.P., L.M., S.D. and D.A. carried out the experiments. G.M.P. wrote the manuscript, together
with L.M., S.D., G.L. and F.S. I.K., M.Z. and F.S. run the simulations. E.M supervised the
microbiology work. G.M.P., G.L. and F.S. conceived and supervised the work. G.M.P and L.M.
contributed equally to this work.
Acknowledgements
This project has received funding from the European Research Council (ERC) under the European
Union's Horizon 2020 research and innovation programme (grant agreement No. [816313]). We
thank Dr. Stefano Perissinotto for the scanning electron microscopy experiments on 1D PhCs cross
section and Ag films.
Conflict of interest
The authors declare no conflict of interest.
Keywords
Responsive Photonics; Plasmonics; Bacterial contaminants; Antibacterial; Silver
References
1.
J. Ge, Y. Yin, Responsive Photonic Crystals. Angew. Chemie Int. Ed. 50, 1492 -- 1522 (2011).
2.
L. Nucara, F. Greco, V. Mattoli, Electrically responsive photonic crystals: A review. J.
Mater. Chem. C. 3, 8449 -- 8467 (2015).
3. H. Inan et al., Photonic crystals: emerging biosensors and their promise for point-of-care
applications. Chem. Soc. Rev. 46, 366 -- 388 (2017).
4.
C. Fenzl, T. Hirsch, O. S. Wolfbeis, Photonic crystals for chemical sensing and biosensing.
Angew. Chemie - Int. Ed. 53, 3318 -- 3335 (2014).
5. H. Wang, K. Q. Zhang, Photonic crystal structures with tunable structure color as
colorimetric sensors. Sensors (Switzerland). 13 (2013), pp. 4192 -- 4213.
6. Meulebroeck, Thienpont, Ottevaere, Photonics enhanced sensors for food monitoring: part
1. IEEE Instrum. Meas. Mag. 19, 35 -- 45 (2016).
7. W. Meulebroeck, H. Thienpont, H. Ottevaere, Photonics enhanced sensors for food
monitoring: Part 3. IEEE Instrum. Meas. Mag. 20, 31 -- 37 (2017).
8.
E. Yablonovitch, Inhibited spontaneous emission in solid-state physics and electronics.
Phys. Rev. Lett. 58, 2059 -- 2062 (1987).
9. V. Robbiano et al., Room-Temperature Low-Threshold Lasing from Monolithically
Integrated Nanostructured Porous Silicon Hybrid Microcavities. ACS Nano. 12, 4536 -- 4544
(2018).
10. L. D. Bonifacio, B. V. Lotsch, D. P. Puzzo, F. Scotognella, G. A. Ozin, Stacking the
nanochemistry deck: Structural And compositional diversity in one-dimensional photonic
crystals. Adv. Mater. 21, 1641 -- 1646 (2009).
11. L. Passoni et al., Self-Assembled Hierarchical Nanostructures for High-Efficiency Porous
Photonic Crystals. ACS Nano. 8, 12167 -- 12174 (2014).
12. M. M. Orosco, C. Pacholski, M. J. Sailor, Real-time monitoring of enzyme activity in a
mesoporous silicon double layer. Nat. Nanotechnol. 4, 255 -- 258 (2009).
13. S. Colodrero, M. Ocaña, H. Míguez, Nanoparticle-based one-dimensional photonic crystals.
Langmuir. 24, 4430 -- 4434 (2008).
14. S. Y. Choi, M. Mamak, G. von Freymann, N. Chopra, G. A. Ozin, Mesoporous Bragg Stack
Color Tunable Sensors. Nano Lett. 6, 2456 -- 2461 (2006).
15. M. C. Fuertes et al., Photonic Crystals from Ordered Mesoporous Thin-Film Functional
Building Blocks. Adv. Funct. Mater. 17, 1247 -- 1254 (2007).
16. V. G. Pedro, M. E. Calvo, H. Míguez, Á. Maquieira, Nanoparticle Bragg reflectors: A smart
analytical tool for biosensing. Biosens. Bioelectron. X, 100012 (2019).
17. S. Chan, S. R. Horner, P. M. Fauchet, B. L. Miller, Identification of Gram negative bacteria
using nanoscale silicon microcavities [1]. J. Am. Chem. Soc. 123, 11797 -- 11798 (2001).
18. L. D. Bonifacio et al., Towards the Photonic Nose: A Novel Platform for Molecule and
Bacteria Identification. Adv. Mater. 22, 1351 -- 1354 (2010).
19. L. D. Bonifacio, G. A. Ozin, A. C. Arsenault, Photonic Nose-Sensor Platform for Water and
Food Quality Control. Small. 7, 3153 -- 3157 (2011).
20. C. Adley, Past, Present and Future of Sensors in Food Production. Foods. 3, 491 -- 510
(2014).
21. V. Morandi, F. Marabelli, V. Amendola, M. Meneghetti, D. Comoretto, Colloidal photonic
crystals doped with gold nanoparticles: Spectroscopy and optical switching properties. Adv.
Funct. Mater. 17, 2779 -- 2786 (2007).
22. V. Robbiano et al., Hybrid Plasmonic-Photonic Nanostructures: Gold Nanocrescents Over
Opals. Adv. Opt. Mater. 1, 389 -- 396 (2013).
23.
J. M. Luther, P. K. Jain, T. Ewers, A. P. Alivisatos, Localized surface plasmon resonances
arising from free carriers in doped quantum dots. Nat. Mater. 10, 361 -- 366 (2011).
24. D. Dorfs et al., Reversible tunability of the near-infrared valence band plasmon resonance
in Cu2- xSe nanocrystals. J. Am. Chem. Soc. (2011), doi:10.1021/ja2016284.
25. F. Scotognella et al., Plasmon dynamics in colloidal Cu2-xSe nanocrystals. Nano Lett. 11,
4711 -- 4717 (2011).
26. G. M. Paternò et al., Solution processable and optically switchable 1D photonic structures.
Sci. Rep. 8, 3517 (2018).
27. G. M. Paternò, L. Moscardi, I. Kriegel, F. Scotognella, G. Lanzani, Electro-optic and
magneto-optic photonic devices based on multilayer photonic structures. J. Photonics
Energy. 8, 1 (2018).
28. E. Aluicio-Sarduy et al., Electric field induced structural colour tuning of a silver/titanium
dioxide nanoparticle one-dimensional photonic crystal. Beilstein J. Nanotechnol. 7, 1404 --
1410 (2016).
29.
J. N. Anker et al., Biosensing with plasmonic nanosensors. Nat. Mater. 7, 442 -- 453 (2008).
30. S. Unser, I. Bruzas, J. He, L. Sagle, Localized surface plasmon resonance biosensing:
Current challenges and approaches. Sensors (Switzerland). 15, 15684 -- 15716 (2015).
31. G. Chen et al., A mechanistic study of the antibacterial effect of silver ions on Escherichia
coli and Staphylococcus aureus. J. Biomed. Mater. Res. 52, 662 -- 668 (2000).
32. S. D. Ponja et al., Enhanced Bactericidal Activity of Silver Thin Films Deposited via
Aerosol-Assisted Chemical Vapor Deposition. ACS Appl. Mater. Interfaces. 7, 28616 --
28623 (2015).
33.
J. S. Kim et al., Antimicrobial effects of silver nanoparticles. Nanomedicine
Nanotechnology, Biol. Med. 3, 95 -- 101 (2007).
34. G. Franci et al., Silver Nanoparticles as Potential Antibacterial Agents. Molecules. 20,
8856 -- 8874 (2015).
35.
I. Sondi, B. Salopek-Sondi, Silver nanoparticles as antimicrobial agent: a case study on E.
coli as a model for Gram-negative bacteria. J. Colloid Interface Sci. 275, 177 -- 182 (2004).
36. A. Dror-Ehre, H. Mamane, T. Belenkova, G. Markovich, A. Adin, Silver nanoparticle -- E.
coli colloidal interaction in water and effect on E. coli survival. J. Colloid Interface Sci.
339, 521 -- 526 (2009).
37. T. C. Dakal, A. Kumar, R. S. Majumdar, V. Yadav, Mechanistic Basis of Antimicrobial
Actions of Silver Nanoparticles. Front. Microbiol. 7, 1 -- 17 (2016).
38. A. Abbaszadegan et al., The Effect of Charge at the Surface of Silver Nanoparticles on
Antimicrobial Activity against Gram-Positive and Gram-Negative Bacteria: A Preliminary
Study. J. Nanomater. 2015, 1 -- 8 (2015).
39. A. M. El Badawy et al., Surface Charge-Dependent Toxicity of Silver Nanoparticles.
Environ. Sci. Technol. 45, 283 -- 287 (2011).
40. D. Esteban-Ferrer, M. A. Edwards, L. Fumagalli, A. Juárez, G. Gomila, Electric
Polarization Properties of Single Bacteria Measured with Electrostatic Force Microscopy.
ACS Nano. 8, 9843 -- 9849 (2014).
41. P. Drude, Zur Elektronentheorie der Metalle. Ann. Phys. 312, 687 -- 692 (1902).
42.
J. J. Mock, D. R. Smith, S. Schultz, Local Refractive Index Dependence of Plasmon
Resonance Spectra from Individual Nanoparticles. Nano Lett. 3, 485 -- 491 (2003).
43. B. V. Lotsch, C. B. Knobbe, G. A. Ozin, A Step Towards Optically Encoded Silver Release
in 1D Photonic Crystals. Small. 5, 1498 -- 1503 (2009).
44. N. De Kruijf et al., Active and intelligent packaging: applications and regulatory aspects.
Food Addit. Contam. 19, 144 -- 162 (2002).
45. K. J. Lee, J. W. Wu, K. Kim, Enhanced nonlinear optical effects due to the excitation of
optical Tamm plasmon polaritons in one-dimensional photonic crystal structures. Opt.
Express. 21, 28817 (2013).
46. C. Symonds et al., High quality factor confined Tamm modes. Sci. Rep. 7, 3859 (2017).
47. M. Shaban, A. M. Ahmed, E. Abdel-Rahman, H. Hamdy, Tunability and Sensing Properties
of Plasmonic/1D Photonic Crystal. Sci. Rep. 7, 41983 (2017).
48. D. Barchiesi, T. Grosges, Fitting the optical constants of gold, silver, chromium, titanium
and aluminum in the visible bandwidth. J. Nanophotonics. 8, 089996 (2015).
49. A. D. Rakić, Algorithm for the determination of intrinsic optical constants of metal films:
application to aluminum. Appl. Opt. 34, 4755 (2009).
50. A. D. Rakic, A. B. Djurisic, J. M. Elazar, M. L. Majewski, Optical properties of metallic
films for vertical-cavity optoelectronic devices. Appl. Opt. 37, 5271 -- 83 (1998).
51.
I. Kriegel, F. Scotognella, L. Manna, Plasmonic doped semiconductor nanocrystals:
Properties, fabrication, applications and perspectives. Phys. Rep. 674, 1 -- 52 (2017).
52. F. Scotognella et al., Metal oxide one dimensional photonic crystals made by RF sputtering
and spin coating. Ceram. Int. 41, 8655 -- 8659 (2015).
53.
I. H. Malitson, Interspecimen Comparison of the Refractive Index of Fused Silica*,†. J.
Opt. Soc. Am. 55, 1205 (1965).
54. R. J. Mendelsberg, G. Garcia, H. Li, L. Manna, D. J. Milliron, Understanding the plasmon
resonance in ensembles of degenerately doped semiconductor nanocrystals. J. Phys. Chem.
C. 116, 12226 -- 12231 (2012).
55.
J. C. M. Garnett, Colours in Metal Glasses, in Metallic Films, and in Metallic Solutions. II.
Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 205, 237 -- 288 (1906).
56. M. Bellingeri, A. Chiasera, I. Kriegel, F. Scotognella, Optical properties of periodic, quasi-
periodic, and disordered one-dimensional photonic structures. Opt. Mater. (Amst). 72, 403 --
421 (2017).
57. M. Born, E. Wolf, Principles of optics: electromagnetic theory of propagation, interference
(1999).
Supplementary information for
A Hybrid 1D Plasmonic/Photonic Crystals are Responsive to
Escherichia Coli
Giuseppe Maria Paternò1†*, Liliana Moscardi1,2†, Stefano Donini1, Davide Ariodanti3, Ilka Kriegel4, Maurizio Zani2,
Emilio Parisini1, * Francesco Scotognella1,2 and Guglielmo Lanzani1,2*
1Center for Nano Science and Technology@PoliMi, Istituto Italiano di Tecnologia, Via Giovanni Pascoli, 70/3, 20133 Milano, Italy;
2Dipartimento di Fisica, Politecnico di Milano, Piazza Leonardo da Vinci 32, 20133 Milano, Italy.
3Dipartimento di Chimica, Materiali e Ingegneria Chimica "Giulio Natta", Piazza Leonardo da Vinci 32, 20133 Milano, Italy.
4Department of Nanochemistry, Istituto Italiano di Tecnologia (IIT), via Morego, 30, 16163 Genova, Italy
†These authors contributed equally to this work
* Corresponding authors
This file includes: Discussion on the biocidal activity of silver; scanning electron microscopy
images of the silver film and cross-section of the 1D photonic crystal; UV-Vis absorption spectra
of the silver films after dipping in liquid LB and E. coli/LB mixture; zone of inhibition in the agar
plate inoculated with E. coli; UV-Vis transmission spectra of the Ag/BSs samples after dipping in
liquid LB and E. coli/LB mixture.
Biocidal activity of silver
The antimicrobial properties of silver are well-known, and are exploited in a variety of everyday
life applications(1, 2). For this reason, silver thin film and nanoparticles (NPs) have been employed
in a range of applications as antibacterial agent, which can be also useful for addressing the
problem of antibiotic resistance that is occurring in the last decade (3). However, a unified and
definitive explanation to describe this property has not been provided yet (4).
Various studies have related the cytotoxic effect of Ag to a combination of different reactions that
take place within the prokaryotic cell(4). Furthermore, it has been observed that the anti-bacterial
efficacy depends on the type of bacteria(5). The bacteria can be cataloged in Gram-positive and
Gram-negative, based on the conformation of their membrane. In the former, the cell wall is
composed of a thick layer of peptidoglycans, while the Gram-negatives have an outer membrane
formed mainly by lipopolysaccharides and an inner one of peptidoglycans, much thinner than the
previous ones(6). Given the larger size of the Gram-positive cell wall, the endocytosis of the
nanoparticles is more difficult, and therefore less effective for this kind of bacteria(3).
The cytotoxic effect of silver nanoparticles is more or less marked depending on concentration,
shape, time and size(3, 7, 8). NPs with dimensions smaller than 10 nm are more easily absorbed
by endocytosis from the cell and interact with lysosomes and endosomes, in fact, thanks to their
large specific surface area they are particularly reactive(3, 7, 9). The acidic environment present
inside the lysosomes favors chemical reactions that increase the presence of ROS (reactive oxygen
species) and of superoxide anion (O2-) produced by them, leading to an arise of oxidative stress.
ROS cause an imbalance between the cell's ability to eliminate reactive intermediates and oxygen
production(9). Hydrogen peroxide (H2O2) contained in the ROS reacts with Ag NPs leading to the
formation of Ag ions: 2Ag + H2O2 + 2H+ → 2Ag++ 2H2O (7). Moreover, H2O2 can lead to the
formation of ·OH, considered one of the most oxidative ROS, able to oxidize the whole cell.
Nanoparticles and silver ions can escape from lysosomes, increasing intracellular ROS
concentration(10). Ag NPs and Ag ions reduce glutathione, thioredoxin, superoxide dismutase and
thioredoxin peroxidase, as they react with thiol groups, contained in most of the cell including
cytoplasm, mitochondria and cell membrane. Damage to the cell wall causes an increase in its
permeability and a cytoplasm leak, thus leading to necrosis(11). Also the lysosomal membrane
rupture causes a lysosome-mediated apoptosis, pouring the cathepsins into the cytoplasm(11). Ag
NPs and Ag+ cause damage to mitochondria by inhibiting the production of adenosine triphosphate
(ATP), increasing oxidative stress, interfering with mitochondria impairs electron transfer, causing
swelling and acceleration of mitochondrial respiration itself and leading to apoptosis(11, 12).
Nuclear pore complex has an average diameter of 9-10 nm, hence small AgNPs can penetrate and
deposit inside with a subsequent production of ROS that damages the DNA and generates
chromosomal abnormalities(11). Also Ag+ has been seen to cause problems to DNA and induce
apoptosis of the cells(7, 10).
Figure S1 (A) Scanning electron microscopy (SEM) image of Ag/BS sample (cross section). (B) SEM of the Ag layer exhibiting
the typical granular morphology.
Figure S2. Average optical absorption of a Ag thin film (8 nm on glass) before and after dipping in before and after dipping in (A)
LB, (B) E. coli at 0.1 OD600nm, (C) 0.5 OD600nm, (D)1.2 OD600nm.
Figure S3. Picture of the Agar plate embedded with LB medium or inoculated with E. coli before (left) and after removal of our
Ag/BSs.
Figure S4. Average transmission spectrum of the Ag/BSs before and after dipping in (A) LB, (B) E. coli at 0.1 OD600nm, (C) 0.5
OD600nm, (D)1.2 OD600nm.
1.
J. H. Crabtree et al., The efficacy of silver-ion implanted catheters in reducing peritoneal
dialysis-related infections. Perit. Dial. Int. 23, 368 -- 74.
2. M. Catauro, M. G. Raucci, F. de Gaetano, A. Marotta, Antibacterial and bioactive silver-
containing Na 2 O·CaO·2SiO 2 glass prepared by sol -- gel method. J. Mater. Sci. Mater.
Med. 15, 831 -- 837 (2004).
3. M. Rai, A. Yadav, A. Gade, Silver nanoparticles as a new generation of antimicrobials.
Biotechnol. Adv. 27, 76 -- 83 (2009).
4. G. Franci et al., Silver Nanoparticles as Potential Antibacterial Agents. Molecules. 20,
8856 -- 8874 (2015).
5.
S. Shrivastava et al., Characterization of enhanced antibacterial effects of novel silver
nanoparticles. Nanotechnology. 18, 225103 (2007).
6.
T. J. Silhavy, D. Kahne, S. Walker, The bacterial cell envelope. Cold Spring Harb. Perspect.
Biol. 2 (2010), p. a000414.
7.
L. Wei et al., Silver nanoparticles: Synthesis, properties, and therapeutic applications. Drug
Discov. Today. 20, 595 -- 601 (2015).
8.
S. Pal, Y. K. Tak, J. M. Song, Does the antibacterial activity of silver nanoparticles depend
on the shape of the nanoparticle? A study of the gram-negative bacterium Escherichia coli.
J. Biol. Chem. 290, 1712 -- 1720 (2015).
9. Y.-N. Chang, M. Zhang, L. Xia, J. Zhang, G. Xing, The Toxic Effects and Mechanisms of
CuO and ZnO Nanoparticles. Materials (Basel). 5, 2850 -- 2871 (2012).
10. D. Guo et al., Anti-leukemia activity of PVP-coated silver nanoparticles via generation of
reactive oxygen species and release of silver ions. Biomaterials. 34, 7884 -- 7894 (2013).
11. T. Zhang, L. Wang, Q. Chen, C. Chen, Cytotoxic potential of silver nanoparticles. Yonsei
Med. J. 55, 283 -- 91 (2014).
12. M. R. Almofti, T. Ichikawa1, K. Yamashita3, H. Terada5, Y. Shinohara1, Hayashi cho,
Takamatsu 761-0395; 3School of Dentistry, The University of Tokushima 4Institute for
Genome Research. J. Biochem. 134, 770 -- 8503 (2003).
|
1711.11083 | 1 | 1711 | 2017-11-28T09:13:41 | Tunable kinoform x-ray beam splitter | [
"physics.app-ph"
] | We demonstrate an x-ray beam splitter with high performances for multi-kilo-electron-volt photons. The device is based on diffraction on kinoform structures, which overcome the limitations of binary diffraction gratings. This beam splitter achieves a dynamical splitting ratio in the range 0-99.1% by tilting the optics and is tunable, here shown in a photon energy range of 7.2-19 keV. High diffraction efficiency of 62.6% together with an extinction ratio of 0.6% is demonstrated at 12.4 keV, with angular separation for the split beam of 0.5 mrad. This device can find applications in beam monitoring at synchrotrons, at x-ray free electron lasers for online diagnostics and beamline multiplexing and, possibly, as key elements for delay lines or ultrashort x-ray pulses manipulation. | physics.app-ph | physics | Tunable kinoform x-ray beam splitter
M. LEBUGLE,*
AND C. DAVID
G. SENIUTINAS, F. MARSCHALL, V. A. GUZENKO, D. GROLIMUND,
Paul Scherrer Institut, CH 5232 Villigen-PSI, Switzerland
*Corresponding author: [email protected]
Received 9 August 2017; accepted 6 September 2017; posted 21 September 2017 (Doc. ID 301376); published 19 October 2017
We demonstrate an x-ray beam splitter with high
performances for multi-kilo-electron-volt photons. The
device is based on diffraction on kinoform structures,
which overcome the limitations of binary diffraction
gratings. This beam splitter achieves a dynamical splitting
ratio in the range 0-99.1% by tilting the optics and is
tunable, here shown in a photon energy range of
7.2-19 keV. High diffraction efficiency of 62.6% together
with an extinction ratio of 0.6% is demonstrated at
12.4 keV, with angular separation for the split beam of
0.5 mrad. This device can find applications in beam
monitoring at synchrotrons, at x-ray free electron lasers
for online diagnostics and beamline multiplexing and,
possibly, as key elements for delay lines or ultrashort
x-ray pulses manipulation. © 2017 Optical Society of
America
phase gratings can also serve as beam splitters [7], for example, for
x-ray monitoring in large-scale facilities [8,9]. At multi-kilo-electron-
volt energies, efficient manipulation of light by diffractive optics is
particularly difficult due to the long required material length to
obtain a significant phase shift (e.g., of 𝜋 radians in phase-shifting
zones of a binary phase grating) because of a low refractive index
contrast. In this regime, the diffraction angle
in good
approximation given by the ratio of the wavelength over the pitch,
and diffraction angles that are as large as possible are usually
sought. Therefore, high-aspect ratio nanostructures are required,
continuously challenging the limits of nanotechnology [10-12].
is
OCIS codes: (340.0340) X-ray optics; (230.1360) Beam splitters;
(050.1950) Diffraction gratings; (220.4241) Nanostructure fabrication.
https://dx.doi.org/10.1364/OL.42.004327
Splitting a beam is a fundamental operation in optics. At visible light
wavelengths, the design of beam splitters was achieved already in
the early days of optics by exploiting basic mechanisms of linear
light-matter interaction. In bulk optics, a plethora of devices is
commonly found to realize this operation. A few examples are cube
beam splitters based on total internal reflection, devices based on
polarization splitting such as the Wollaston or the Glan-Taylor
prism, or those for which coatings (metallic or dielectric) are used
for partial transmission and reflection at single or multiple
interfaces. Beam splitters are key components in a broad range of
experiments, for instance, in interferometry, in laser building as
output couplers, or in pump-probe experiments, to name a few. In
the x-ray realm, however, splitting a beam is not straightforward,
particularly in view of obtaining a high splitting efficiency, a high
extinction ratio, a dynamical control over the splitting ratio, and an
energy tunable device. Various implementations were proposed, for
example using single crystals (e.g., silicon or diamond) placed for
Bragg reflection [1-3] possibly using an additional beam stop [4],
Laue diffraction [5] or a combination of both [6]. Diffractive binary
In this Letter, we present and implement the novel concept of a
kinoform x-ray beam splitter, which has a dynamical splitting ratio
and can be tuned in photon energy, here demonstrated in the range
of 7.2-19 keV. At 12.4 keV, we demonstrated a splitting ratio up to
99.1%, with a splitting efficiency of 62.6%, or -2.0 dB, and an
extinction ratio of 0.6%, or -22.2 dB. Other values of the splitting
ratio, down to zero, are achieved by using intermediate tilt angles of
the beam splitter. The fundamental idea underlying this Letter is to
use the 0th and the 1st diffraction orders (DOs) as output ports of
the beam splitter, occurring while an incoming x-ray beam
impinges on a kinoform grating.
First, to overcome the limitations of binary phase gratings that have
a maximum efficiency of 40.5%, our design is based on the
kinoform profile. The latter was introduced in context of x-ray
lenses and consists of introducing a progressive phase shifting up to
2𝜋 radians in every zone of a zone plate [13-15]. We designed our
beam splitter using a kinoform profile; however, here the pitch is
constant over the entire aperture, realizing a kinoform diffraction
grating. As an asymmetry in the transmission function is
introduced, our beam splitter can also be seen as a blazed grating,
enhancing the fraction of intensity placed in a given DO. For
optimizing the efficiency of the 𝑚th DO, the optimal phase shift to
imprint onto the beam in the thickest part of each grating line is
Δφm = 2𝜋𝑚, where 𝑚 is an integer. For the 1st DO (𝑚 = 1), the
realization for multi-kilo-electron-volt photons is, however, already
a challenge, since the diffractive structure height required to
provide a phase shift of 2𝜋 radians is considerable with common
materials for x-ray optics such as gold, nickel, or silicon. For
increasing the effective structure height, and as the beam splitter is
⁄
intrinsically one-dimensional, one can tilt the optics [16] (see Fig. 1).
When x-rays propagate through the beam splitter in transmission
with an angle of incidence 𝛼, an increase of path length that is
proportional to 1 tan(𝛼)
is obtained. Simultaneously, the tilted
geometry makes it possible to use only binary structures for
realizing a continuous phase profile and, thus, avoids resorting to
complex fabrication processes such as multi-level stacking [12,17].
To this end, we pattern stripes of triangle nanopillars, which
approximate the optimal kinoform phase profile with constant
pitch, as depicted in Fig. 1. The effective phase profile equals the
sum of the phase profile of each triangular nanopillar experienced
by the x-rays, which depends on the real part of the refractive index
of the material and the angle of incidence, or the tilt angle. As the
beam propagates through this tilted array of triangular nanopillars,
the sawtooth phase shift is realized within the accuracy of the
fabrication process. Importantly, in contrast to planar refractive
lenses [18,19], the aperture is not limited by the structure height
and, with current lithographic fabrication techniques, the time for
structure patterning does not constrain the area of the lens.
Diffraction occurs in the xz-plane defined by the incoming beam and
the tilt axis (light gray plane in Fig. 1). The tilt angle is equal to zero
when the beam splitter surface is parallel to the beam.
Fig. 1. Schematic view of the kinoform x-ray beam splitter.
Secondly, by using the tilt angle as a degree of freedom, one can
dynamically set the magnitude of the phase profile to vary the
relative intensity distribution between the DOs. This makes it
possible to adjust the splitting ratio between the two output ports of
the beam splitter, which is similar to adjusting the blazed angle of
blazed gratings for given DO and energy. The phase shift can be set
from small values to 2𝜋𝑚 radians in the thickest part of each
grating line, the upper value being limited by the optics aperture.
The set of tilt angles 𝛼𝑚 for maximum efficiency of the 𝑚th DO is
𝛼𝑚 = asin (
ℎ𝛿𝑔
𝑚𝜆
) ,(1)
with ℎ being the height of the structures, 𝛿 being the real part of the
refractive index, 𝑔 being a factor accounting for the gap separating
each lenslet element due to fabrication, and 𝜆 being the x-ray
wavelength. The performances can be assessed by measuring the
diffraction efficiency (DE) of the 𝑚th DO, 𝑑𝑚(𝛼), a function of the
tilt angle 𝛼:
𝑑𝑚(𝛼) =
𝐼𝑚(𝛼)
𝐼𝑖𝑛𝑐
,(2)
where 𝐼𝑚(𝛼) is the intensity of the 𝑚th DO at the tilt angle 𝛼, and
𝐼𝑖𝑛𝑐 is the incident intensity. Thus, in the following, the DE includes
absorption. As the two output ports of the beam splitter are the 0th
and the 1st DO, an ideal device should achieve 𝑑0(0) = 1 and
𝑑1(0) = 0, a trivial case achieved when the component is placed
out of the beam; then 𝑑0(𝛼1) = 0 and 𝑑1(𝛼1) = 1. This corresponds
to a lossless material being able to route all the incoming flux
towards the 1st DO, with a perfect extinction ratio. All intermediate
values of 𝛼 < 𝛼1 allow adjusting the splitting ratio between the
output ports. The splitting ratio is
𝑠(𝛼) =
𝑑1(𝛼)
𝑑0(𝛼) + 𝑑1(𝛼)
⋅ (3)
Finally, using a low-Z element for fabricating an x-ray beam splitter
is crucial to keep absorption as low as possible. We target a range of
x-ray energies centered on 12.4 keV (wavelength of 1 Å). Therefore,
silicon (Si, Z=14) is an interesting candidate, since its K-edge has a
characteristic energy of 1.84 keV and has a rather small imaginary
part of the refractive index of 3.16×10-8 at 12.4 keV. While an x-ray
beam with such energy travels through the required length for a 2𝜋
radians phase shift of ~31.3 μm (optimized DE of the 1st DO), a
moderate absorption of ~11.7% occurs. Importantly, our beam
splitter is also energy tunable, achieved by using a different set of tilt
angles 𝛼𝑚 [see eq. (1)] for reaching the optimal phase shift of a
given DO. The use of Si prevents the use of the device at photon
energies close the K-edge while, at high energies, the required
extremely shallow angles become limiting. A representative range
of energies for high performances is 7.2-19 keV, as further
demonstrated.
We fabricated kinoform beam splitters with pitches of 200, 300, and
400 nm (Fig. 2). To achieve Si patterning in triangular nanopillars
with sufficient height, we used a metal-assisted chemical etching
(MAC-etch) process for etching 10-μm thick Si membranes
[orientation (100), boron-doped with conductivity of 1-20 Ω·cm].
This technique transfers a pattern with high fidelity in Si and
produces high aspect ratio nanostructures [11,20,21], see Fig. 2(g).
E-beam lithography at 100 keV (Vistec EBPG 5000plus, Raith
GmbH) was performed on a bilayer of MMA/PMMA resists to
expose a triangular stripe negative layout. After development in a
solution of isopropanol and water (7:3 in volume), thermal
evaporation was used to deposit a 30 nm thick gold film at a rate of
0.25 nm/s. A lift-off procedure was performed in acetone to obtain a
negative mask pattern corresponding to the kinoform grating
profile. The MAC-etch technique was realized at room temperature
with hydrogen peroxide (H2O2) as an oxidizer and hydrofluoric acid
(HF) as an etching agent in a water-based solution. As a result, Si
could be etched with the gold pattern acting as a negative mask. The
molar concentrations were [HF]=4.93 M and [H2O2]=0.55 M,
resulting in the molar ratio 𝜌 = [HF] ([HF] + [H2O2])
of 0.9. The
reduction reaction of H2O2 is the rate-limiting step of the redox
reaction, which limits the variation in the etch rate as the diffusion
lengths for different feature sizes of the catalyst are minimized, as
interestingly put forward recently by Chang et al. [11]. The etching
rate was about 0.3 μm/min. The samples shown here have a depth
of about 5.7 μm, but varies slightly among structures with different
pitch due to the influence of the mask on the MAC-etch [21,11]. The
resulting aspect ratio is 28.5:1. The gold mask was removed in an
aqueous solution of potassium iodide, followed by critical-point
drying (Leica EM CPD300 Auto). The support membrane was
⁄
thinned by about 2 μm from the back side using deep reactive ion
etching (Oxford Plasmalab100).
Fig. 2. SEM micrographs of kinoform beam splitters on Si membranes
with pitches of (a), (d) 200 , (b), (e) 300 , and (c), (f) 400 nm. (a)-(c) Top
view and (d)-(f) tilted view with an angle of 19.5°. (g) SEM micrograph
of a cross section of structures with a 250 nm pitch etched into bulk Si,
further cleaved for inspection.
X-rays experiments were performed at the microXAS beamline,
Swiss Light Source (Paul Scherrer Institute, Switzerland). A
fixed-exit double-crystal Si(111) monochromator defined the x-ray
energy with a bandwidth of about 2×10-4. The beam splitters were
aligned using a hexapod (SmarAct) with piezo nanopositioning
stages for all six degrees of freedom of translation and rotation. We
first characterized at 12.4 keV a kinoform beam splitter with a pitch
of 200 nm and diffracting along the vertical dimension. A beam with
an aperture of 80 μm x 80 μm was defined by a pair of slits. The
detector was at 0.42 m downstream, where a spatial separation of
210 μm between the DOs is found. By vertically scanning a 20-μm
slit and measuring the transmitted flux using a photodiode, we
obtained the DE 𝑑𝑚(𝛼) of the -8th to the 8th DOs at tilt angles
between 4° and 40° (Fig. 3 and inset).
Fig. 3. DEs of the -8th to the 8th DOs at 12.4 keV of the beam splitter
with a 200 nm pitch versus the tilt angle. The inset shows the raw
signal in logarithmic scale obtained by scanning a 20 μm slit across
the DOs (vertical axis) versus the tilt angle (horizontal axis). The
color bar is the exponent in base 10 of the photodiode current (in
A).
As the tilt angle becomes shallower, the 0th DO is attenuated, while
the DE of the 1st DO increases, and reaches a maximum of 62.6% at
𝑒𝑥𝑝 =9°, theoretically expected at 7.4° using Eq. (1).
a tilt angle of 𝛼1
This difference probably arises from the fact that, in Eq. (1), a
lossless material is assumed, shifting the expected angles towards
shallower values. At the tilt angle optimum for the 1st DO, our tilt
geometry permits a six-fold increase of the effective aspect ratio of
the nanostructures. Other maxima of higher DOs, such as
the -1st, -2nd, and 2nd also occur and may be attributed to either
the imperfect shape of the kinoform elements or secondary maxima
of the DE as a function of the material length passed through, i.e., not
occurring at the optimum phase shift of Δφ𝑚 = 2𝜋𝑚. At an angle of
7°, a significant fraction of 24.6 % is found in the sum of DEs of the
±3rd, ±4th, ±5th, ±6th, ±7th, and ±8th DOs (see the light blue line
with star markers). This reveals the higher sensitivity of high DOs to
small imperfections of the kinoform shape. The sum of all measured
DEs (dark red line) agrees well with the calculated overall
transmission (dark blue line). A deviation is observed at 18° that
may come from the (400) Si Bragg reflection (expected at 21.6°),
which would indicate a miscut of the crystal of about 3.6°. In view of
our application, the DEs of the beam splitter with a 200 nm pitch are
𝑒𝑥𝑝) = 62.6%, for the 0th and the 1st
𝑑0(𝛼1
DOs, respectively. This leads to a maximum splitting ratio of
𝑒𝑥𝑝) = 99.1%. At an angle of 12.9°, the device acts as a 50:50
𝑠(𝛼1
beam splitter, with DEs of the 0th and the 1st DOs nearly identical
and equal to 35.1%. Table 1 and Fig. 4 summarize similar
measurements of the DEs with larger pitches of 300 and 400 nm,
also expressed in decibels.
𝑒𝑥𝑝) = 0.6% and 𝑑1(𝛼1
Table 1. Optimum of DEs of the 0th and 1st DOs at 12.4 keV of
Beam Splitters with Pitches of 200 nm, 300 nm, and 400 nm
Pitch
𝑒𝑥𝑝)
𝑑0(𝛼1
𝑑1(𝛼1
𝑒𝑥𝑝)
Splitting
angle
200 nm
0.6%
-22.2 dB
62.6%
-2.0 dB
300 nm
0.6%
-22.2 dB
71.4%
-1.5 dB
400 nm
0.8%
-21.0 dB
74.8%
-1.3 dB
0.5 mrad
0.33 mrad
0.25 mrad
The maximum DE increases with pitch, being as high as 74.8%
(-1.3 dB) for a pitch of 400 nm, as the shape of larger nanostructures
is better controlled [see Figs. 2(c) and 2(f)]. The beam splitter
extinction ratio remains strong, less than 0.8% (-21.0 dB), and
50:50 splitting is possible for all devices. Using the beam splitter
with a pitch of 300 nm, images of both output ports on a scintillator
screen were obtained along with the splitting ratio s(α) and
1 − s(α); see Figs. 4(d) and 4(e). We also realized efficiency
measurements
from 7.2 keV to 19 keV (Table 2). Similar
performances were obtained while keeping an aperture of
80 μm x 80 μm. A slight increase in DE of the 1st DO with energy is
observed, as
the silicon absorption coefficient decreases
accordingly.
Table 2. Optimum of DEs of the 0th and 1st DOs at Several
Energies of a Beam Splitter with a Pitch of 200 nm
Energy
𝑒𝑥𝑝)
𝑑0(𝛼1
𝑑1(𝛼1
𝑒𝑥𝑝)
7.2 keV
1.0%
-20.2 dB
51.3%
-2.9 dB
16 keV
1.2 %
-19.2 dB
62.1%
-2.1 dB
19 keV
1.9%
-17.2 dB
64.1%
-1.9 dB
REFERENCES
[1] U. Bonse, and M. Hart, Appl. Phys. Lett., 6, 155 (1965).
[2] T. Osaka, M. Yabashi, Y. Sano, K. Tono, Y. Inubushi, T. Sato, S.
Matsuyama, T. Ishikawa, and K. Yamauchi, J. Phys. 425, 052014
(2013).
[3] S. Stoupin, S. A. Terentyev, V. D. Blank, Y. V. Shvyd'ko, K. Goetze, L.
Assoufid, S. N. Polyakov, M. S. Kuznetsov, N. V. Kornilov, J. Katsoudas,
R. Alonso-Mori, M. Chollet, Y. Feng, J. M. Glownia, H. Lemke, A.
Robert, M. Sikorski, S. Song, and D. Zhu, J. Appl. Cryst. 47, 1329
(2014).
[4] M. Hoshino, K. Uesugi, J. Pearson, T. Sonobe, M. Shirai, and N.
Yagi, J. Synchrotron Radiat. 18, 569 (2011).
[5] W. K. Lee, K. Fezzaa, and T. Uemura, J. Synchrotron Radiat. 18,
302 (2011).
[6] P. Oberta, and R. Mokso, Nucl. Instrum. Methods Phys. Res. A
703, 59 (2013).
[7] T. Weitkamp, A. Diaz, C. David, F. Pfeiffer, M. Stampanoni, P.
Cloetens, and E. Ziegler, Opt. Express. 13, 6296 (2005).
[8] P. Karvinen, S. Rutishauser, A. Mozzanica, D. Greiffenberg, P. N.
Juranić, A. Menzel, A. Lutman, J. Krzywinski, D. M. Fritz, H. T. Lemke,
M. Cammarata, and C. David, Opt. Lett. 37, 5073 (2012).
[9] M. Makita, P. Karvinen, D. Zhu, P. N. Juranic, J. Grünert, S. Cartier,
J. H. Jungmann-Smith, H. T. Lemke, A. Mozzanica, S. Nelson, L.
Patthey, M. Sikorski, S. Song, Y. Feng, and C. David, Optica 2, 912
(2015).
[10] S. Gorelick, J. Vila-Comamala, V. A. Guzenko, R. Barrett, M.
Salomé, and C. David, J. Synchrotron Radiat. 18, 442 (2011).
[11] C. Chang, and A. Sakdinawat. Nat. Commun. 5, 4243 (2014).
[12] I. Mohacsi, I. Vartiainen, M. Guizar-Sicairos, P. Karvinen, V. A.
Guzenko, E. Müller, C. M. Kewish, A. Somogyi, and C. David, Opt. Lett.
41, 281 (2016).
[13] J. A. Jordan, Jr., P. M. Hirsch, L. B. Lesem, and D. L. Van Rooy
Appl. Opt. 9, 1883 (1970).
[14] A. F. Isakovic, A. Stein, J. B. Warren, S. Narayanan, M. Sprung, A.
R. Sandy, and K. Evans-Lutterodt, J. Synchrotron Radiat. 16,
8(2009).
[15] P. Karvinen, D. Grolimund, M. Willimann, B. Meyer, M. Birri, C.
Borca, J. Patommel, G. Wellenreuther, G. Falkenberg, M. Guizar-
Sicairos, A. Menzel, C. David, Opt. Express. 22, 16676 (2014).
[16] C. David, B. Nöhammer, and E. Ziegler, Appl. Phys. Lett. 79,
1088 (2001).
[17] E. Di Fabrizio, F. Romanato, M. Gentili, S. Cabrini, B. Kaulich, J.
Susini, and R. Barrett, Nature 401, 895 (1999).
[18] C. G. Schroer, O. Kurapova, J. Patommel, P. Boye, J. Feldkamp, B.
Lengeler, M. Burghammer, C. Riekel, L. Vincze, A. van der Hart, and
M. Küchler, Appl. Phys. Lett. 87, 124103 (2005).
[19] G. B. M. Vaughan, J. P. Wright, A. Bytchkov, M. Rossat, H.
Gleyzolle, I. Snigireva, and A. Snigirev, J. Synchrotron Radiat. 18, 125
(2011).
[20] X. Li, and P. W. Bohn, Appl. Phys. Lett. 77, 2572 (2000).
[21] Z. Huang, N. Geyer, P. Werner, J. De Boor, and U. Gösele, Adv.
Mater. 23, 285 (2011).
[22] A. F. Isakovic, A. Stein, J. B. Warren, A. R. Sandy, S. Narayanan, M.
Sprung, J. M. Ablett, D. P. Siddons, M. Metzler and K. Evans-Lutterodt,
J. Synchrotron Radiat. 17, 451 (2010).
Fig. 4. DEs of the 0th and 1st DOs at 12.4 keV of beam splitters with
pitches of (a) 200, (b) 300, and (c) 400 nm, versus the tilt angle. (d)
DOs imaged on a scintillator screen using a 300 nm pitch beam
splitter, with (e) the corresponding splitting ratios (see text).
Our beam splitter could be used for x-ray beam diagnostics
requiring variable flux, either at synchrotrons or x-ray free-electron
lasers (X-FELs). At X-FELs, multiplexing experiments could be
enabled by routing part of the photon flux to a secondary setup. To
increase the angular separation between beams, Bragg reflection
onto a thin Si crystal can be used. As the splitting angle (see Table 1)
is larger than the corresponding rocking curve width, one could
tune only one split beam for Bragg reflection to obtain a larger split
angle. One application could be a split-and-delay line for ultrashort
x-ray pulses, where two pulses are generated and recombined with
varied delay and relative intensities, to perform pump-probe or
double-pump experiments in a non-collinear or collinear geometry.
Eventually, the nanostructures of the kinoform beam splitter can
form a Fresnel bi-prism, i.e., two thin prisms joined at their base. In
such a device, the left portion of the wavefront is deflected right, and
vice-versa for the right portion, creating a zone of interference as
two virtual sources exist. This can be used for source size or
coherence length measurements [22].
Summarizing, we devised a tunable beam splitter for multi-kilo-
electron-volt x-ray wavelengths with high efficiency and extinction
ratio, and whose splitting ratio can be dynamically adjusted. We
foresee the use of such beam splitters for beam diagnostics in large-
scale facilities, and possibly as key element in ultrafast x-ray optics.
Funding. Horizon 2020 Framework Programme (H2020) (654360
NFFA-Europe).
Acknowledgment. The x-ray experiments were performed at the
microXAS beamline of the Swiss Light Source, Paul Scherrer
Institute, Villigen, Switzerland. The authors
thank Florian
Dworkowski for fruitful discussions about kinoform optics. They
are also grateful to Lucia Romano and Dario Marty for their help
while developing the MAC-Etch process, and to Dario Ferreira
Sanchez for his dedicated support during the measurements at
microXAS.
|
1902.07470 | 1 | 1902 | 2019-02-20T09:43:12 | Improved Photoelectrochemical Water Splitting of CaNbO2N Photoanodes by Co-Pi Photodeposition and Surface Passivation | [
"physics.app-ph"
] | Photoelectrochemical solar water splitting is a promising approach to convert solar energy into sustainable hydrogen fuel using semiconductor electrodes. Due to their visible light absorption properties, oxynitrides have shown to be attractive photocatalysts for this application. In this study, the influence of the preparation method of CaNbO2N particles on their morphological and optical properties, and thereby their photoelectrochemical performance, is investigated. The best performing CaNbO2N photoanode is produced by ammonolysis of Nb enriched calcium niobium oxide. The enhanced photoactivity arises from an enlarged surface area and superior visible light absorption properties. The photoactivity of this photoanode was further enhanced by photodeposition of Co-Pi co-catalyst and by atomic layer deposition of an Al2O3 overlayer. A photocurrent density of 70 microA.cm-2 at 1.23 V vs RHE was achieved. The observed enhancement of the photoelectrochemical performance after Co-Pi/Al2O3 deposition is the combined effect of the improved kinetics of oxygen evolution due to the Co-Pi co-catalyst and the reduced surface recombination of the photogenerated carriers at the Al2O3 surface layer. | physics.app-ph | physics | Improved Photoelectrochemical Water Splitting of CaNbO2N
Photoanodes by Co-Pi Photodeposition and Surface Passivation
Fatima Haydous,† Wenping Si,†§ Vitaliy A. Guzenko,‡ Friedrich Waag,┴ Ekaterina
Pomjakushina,† Mario El Kazzi,∥ Laurent Sévery,● Alexander Wokaun,∥ Daniele Pergolesi*,†,∥ and
Thomas Lippert*,†,#,∇
† Division for Research with Neutrons and Muons, Paul Scherrer Institut, 5232 Villigen-PSI, Switzerland
‡ Photon Science Division, Paul Scherrer Institut, 5232 Villigen-PSI, Switzerland
┴ Center for Nanointegration Duisburg-Essen, Technical Chemistry I, University of Duisburg-Essen
║ Energy and Environment Research Division, Paul Scherrer Institut, 5232 Villigen PSI, Switzerland
● Department of Chemistry, University of Zurich, 8057 Zurich, Switzerland
# Laboratory of Inorganic Chemistry, Department of Chemistry and Applied Biosciences, ETH Zurich, 8093 Zurich,
Switzerland
∇ Molecular Photoconversion Devices Division, International Institute for Carbon-Neutral Energy Research
(I2CNER) Kyushu University 744 Motooka, 819-0395 Fukuoka, Japan
ABSTRACT: Photoelectrochemical solar water splitting is a promising approach to convert solar energy into
sustainable hydrogen fuel using semiconductor electrodes. Due to their visible light absorption properties,
oxynitrides have shown to be attractive photocatalysts for this application. In this study, the influence of
the preparation method of CaNbO2N particles on their morphological and optical properties, and thereby
their photoelectrochemical performance, is investigated. The best performing CaNbO2N photoanode is
produced by ammonolysis of Nb enriched calcium niobium oxide. The enhanced photoactivity arises from
an enlarged surface area and superior visible light absorption properties. The photoactivity of this
photoanode was further enhanced by photodeposition of Co-Pi co-catalyst and by atomic layer deposition
of an Al2O3 overlayer. A photocurrent density of 70 µA.cm-2 at 1.23 V vs RHE was achieved. The observed
enhancement of the photoelectrochemical performance after Co-Pi/Al2O3 deposition is the combined
effect of the improved kinetics of oxygen evolution due to the Co-Pi co-catalyst and the reduced surface
recombination of the photogenerated carriers at the Al2O3 surface layer.
1.
INTRODUCTION
One of the biggest challenges our world is facing today is the increase in the energy demand associated with the shortage
of the fossil fuel supplies. This, in addition to the global warming caused by the carbon dioxide produced from fossil fuels,
points at the necessity of finding alternative renewable energy resources. As the sun is capable of supplying Earth with an
amount of energy per hour which surpasses the annual global energy need, it stands out to be the most appealing choice
among all potential renewable energy sources.1 The intermittent nature of the solar energy has driven many research efforts
towards the conversion of the solar energy into storable and transportable forms.2 The photoelectrochemical (PEC) water
splitting into hydrogen and oxygen using semiconductor electrodes is one of the most promising strategies to efficiently
harvest and store the solar energy. Within such an approach, photons in the visible light energy range are absorbed by a
semiconducting material and used for the creation of electron-hole pairs. The photogenerated electrons and holes migrate
to the surface of the semiconductor where they are used to drive the water reduction and oxidation reaction respectively.
Unfortunately, to date there is not a single photoactive material available that can fulfill all the requirements for a
commercially feasible solar hydrogen production process.
The main bottleneck is that most of the materials that are chemically stable in operating conditions are wide band gap
metal oxides, with light absorption limited to the ultraviolet spectral region.3 Many efforts have been devoted to extend the
light absorption to the visible light energy range by tuning the metal oxide band gap by doping or sensitizing with dye
molecules.2 On the other hand, for oxides with narrow band gaps, as Fe2O3 and WO3, the photoactivity is limited by their
chemical instability and inappropriate band alignment with water redox potentials.3
The narrowed band gap of oxynitrides relative to oxides mainly arises from the negative shift of the valence band
maximum due to the hybridization of N2p and O2p orbitals. 4-5 Also the energy position of the conduction band minimum
can be affected by the N substitution giving a substantial contribution to the narrowing of the band gap.6 This characteristic
makes oxynitrides very promising photocatalysts for water splitting.
The photoactivity of several oxynitrides was investigated. Currently, LaTiO2N,7-9 BaTaO2N10-12 and TaON2, 13-15 are
considered to be the best performing materials. Also Nb-based perovskite oxynitrides are expected to achieve good
performance mainly due to their band edge structure,16-17 which is particularly favorable for water splitting. Theoretical
calculations18 suggest for example CaNbO2N (CNON) as a very promising photocatalyst. It has been demonstrated that this
material is in fact capable of both oxidation and reduction of water in the presence of sacrificial agents.16
In spite of the good expectations, only a few studies have been conducted so far on this family of oxynitrides which still
show inefficient performances.16-17, 19-20 The main drawback associated with Nb-containing oxynitrides is the ease of
reduction of Nb, as compared to Ta for instance, during ammonolysis which results in more defects affecting the optical
and conducting properties.19 In general, the performance of a photocatalyst is limited by different factors including: slow
kinetics of hole transfer at the semiconductor's surface to the electrolyte, charge-carrier transport losses, low electron−hole
separation rates, surface recombination, and corrosion.21 The problem of recombination losses has been addressed by
facilitating the charge separation and transport and the oxygen evolution reaction (OER) kinetics by loading the
photocatalyst with well-known water oxidation catalysts, such as IrO2,14, 22-23 CoOx,7, 9, 23-25 and NiOx
23, 26. IrO2 is one of the
most active oxygen evolution co-catalysts, but for large-scale applications co-catalysts consisting of earth-abundant and
low-cost elements such as cobalt phosphate (Co-Pi) becomes more appealing.27 For instance, it was shown that combining
La(Ta,Nb)O2N with Co-Pi co-catalyst improves the charge separation and collection of holes produced at the oxynitride
surface, which consequently enhanced its photocatalytic activity. 28
The recombination losses can also be lowered by reducing the density of defects which improves the crystal quality.
Concerning Nb-based compounds, it was recently reported that an excess of Nb in the preparation of Ba and Sr niobium
oxides was beneficial for the crystallinity and thereby the photoactivity of their corresponding oxynitrides. This was
explained considering that the Nb-enriched precursor oxides are isostructural to the corresponding oxynitrides, thus no
structural changes occur during the ammonolysis process.17
This study provides new insights into the properties and potentials of CNON photoanodes toward solar water splitting.
The preparation procedure of the oxynitride is shown to play a key role in determining its PEC performance. We also show
here that the photo-assisted deposition of a Co-Pi co-catalyst and the atomic layer deposition (ALD) of a thin passivation
layer of Al2O3 significantly enhance the photoactivity.
2. EXPERIMENTAL SECTION
Synthesis of CaNbO2N Powders. CaNbO2N was prepared by both solid state (SS) and polymerized complex (PC)
methods.
For the SS reaction, CaCO3 (Alfa Aesar, 99.0% min.) and Nb2O5 (Merck, 99+%) in stoichiometric amounts were annealed
at 1000oC for 12hrs in the presence of a NaCl flux to prepare the Ca2Nb2O7 oxide precursor. Then, by thermal ammonolysis
of the oxide precursor at 800 oC for 24hrs with mixing in between, CaNbO2N-SS was prepared.
For the PC method, the Ca2Nb2O7 oxide precursor was prepared by dissolving stoichiometric amounts of CaCO3 and
NbCl5 (ChemPUR, 99+%) in methanol. Then, citric acid (CA) and ethylene glycol (EG) were added with a molar ratio of
1:1:6:2 for Ca:Nb:CA:EG. The solution was left to polymerize at 200oC overnight. Then, the obtained yellow resin was heated
at 400oC for 2hrs followed by annealing at 650oC and 800oC for 2hrs each with grinding in between. CaNbO2N-PC was
prepared by ammonolysis of the prepared oxide at 800oC for 24hrs.
The PC method was also applied for the preparation of Nb-enriched samples CaNbO2N(Nb)-PC by adding excess Nb up
to 1:2 ratio of CaCO3:NbCl5 in the initial oxide mixture.
An ammonia flow of 250mL/min was used for the preparation of all the oxynitrides. The ammonolysis temperature and
time were optimized in order to have a single phase of the oxynitride; since at lower temperatures the oxide coexist with
the oxynitride and at higher temperatures a NbOxNy phase is formed.
Preparation of Photoanodes. The photoanodes were prepared via electrophoretic deposition (EPD) where 40 mg of
CaNbO2N powder were mixed with 10 mg of iodine in 50 ml of acetone followed by 1 hour sonication to obtain dispersed
CaNbO2N powders. Electrophoretic deposition (EPD) was conducted between two parallel fluorine-doped tin oxide (FTO)
substrates (1×2 cm) placed in the CaNbO2N dispersion with a distance of 7 mm under a bias of 20 V for 3 min. CaNbO2N
powders were deposited on the negative electrode. A post-necking treatment was done after the deposition of CaNbO2N
on the FTO substrates by dropping 30 µL of 10 mM TaCl5 methanol solution on the CaNbO2N photoanodes followed by
drying in air. After repeating this cycle three times, the photoanodes were annealed at 300 oC for 30 min in air, followed by
a heating cycle under ammonia flow for 1 hour at 450 oC.
The Co-Pi co-catalyst was loaded on the CaNbO2N photoanodes by photodeposition. A solution of 0.5 mM CoCl2 was
prepared and 0.1 M potassium phosphate was added to adjust the pH to 7. Afterwards, the sample was dipped in this solution
and the Co-Pi co-catalyst was deposited on the CaNbO2N particles under UV-light illumination for 30 min.
Al2O3 and TiO2 overlayers were deposited on the Co-Pi loaded CaNbO2N photoanodes by atomic layer deposition (ALD).
For the Al2O3 layer, successive pulses of trimethylaluminum (TMA) and water vapor were performed in a closed chamber
at a temperature of 300 oC with nitrogen as a carrier gas. 20 cycles (TMA/N2 purging/water/N2 purging) were applied
resulting in a thickness of around 2 nm for the Al2O3 layer. For TiO2 layers (about 2 nm thick as well, as measured by
ellipsometry on a silicon wafer piece), the deposition was carried out using a Picosun R-200 tool at 120 oC with 32 sequential
pulses of tetrakis(dimethylamino)titanium (TDMAT, 99.999% trace metal basis, Sigma-Aldrich) preheated to 85 °C and
water vapor.
Photoelectrochemical (PEC) Measurements. PEC measurements were performed using a three electrode
configuration in 0.5 M NaOH (pH=13.0) aqueous solution with the CaNbO2N photoelectrode being used as the working
electrode. A coiled Pt wire and Ag/AgCl were used as the counter and the reference electrodes, respectively. The electrolyte
was purged with Ar for 1 hour prior to PEC measurements. The photoanodes were irradiated from the front side in all the
experiments with a 150 W Xe arc lamp (Newport 66477) equipped with AM 1.5 G filter and with an output intensity of 100
mW.cm-2 calibrated by a photodetector (Gentec-EO).
Characterization. X-ray diffraction (XRD) measurements, conducted by a Bruker -- Siemens D500 X-ray Diffractometer,
were used to characterize the calcium niobium oxide powders prepared by the solid-state and polymerized-complex routes
and their corresponding oxynitrides. The diffuse reflectance of the powders was measured with a Cary 500 Scan UV-Vis-
NIR spectrophotometer using an integrating sphere to determine their band gaps. Scanning electron microscopy (SEM)
images were obtained with a Zeiss Supra VP55 Scanning Electron Microscope. The surface area was calculated according to
the Brunauer -- Emmett -- Teller (BET) theory by conducting the N2 adsorption -- desorption analysis on a Quantachrome Nova
2200 at 77 K. A VG ESCALAB 220iXL spectrometer (Thermo Fischer Scientific) equipped with an Al Kα monochromatic
source and a magnetic lens system was used for the X-ray photoelectron spectroscopy (XPS) measurements.
Thermogravimetric (TG) measurements were acquired using NETZSCH STA 449C analyzer equipped with PFEIFFER
VACUUM ThermoStar mass spectrometer, where aliquots of 20-60 mg of the oxynitride powders were heated in alumina
crucibles to 1400 oC with a heating rate of 10 oC /min in 36.8 mL.min-1 synthetic air.
3. RESULTS AND DISCUSSION
Characterization of CaNbO2N and its oxide precursors. For an efficient PEC water splitting, the optimization of the
morphology of the powders is of great importance as it affects the properties of the photocatalyst in terms of active surface
area, light absorption, recombination, and charge transfer of the photogenerated carriers.29 In general, the preparation
method affects the morphologies of the resulting oxides and their corresponding oxynitrides. For this purpose, two
synthesis routes were used for the preparation of the CNON oxynitrides: the polymerized complex (PC) and solid state (SS)
methods.
Figure 1a shows the XRD patterns for the different oxides. The oxide prepared by the SS and PC route starting from
stoichiometric ratios of Ca and Nb precursors showed the monoclinic Ca2Nb2O7 single phase structure. The PC method was
also applied for the preparation of Nb-enriched samples used to probe the effect of the Nb content on the PEC performance.
When excess Nb was added in the initial mixture of CaCO3 and NbCl5, the CaNb2O6 orthorhombic columbite structure was
obtained.
CNON was then prepared by thermal ammonolysis at 800oC for 24 hours starting from the Ca2Nb2O7 and CaNb2O6 oxide
precursors. Figure 1b shows the XRD patterns of the CNON photoanodes produced from Ca2Nb2O7 -- SS, Ca2Nb2O7 -- PC,
CaNb2O6 -- PC denoted as CNON-SS, CNON-PC and CNON(Nb)-PC respectively. For all oxynitrides, the orthorhombic
perovskite structure was observed, in good agreement with the ICSD reference data, with no evidence of secondary phases.
The structures and lattice parameters of the oxides and oxynitrides are reported in Table S1.
We note that the addition of niobium did not result in an oxide isostructural to the oxynitride, as one might have expected
by comparison with similar materials. However, a clear difference in the morphologies of the Nb-enriched oxide and
oxynitride relative to the other samples is visible in the SEM images.
Figure 1. XRD patterns obtained from (a) the calcium niobium oxides and (b) their respective oxynitride photoanodes. The peaks
corresponding to the FTO substrate and the perovskite CNON oxynitride are marked with (●) and (∗), respectively.
Figure 2. SEM images for (a) Ca2Nb2O7-SS (b) CNON-SS (c) Ca2Nb2O7-PC (d) CNON-PC (e) CaNb2O6-PC (f) CNON(Nb)-PC
powders.
As shown in Figure 2a, Ca2Nb2O7-SS is obtained in the form of platelets of few µm in size. After ammonolysis, CNON-SS
revealed the same brick-like morphology as its original oxide but with a porous structure (Figure 2b). The porosity of the
oxynitride is attributed to the exchange of three O2- anions with two N3- anions during the ammonolysis process. A similar
morphology has been reported in many studies for LaTiO2N prepared by the solid state route.30-31 Figure 2c shows the
irregularly shaped Ca2Nb2O7-PC particles with a broad size distribution ranging between 0.1 µm and 4 µm. The nitration of
this oxide resulted in large irregular aggregated CNON-PC particles (Figure 2d). A different morphology was observed for
CaNb2O6-PC, the Nb-enriched oxide obtained by PC method. As shown in Figure 2e, highly aggregated foamy particles with
sizes ranging from 0.5 µm to 1.5 µm were obtained for this oxide. After ammonolysis, agglomerates of dense particles are
observed for CNON(Nb)-PC with very high porosity (Figure 2f). These agglomerates consist of particles with a small grain
size ranging between 50 and 120 nm as shown in Figure S1a.
Figure 3. The absorption spectra of (a) the oxides prepared from the solid state and polymerized complex routes (with and without
excess of Nb) and (b) their corresponding oxynitrides. The photograph in (c) shows the colors of the three photoanodes.
The absorbance spectra of the three oxynitrides and their corresponding oxides are presented in Figure 3a and b. It can
be seen clearly in Figure 3a that CaNb2O6 has an absorption edge at around 400 nm, more red-shifted compared to Ca2Nb2O7
synthesized by the two other routes. In addition, Ca2Nb2O7 -- PC shows a higher absorption background above the light
absorption edge compared to the other two oxides which might be attributed to an increased number of defects in this
oxide. Crystallographic defects create energy states within the energy gap of the materials thus resulting in absorption at
energies smaller than the band gap. Other defects that might be present in Ca2Nb2O7 -- PC are reduced Nb and/or oxygen
vacancies which have been shown to increase the background absorption.17
After ammonolysis, a red shift in the absorbance is observed for CNON-SS and CNON(Nb)-PC compared to their
precursor oxides. The band gaps of these CNON samples were determined from the direct transition in the Tauc plots
(Figure S2) to be 1.9 and 2.1 eV for CNON-SS and CNON(Nb)-PC, respectively. For CNON-PC, the Tauc plot in Figure S2b
shows a pronounced shoulder at around 2.5 eV which can be attributed to defects inherited from the oxide precursor and/or
to the reduced Nb species that contribute to the total absorption. Therefore, the band gap of CNON-PC could not be
precisely determined. This effect is also responsible of the high background absorbance shown in Figure 3b which makes
impossible to determine the absorption onset of CNON-PC. In general for niobates, the easy reduction of Nb5+ during
ammonolysis compared to other metal ions (as Ta5+ )32 leads to the formation of reduced Nb (Nb3+ and Nb4+) species which
causes an increase in the absorbance background.17 This also explains the observed high absorbance at wavelengths longer
than the absorption edge of CNON-SS. Concerning the Nb-enriched oxynitride, the observed lower background absorbance
might be due to the different reactivity of the starting oxide (CaNb2O6) with NH3 during ammonolysis compared to
Ca2Nb2O7. Indeed, this is confirmed by the dark brown color of CNON-SS and CNON-PC deposited on FTO compared to
the yellow-orange color of CNON(Nb)-PC as can be seen in the photograph of Figure 3c. However, from the absorbance
measurement, a clear conclusive result about the nitrogen content of the samples couldn't be obtained. Therefore,
thermogravimetric (TG) analysis was used to compare the N content of the three different oxynitrides through the change
of the mass with respect to temperature.
From the TG measurements shown in Figure 4, it can be observed that the three oxynitrides were stable up to a
temperature of at least 200 oC. At higher temperatures the mass increased sharply due to the uptake of oxygen before
decreasing again with the release of nitrogen. This qualitative behavior is in agreement with previous measurements
reported in literature.33 The onset temperature of the oxygen uptake was lower for the Nb enriched oxynitride with respect
to the other two samples. The weight gain ∆m after the complete oxidation of the oxynitride can be used to compare the N
content of the three oxynitrides. From Figure 4, it can be clearly seen that the mass gain is more than three times higher
for CNON(Nb)-PC compared to CNON-SS and CNON-PC. The mass difference is inversely proportional to the molar
masses of the resulting oxides. However, since the difference in the molar masses of CaNb2O6 and Ca2Nb2O7 obtained after
TG is not significant, the higher mass difference of CNON(Nb)-PC is attributed to a higher N content than in CNON-SS
and CNON-PC. Similar results were observed for La(Ti1-xNbx)O2N where increasing the Nb content lowered the temperature
at which oxygen uptake started and increased the N content in the oxynitrides.34
Figure 4. Thermogravimetric (TG) curves of the three calcium niobium oxynitride powders.
Effect of Morphology on the PEC Performance of CNON. CNON photoanodes were prepared by electrophoretic
deposition of the oxynitride powders. The deposition was followed by a post necking treatment similar to that reported in
literature for other oxynitrides in order to interconnect the particles and enhance the charge transport.30 In brief, TaCl5 was
dropped on the as-prepared photoanode, then by annealing in air Ta2O5 connections were obtained. Post annealing in
ammonia lead to the formation of Ta(O,N) that interconnected the CNON particles.
Figure S3 presents the current-voltage curves of CNON(Nb)-PC photoanodes in 0.5 M NaOH solution under chopped
illumination with Xe lamp. The as-prepared photoanode showed a negligible photocurrent due to bad electrical contact
between the particles. After necking with Ta2O5, the photocurrent increased to about 2 µA.cm-2 at 1.23 V vs RHE. A
significant enhancement in the photocurrent was achieved by annealing in NH3 reaching 11.3 µA.cm-2 at 1.23 V vs RHE. The
post-necking treatment was also applied to CNON-SS and CNON-PC photoanodes.
Figure 5 shows the photoelectrochemical behavior of the oxynitrides with different morphologies. CNON(Nb)-PC showed
the highest photocurrents in comparison to CNON-PC and CNON-SS. While photocurrents of about 10 and 14 µA.cm-2 were
achieved for CNON(Nb)-PC at 1.23 and 1.5 V vs RHE, respectively; CaNbO2N-SS resulted in much lower photocurrents (4
and 6 µA.cm-2 at 1.23 and 1.5 V vs RHE, respectively). CNON-PC resulted in the lowest photocurrents among the three
different CNON photoanodes (≈1 and 3 µA.cm-2 at 1.23 and 1.5 V vs RHE, correspondingly).
Figure 5. The potentiodynamic measurements of CNON-PC, CNON-SS, CNON(Nb)-PC acquired in NaOH (0.5M) solution under
chopped light illumination with a Xe lamp.
The difference in morphology between the oxynitride powders, which was observed in the SEM images (Figure 2), is
considered for a better understanding of the obtained photoactivities. The crystallinity of the photocatalysts is one of the
major factors affecting the activity. Indeed, higher crystallinity leads to a better separation of the photogenerated charge
carriers, thus leading to higher performance. However, CNON(Nb)-PC, which showed the smallest average grain size (lower
crystallinity) among the three samples (see Figure 2f and Figure S1) exhibited higher photocurrents compared to the non-
Nb enriched photoanodes with larger grain size. Hence, the photoactivity of CNON seems to be ruled mainly by factors
other than crystallinity. In fact, it was reported that for LaTiO2N photoanodes with different morphologies, the surface area
had the most important effect on the photoactivity.30 Thus, the surface area of the oxynitrides fabricated for this study was
determined using BET. As shown in Figure 6, a clear trend was observed between the photocurrent densities of the CNON
photoanodes and their BET surface areas. CNON(Nb)-PC had a surface area more than two and six times higher than
CNON-SS and CNON-PC, respectively. The larger surface would result in more sites for hole injection into the electrolyte,
that is more active sites for oxygen evolution.
Another advantage of the agglomerate structure of CNON(Nb)-PC is the low angle interparticle boundaries which allow
the formation of densely packed films that would result in improved electron transfer. For CNON-SS particles, a closely
packed film couldn't be formed due to the high angle interparticle boundaries as seen in the SEM image (Figure 2). One of
the main factors to be considered for a rational design of particle-based photoelectrodes is that the particles should
assemble into dense films providing large interparticle interface area.30 This, in addition to the higher surface area, explain
the improved performance of CNON(Nb)-PC electrodes compared to the others. In comparison, the low photoactivity of
CNON-PC is attributed to the low surface area, and to the large nonporous cuboid particles shape which lead to poor charge
transport as a result of its long migration distance.
Beside the morphological features, an important factor that influences the photoactivity of the oxynitride photoanodes
is the visible light absorption properties and the N content. As discussed above, the high absorption background of both
CNON-SS and CNON-PC shown in Figure 3b, suggests a greater amount of defects for these two oxynitrides compared to
CNON(Nb)-PC. The defects act as recombination centers for the photogenerated e--h+ pairs, thus reducing the photoactivity
of the photoanodes. Additionally, the higher N content of CNON(Nb)-PC measured by TG improves the visible light
absorption and thereby increases the PEC performance in comparison to the other two oxynitrides.
Figure 6. The photocurrent density of CNON oxynitrides measured at 1.23 V vs RHE (black squares) as a function of the BET surface
area and a line shows the linear fitting of the data with slope of 1.04 µA.cm-2 and R2=0.989.
Indeed, it is expected that better performance could be achieved by optimizing the Nb enrichment in the oxide precursor.
The XRD analysis of CNON(Nb)-PC did not show evidence of formation of secondary phases, whose presence may be
expected in case that part of the excess Nb is not incorporated in the oxynitride. However, XRD analysis would not detect
easily the presence of secondary phases with low crystallinity and/or very small average grain size. It was reported that up
to 0.4 molar excess Nb improves PEC performance of BaNbO2N, but for larger quantities of Nb excess the PEC performance
decreases and the particle morphology changed showing cracks and the typical features of local segregation and
agglomeration which were ascribed to secondary phases formed by the unreacted excess Nb. It was also shown that
interestingly the 0.4 molar Nb excess corresponded to the value that resulted in a stoichiometric 1:1 ratio of the Ba and Nb
content at the surface of the photoanode.17 Future studies are planned to identify the optimal level of Nb excess in the
synthesis of CNON.
Effect of Co-catalyst and ALD layers on PEC Performance. The slow kinetics of the oxygen evolution and the surface
recombination of the photogenerated carriers usually limit the performance of photoanodes used for solar water splitting.2,
14-15, 27 To tackle these issues, a co-catalyst is loaded on the semiconductor photoelectrode. The co-catalyst first captures the
photogenerated holes thereby reducing the recombination of the photoinduced charge carriers. Then, the oxygen evolution
reaction proceeds at the co-catalyst surface.
Among the different co-catalysts used to improve the photoactivity of photoanodes, Co-Pi is advantageous as it can be
regenerated during the PEC experiments and also because of the ease of incorporating it into complex morphologies.35
Several photoanodes, such as WO3,36 ZnO,37 α-Fe2O3,38-39 and W:BiVO4
35, showed improved photoactivity when loaded with
the Co-Pi co-catalyst. For the present study, Co-Pi was loaded on CNON photoanodes via photodeposition. This method
was selected because it allows the deposition of the co-catalyst specifically in the sites where the photogenerated holes are
more readily accessible,37 thus, directly in the most active sites leading to better performance with reduced amount of Co-
Pi.40 The effect of Co-Pi loading was investigated using the best performing photoanode CNON(Nb)-PC. The
photodeposition of Co-Pi was confirmed by SEM images. As clearly seen in Figure S1b, the co-catalyst nanoparticles (with
average grain size of 10-15 nm) appeared homogeneously distributed on the CNON(Nb)-PC photoanode. The effect of Co-
Pi on the photoactivity of the CaNbO2N is assessed by comparing the PEC activity of bare CNON(Nb)-PC and CNON(Nb)-
PC/Co-Pi photoanodes (Figure 7).
The photocurrent was enhanced with the photodeposition of Co-Pi to reach 19 and 25µA.cm-2 at 1.23 and 1.5V vs RHE,
respectively. This corresponds to an increment of about 90 and 80% of the photoactivity.
To complete the investigation of the potentials of CNON toward solar water splitting, we also probed the effect of two
43-
different passivation layers on the PEC activity. Several studies have shown that overlayers of oxides such as TiO2,41-42 Al2O3
46 and Ga2O3
47-48 have beneficial effects on the performance of photoanodes towards solar water splitting. These overlayers
passivate the surface of the photoanodes by reducing the average density of surface defects. The surface defects, typically
oxygen vacancies, act as recombination sites of the photogenerated charge carriers thus lowering the PEC performance of
the photoelectrodes. For this study, overlayers of TiO2 and Al2O3 were deposited by atomic layer deposition (ALD) on
CNON(Nb)-PC/Co-Pi photoanodes.
As can be seen in Figure 7, the PEC performance of CNON(Nb)-PC/Co-Pi photoanodes remained almost the same after
the deposition of TiO2 overlayers with photocurrents of 17 and 24 µA.cm-2 at 1.23 and 1.5 V vs RHE, respectively. Instead,
after the deposition of Al2O3 an enhancement in the photocurrent up to 70 and 96 µA.cm-2 at 1.23 and 1.5 V vs RHE was
observed. This represents a 7-fold increase in the photocurrent compared to the bare photoanode. A similar observation
was reported for hematite photoanodes whereby an increase by a factor of three of the photocurrent was achieved when
passivating the surface with an alumina overlayer while no beneficial effect was observed with a TiO2 layer. The
improvement of the photocurrent was shown to be due to the passivation of the surface by the Al2O3 layer and not due to
a catalytic effect. 46
Figure 7. The potentiodynamic measurements of (a) bare CNON(Nb)-PC, (b) CNON(Nb)-PC/Co-Pi, (c) CNON(Nb)-PC/Co-
Pi/TiO2 and (d) CNON(Nb)-PC/Co-Pi/Al2O3 photoanodes.
For CNON(Nb)-PC, a significant surface density of defects, such as for instance oxygen vacancies, acting as recombination
sites, may arise from the presence of reduced Nb ions near the surface. The deposition of an alumina overlayer may
compensate these vacancies at the surface of the photoanodes. In addition, the alumina layer with more O2- ions compared
to the oxynitride acts as an electron-rich layer which helps repelling of the photogenerated electrons from the surface thus
reducing the extent of the surface recombination with holes.49
The photoelectrochemical stability of CNON photoanodes in NaOH (0.5 M) was studied by measuring the photocurrent
at an applied potential of 1.23 V vs RHE for 30 min under illumination. Figure 8a shows the results of the stability tests
performed for the bare CNON(Nb)-PC, CNON(Nb)-PC/Co-Pi, CNON(Nb)-PC/Co-Pi/TiO2 and CNON(Nb)-PC/Co-Pi/Al2O3
photoanodes.
The rapid reduction of the photoactivity of the bare CNON(Nb)-PC photoanode is attributed to photooxidation of CNON,
whereby the photogenerated holes promote the oxidation of N3- to N2. This is a common problem reported for different
oxynitrides.10
After the photodeposition of Co-Pi, a decrease of the photocurrent is still observed; that is mainly due to the incomplete
coverage of the CNON surface by Co-Pi. However, after the modification of CNON surface withthe TiO2 layer, the
photostability improved. The comparison of Figure 7 and 8a show that, though the TiO2 layer did not improve the
photocurrent nor reduce the recombination losses, this layer can indeed be used as a protective layer for the oxynitride.
The TiO2 coating allowed in fact to achieve and keep a stable value of photocurrent higher than that achieved using only
the Co-Pi co-catalyst.
Figure 8. (a) Potentiostatic measurement showing the stability of bare CNON(Nb)-PC, CNON(Nb)-PC/Co-Pi, and CNON(Nb)-
PC/Co-Pi/Al2O3 photoanodes for 30 min at 1.23 V vs RHE. The inset shows an enlarged view for the first 100sec of the stability
measurement. (b) XPS spectra of CNON(Nb)-PC/Co-Pi before the stability test and CNON(Nb)-PC/Co-Pi/Al2O3 after the stability
test.
Differently, the CNON(Nb)-PC/Co-Pi/Al2O3 photoanode showed remarkably better performance compared to the
uncoated sample. The initial photocurrent density was 3 to 4 times higher than that measured for the Co-Pi/TiO2 coated
samples. The photocurrent however decreased to a half of the initial value within the first 10 minutes. This effect is mainly
due to the dissolution of Al2O3 in the NaOH electrolyte, as confirmed by the XPS analyses shown in Figure S4.
This observation is in agreement with what one would expect considering the etching properties of alumina reported in
literature.50 In an alkaline solution with pH of 13, a 2 nm thick alumina layer is expected to be dissolved at room temperature
in about 3 minutes. Nevertheless, an alumina overlayer was reported to be beneficial also to boost the photoactivity of
hematite photoanodes. 46 In that study 1 M NaOH was used as the electrolyte with similar value of pH as in the present
investigation, however the alumina overlayers was reported to be sufficiently stable for at least 30 minutes.
For our sample XPS suggests that the beneficial effect of the Al2O3 overlayer is limited by its solubility in the alkaline
medium used for this study. However, even though this overlayer dissolved during the PEC measurement, it can be observed
in Figure 8a that CNON(Nb)-PC/Co-Pi/Al2O3 still maintained a photoactivity about 50% higher than the CNON(Nb)-
PC/Co-Pi photoanode without a passivation layer.
To gain more insights into the observed enhancement of photocurrent density, XPS measurements were conducted to
investigate the surface chemistry of the Co-Pi loaded CNON(Nb)-PC photoanodes with and without passivation layer
(Figure 8b). XPS indicates that 1) the as prepared sample had a significant Nb surface enrichment (Ca/Nb = 0.77), 2) the Nb
enrichment increased after the deposition of the Al2O3 overlayer (Ca/Nb = 0.48) but 3) the Nb content remained stable after
PEC tests (Ca/Nb = 0.53).
The surface Nb enrichment of the as prepared sample is likely to arise from the thermal treatments required for the
fabrication of the photoanode which results in the segregation of Nb to the surface. The further enrichment detected after
Al2O3 deposition could be due to the additional thermal treatment the sample undergoes during ALD. However, the PEC
characterization of samples that underwent the same thermal treatments but without the Al2O3 deposition showed no
enhancement of photocurrent. We thus conclude that the Nb surface enrichment is not the reason of the improved
performance of CNON(Nb)-PC/Co-Pi/Al2O3. Besides, ascribing the increased PEC performance to Nb segregation would
not agree with previous measurements on BaNbO2N17 showing that a 1:1 Ca/Nb ratio at the surface leads to the highest
photoactivity. Actually, as previously pointed out, assuming similar behavior for CNON and BaNbO2N, the XPS
measurements reported here suggest that the performance of CNON(Nb)-PC/Co-Pi/Al2O3 could be further improved by
optimizing the chemical composition of the surface to reach a 1:1 Ca/Nb ratio.
One alternative possibility to explain the enhanced photoactivity of CNON(Nb)-PC/Co-Pi/Al2O3 is that the presence of
the O-rich alumina layer at the surface favors the oxidation of the Nb ions decreasing the amount of Nb4+ that act as
recombination centers. However, also this scenario can be ruled out. Figure S5 in fact shows that the Nb4+ and Nb5+
components contribute equally to the Nb 3d XPS spectrum of the photoanodes with and without Al2O3 overlayer.
Finally, we suggest that the observed significant and prolonged improvement of the photocurrent density could be
attributed to the increased roughness of the surface after the dissolution of the passivation layer. An increased surface
roughness would result in an increased semiconductor/electrolyte interface that would in turn improve the PEC
performance of the photoanodes. For instance, it was shown that the leaching of Cr from outermost surface of Fe2O3
modified by a Fe20Cr40Ni40Ox layer increases the surface roughness.51 To verify this hypothesis, the electrochemical active
surface area of the CNON(Nb)-PC/Co-Pi and CNON(Nb)-PC/Co-Pi/Al2O3 after PEC experiments was evaluated. From the
change of current in the non-Faradaic region, which corresponds only to the charge and discharge of the electric double
layer, with respect to the scan rate, the electrochemical area can be estimated. Cyclic voltammetry (CV) measurements
were performed at different scan rates for the two photoanodes. Figure 9 shows the values of current measured from the
CV scans versus the scan rate for the CNON(Nb)-PC/Co-Pi with and without Al2O3 overlayer after the stability test shown
in Figure 8a.
The slope of the fitted lines corresponds to the total Helmholtz capacitance (𝐶𝐻) of the photoanode:
CH = CH,sp × A
where 𝐶𝐻,𝑠𝑝 is the specific Helmholtz capacitance (F.cm-2) and 𝐴 is the electrochemical active surface area (in cm2). There
are no previous reports on the value of 𝐶𝐻,𝑠𝑝 for CNON, however, since this value is the same for both photoanodes, the
difference in 𝐶𝐻 (slope of the fitted lines in Figure 9) can only be due to the change of the surface area. For CNON(Nb)-
PC/Co-Pi/Al2O3 the slope of the curve in Figure 9 is 2.3 times higher than that calculated for CNON(Nb)-PC/Co-Pi. This
means that after the stability test, the dissolution of Al2O3 results in an increased surface area of the photoanode.
By comparing the photocurrent values for both photoanodes after the potentiostatic measurements shown in Figure 8a,
an enhancement in the photocurrent by a factor of 2.8 can be observed for the Al2O3-modified sample.
This enhancement is similar to the estimated increase of the electrochemical active surface area. Therefore, we conclude
that even when the Al2O3 overlayer is dissolved, a prolonged photocurrent enhancement is observed mainly due to the
increased surface area.
Further studies are required to investigate the use of other overlayer materials and to understand why Al2O3 and TiO2
have different functionalities.
Figure 9. Plot of current as a function of scan rate for CNON(Nb)-PC/Co-Pi and CNON(Nb)-PC/Co-Pi/Al2O3 photoanodes after
the photostability measurement.
4. CONCLUSION
In summary, we showed in this work the first photoelectrochemical water splitting measurements for CaNbO2N (CNON)
photoanodes. CNON synthesized from different oxide precursors resulted in different performances of the photoanodes,
mainly due to the different morphologies and surface areas. The Nb-enriched oxide resulted in the best performing
photoanode, which was further improved by the photodeposition of Co-Pi co-catalyst and the ALD deposition of Al2O3
overlayer. An improved PEC activity of the CNON(Nb)-PC/Co-Pi/Al2O3 photoanode was observed even after the Al2O3 layer
was completely dissolved in the alkaline electrolyte. This is attributed to the increased surface roughness of the
photoanodes after the dissolution of Al2O3 layer. The modification of CNON(Nb)-PC/Co-Pi photoanodes with an ALD layer
of TiO2 didn't result in an enhanced PEC activity; however it improved the stability of the photoanodes.
These results encourage further investigations using other passivation layers for an improved photoactivity and stability
of the CNON photoanodes.
ASSOCIATED CONTENT
Supporting Information
Lattice parameters of calcium niobium oxides and oxynitride, SEM images of CNON(PC)-Nb photoanode before and after loading
of Co-Pi co-catalyst, Tauc plots of CNON(PC)-Nb, CNON-PC and CNON-SS, potentiodynamic scans of before and after post
treatment, XPS of Al-2p and Nb-3d peaks. This material is available free of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
* Emails: [email protected], [email protected]
Present Addresses
§ (W.Si) Key Laboratory of Advanced Ceramics and Machining Technology, Ministry of Education, School of Materials Science
and Engineering, Tianjin University, Tianjin 300072, P.R. China.
ACKNOWLEDGMENT
This research was supported by the Paul Scherrer Institut and the NCCR MARVEL funded by the Swiss National Science
Foundation. The Swiss Excellence Governmental Scholarship is gratefully acknowledged for the financial support of this work.
Fatima Haydous was a Swiss Government Excellence Scholarship holder for the academic years 2014-2017 (ESKAS No. 2014.0282).
The authors would like to thank Dr. Bilal Gökce from the University of Duisburg-Essen for the support with the BET measurements
and Dr. David Tilley from the University of Zurich for the help with the atomic layer deposition.
REFERENCES
1.
2.
Lewis, N. S. Toward Cost-Effective Solar Energy Use. Science (N.Y.) 2007, 315, 798-801.
Gujral, S. S.; Simonov, A. N.; Higashi, M.; Fang, X.-Y.; Abe, R.; Spiccia, L. Highly Dispersed Cobalt Oxide on Taon as Efficient
Photoanodes for Long-Term Solar Water Splitting. ACS Catal. 2016, 6, 3404-3417.
3.
Allam, N. K.; Shaheen, B. S.; Hafez, A. M. Layered Tantalum Oxynitride Nanorod Array Carpets for Efficient
Photoelectrochemical Conversion of Solar Energy: Experimental and Dft Insights. ACS Appl. Mater. Interfaces 2014, 6, 4609-4615.
4.
Kubota, J.; Domen, K. Photocatalytic Water Splitting Using Oxynitride and Nitride Semiconductor Powders for Production of
Solar Hydrogen. Electrochem. Soc. Interface 2013, 22, 57-62.
5.
Maeda, K.; Higashi, M.; Siritanaratkul, B.; Abe, R.; Domen, K. SrNbO2N as a Water-Splitting Photoanode with a Wide Visible-
Light Absorption Band. J. Am. Chem. Soc. 2011, 133, 12334-12337.
6.
Pichler, M.; Szlachetko, J.; Castelli, I. E.; Marzari, N.; Döbeli, M.; Wokaun, A.; Pergolesi, D.; Lippert, T. Determination of
Conduction and Valence Band Electronic Structure of Latioxny Thin Film. ChemSusChem 2017, 10, 2099-2106.
7.
Maegli, A. E.; Pokrant, S.; Hisatomi, T.; Trottmann, M.; Domen, K.; Weidenkaff, A. Enhancement of Photocatalytic Water
Oxidation by the Morphological Control of LaTiO2N and Cobalt Oxide Catalysts. J. Phys. Chem. C 2014, 118, 16344-16351.
8.
Singh, R. B.; Matsuzaki, H.; Suzuki, Y.; Seki, K.; Minegishi, T.; Hisatomi, T.; Domen, K.; Furube, A. Trapped State Sensitive
Kinetics in LaTiO2N Solid Photocatalyst with and without Cocatalyst Loading. J. Am. Chem. Soc. 2014, 136, 17324-17331.
9.
Yamakata, A.; Kawaguchi, M.; Nishimura, N.; Minegishi, T.; Kubota, J.; Domen, K. Behavior and Energy States of Photogenerated
Charge Carriers on Pt- or CoOx-Loaded Latio2n Photocatalysts: Time-Resolved Visible to Mid-Infrared Absorption Study. J. Phys. Chem.
C 2014, 118, 23897-23906.
10.
Higashi, M.; Domen, K.; Abe, R. Fabrication of an Efficient BaTaO2N Photoanode Harvesting a Wide Range of Visible Light for
Water Splitting. J. Am. Chem. Soc. 2013, 135, 10238-10241.
11.
Higashi, M.; Yamanaka, Y.; Tomita, O.; Abe, R. Fabrication of Cation-Doped BaTaO2N Photoanodes for Efficient
Photoelectrochemical Water Splitting under Visible Light Irradiation. APL Mater. 2015, 3, 104418.
12. Wang, C.; Hisatomi, T.; Minegishi, T.; Wang, Q.; Zhong, M.; Katayama, M.; Kubota, J.; Domen, K. Synthesis of Nanostructured
BaTaO2N Thin Films as Photoanodes for Solar Water Splitting. J. Phys. Chem. C 2016, 120, 15758 -- 15764.
13.
Abe, R.; Higashi, M.; Domen, K. Facile Fabrication of an Efficient Oxynitride TaON Photoanode for Overall Water Splitting into
H2 and O2 under Visible Light Irradiation. J. Am. Chem. Soc. 2010, 132, 11828-11829.
14.
Higashi, M.; Domen, K.; Abe, R. Fabrication of Efficient Taon and Ta3N5 Photoanodes for Water Splitting under Visible Light
Irradiation. Energy Environ. Sci. 2011, 4, 4138-4147.
15.
Higashi, M.; Domen, K.; Abe, R. Highly Stable Water Splitting on Oxynitride TaON Photoanode System under Visible Light
Irradiation. J. Am. Chem. Soc. 2012, 134, 6968-6971.
16.
Siritanaratkul, B.; Maeda, K.; Hisatomi, T.; Domen, K. Synthesis and Photocatalytic Activity of Perovskite Niobium Oxynitrides
with Wide Visible‐Light Absorption Bands. ChemSusChem 2011, 4, 74-78.
17.
Seo, J.; Moriya, Y.; Kodera, M.; Hisatomi, T.; Minegishi, T.; Katayama, M.; Domen, K. Photoelectrochemical Water Splitting on
Particulate ANbO2N (A = Ba, Sr) Photoanodes Prepared from Perovskite-Type ANbO3. Chem. Mater. 2016, 28, 6869-6876.
18.
Castelli, I. E.; Landis, D. D.; Thygesen, K. S.; Dahl, S.; Chorkendorff, I.; Jaramillo, T. F.; Jacobsen, K. W. New Cubic Perovskites
for One- and Two-Photon Water Splitting Using the Computational Materials Repository. Energy Environ. Sci. 2012, 5, 9034-9043.
19. Maeda, K.; Higashi, M.; Siritanaratkul, B.; Abe, R.; Domen, K. SrNbO2N as a Water-Splitting Photoanode with a Wide Visible-
Light Absorption Band. J. Am. Chem. Soc. 2011, 133, 12334-12337.
20.
Hisatomi, T.; Katayama, C.; Moriya, Y.; Minegishi, T.; Katayama, M.; Nishiyama, H.; Yamada, T.; Domen, K. Photocatalytic
Oxygen Evolution Using BaNbO2N Modified with Cobalt Oxide under Photoexcitation up to 740 nm. Energy Environ. Sci. 2013, 6, 3595-
3599.
21.
Landsmann, S.; Surace, Y.; Trottmann, M.; Dilger, S.; Weidenkaff, A.; Pokrant, S. Controlled Design of Functional Nano-
Coatings: Reduction of Loss Mechanisms in Photoelectrochemical Water Splitting. ACS Appl. Mater. Interfaces 2016, 8, 12149−12157.
22.
Dabirian, A.; Van de Krol, R. High-Temperature Ammonolysis of Thin Film Ta2O5 Photoanodes: Evolution of Structural, Optical,
and Photoelectrochemical Properties. Chem. Mater. 2015, 27, 708-715.
23.
Si, W.; Pergolesi, D.; Haydous, F.; Fluri, A.; Wokaun, A.; Lippert, T. Investigating the Behavior of Various Cocatalysts on LaTaON2
Photoanode for Visible Light Water Splitting. Phys. Chem. Chem. Phys. 2017, 19, 656-662.
24.
Kato, H.; Ueda, K.; Kobayashi, M.; Kakihana, M. Photocatalytic Water Oxidation under Visible Light by Valence Band Controlled
Oxynitride Solid Solutions LaTaON2-SrTiO3. J. Mater. Chem. A 2015, 3, 11824-11829.
25.
Oehler, F.; Naumann, R.; Köferstein, R.; Hesse, D.; Ebbinghaus, S. G. Photocatalytic Activity of CaTaO2N Nanocrystals Obtained
from a Hydrothermally Synthesized Oxide Precursor. Mater. Res. Bull. 2016, 73, 276-283.
26.
Lim, Y.-F.; Chua, C. S.; Lee, C. J. J.; Chi, D. Sol-Gel Deposited Cu2O and Cuo Thin Films for Photocatalytic Water Splitting. Phys.
Chem. Chem. Phys. 2014, 16, 25928-25934.
27.
Hisatomi, T.; Kubota, J.; Domen, K. Recent Advances in Semiconductors for Photocatalytic and Photoelectrochemical Water
Splitting. Chem. Soc. Rev. 2014, 43, 7520-7535.
28.
Arunachalam, P.; Al-Mayouf, A.; Ghanem, M. A.; Shaddad, M. N.; Weller, M. T. Photoelectrochemical Oxidation of Water Using
La(Ta, Nb)O2N Modified Electrodes. Int. J. Hydrogen Energy 2016, 41, 11644-11652.
29.
Jang, J.-W.; Chun, D.; Ye, Y.; Lin, Y.; Yao, X.; Thorne, J.; Liu, E.; McMahon, G.; Zhu, J.; Javey, A.; et al. Enabling Unassisted Solar
Water Splitting by Iron Oxide and Silicon. Nat. Commun. 2015, 6, 7447.
30.
Landsmann, S.; Maegli, A. E.; Trottmann, M.; Battaglia, C.; Weidenkaff, A.; Pokrant, S. Design Guidelines for High-Performance
Particle-Based Photoanodes for Water Splitting: Lanthanum Titanium Oxynitride as a Model. ChemSusChem 2015, 8, 3451 -- 3458.
31.
Zhang, F.; Yamakata, A.; Maeda, K.; Moriya, Y.; Takata, T.; Kubota, J.; Teshima, K.; Oishi, S.; Domen, K. Cobalt-Modified Porous
Single-Crystalline LaTiO2N for Highly Efficient Water Oxidation under Visible Light. J. Am. Chem. Soc. 2012, 134, 8348-8351.
32.
Feng, J.; Luo, W.; Fang, T.; Lv, H.; Wang, Z.; Gao, J.; Liu, W.; Yu, T.; Li, Z.; Zou, Z. Highly Photo-Responsive LaTiO2N Photoanodes
by Improvement of Charge Carrier Transport among Film Particles. Adv. Funct. Mater. 2014, 24, 3535-3542.
33.
Le Gendre, L.; Marchand, R.; Laurent, Y. A New Class of Inorganic Compounds Containing Dinitrogen-Metal Bonds. J. Eur.
Ceram. Soc. 1997, 17, 1813-1818.
34.
Yoon, S.; Maegli, A. E.; Eyssler, A.; Trottmann, M.; Hisatomi, T.; Leroy, C. M.; Grätzel, M.; Weidenkaff, A. Synthesis and
Characterization of La(Ti,Nb)(O,N)3 for Photocatalytic Water Oxidation. Energy Procedia 2012, 22, 41-47.
35.
Zhong, D. K.; Choi, S.; Gamelin, D. R. Near-Complete Suppression of Surface Recombination in Solar Photoelectrolysis by "Co-
Pi" Catalyst-Modified W:BiVO4. J. Am. Chem. Soc. 2011, 133, 18370-18377.
36.
Seabold, J. A.; Choi, K.-S. Effect of a Cobalt-Based Oxygen Evolution Catalyst on the Stability and the Selectivity of Photo-
Oxidation Reactions of a WO3 Photoanode. Chem. Mater. 2011, 23, 1105-1112.
37.
Steinmiller, E. M. P.; Choi, K.-S. Photochemical Deposition of Cobalt-Based Oxygen Evolving Catalyst on a Semiconductor
Photoanode for Solar Oxygen Production. Proc. Natl. Acad. Sci. 2009, 106, 20633-20636.
38.
Zhong, D. K.; Sun, J.; Inumaru, H.; Gamelin, D. R. Solar Water Oxidation by Composite Catalyst/Α-Fe2O3 Photoanodes. J. Am.
Chem. Soc. 2009, 131, 6086-6087.
39.
Zhong, D. K.; Cornuz, M.; Sivula, K.; Grätzel, M.; Gamelin, D. R. Photo-Assisted Electrodeposition of Cobalt -- Phosphate (Co -- Pi)
Catalyst on Hematite Photoanodes for Solar Water Oxidation. Energy Environ. Sci. 2011, 4, 1759-1764.
40. McDonald, K. J.; Choi, K.-S. Photodeposition of Co-Based Oxygen Evolution Catalysts on a-Fe2O3 Photoanodes. Chem. Mater.
2011, 23, 1686-1693.
41.
Yang, X.; Liu, R.; Du, C.; Dai, P.; Zheng, Z.; Wang, D. Improving Hematite-Based Photoelectrochemical Water Splitting with
Ultrathin TiO2 by Atomic Layer Deposition. ACS Appl. Mater. Interfaces 2014, 6, 12005-12011.
42.
Li, C.; Hisatomi, T.; Watanabe, O.; Nakabayashi, M.; Shibata, N.; Domen, K.; Delaunay, J.-J. Positive Onset Potential and Stability
of Cu2O-Based Photocathodes in Water Splitting by Atomic Layer Deposition of a Ga2O3 Buffer Layer. Energy Environ. Sci. 2015, 8, 1493-
1500.
43.
Zeng, M.; Peng, X.; Liao, J.; Wang, G.; Li, Y.; Li, J.; Qin, Y.; Wilson, J.; Song, A.; Lin, S. Enhanced Photoelectrochemical
Performance of Quantum Dot-Sensitized TiO2 Nanotube Arrays with Al2O3 Overcoating by Atomic Layer Deposition. Phys. Chem. Chem.
Phys. 2016, 18, 17404-17413.
44.
Neufeld, O.; Yatom, N.; Caspary Toroker, M. A First-Principles Study on the Role of an Al2O3 Overlayer on Fe2O3 for Water
Splitting. ACS Catal. 2015, 5, 7237-7243.
45.
Roelofs, K. E.; Brennan, T. P.; Dominguez, J. C.; Bailie, C. D.; Margulis, G. Y.; Hoke, E. T.; McGehee, M. D.; Bent, S. F. Effect of
Al2O3 Recombination Barrier Layers Deposited by Atomic Layer Deposition in Solid-State CdS Quantum Dot-Sensitized Solar Cells. J.
Phys. Chem. C 2013, 117, 5584-5592.
46.
Le Formal, F.; Tetreault, N.; Cornuz, M.; Moehl, T.; Gratzel, M.; Sivula, K. Passivating Surface States on Water Splitting Hematite
Photoanodes with Alumina Overlayers. Chem. Sci. 2011, 2, 737-743.
47.
Barroso, M.; Mesa, C. A.; Pendlebury, S. R.; Cowan, A. J.; Hisatomi, T.; Sivula, K.; Grätzel, M.; Klug, D. R.; Durrant, J. R. Dynamics
of Photogenerated Holes in Surface Modified Α-Fe2O3 Photoanodes for Solar Water Splitting. Proc. Natl. Acad. Sci. 2012, 109, 15640-15645.
Steier, L.; Herraiz-Cardona, I.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.; Tilley, S. D.; Grätzel, M. Understanding the Role
48.
of Underlayers and Overlayers in Thin Film Hematite Photoanodes. Adv. Funct. Mater. 2014, 24, 7681-7688.
49.
Le Formal, F.; Sivula, K.; Grätzel, M. The Transient Photocurrent and Photovoltage Behavior of a Hematite Photoanode under
Working Conditions and the Influence of Surface Treatments. J. Phys. Chem. C 2012, 116, 26707-26720.
50.
Sun, K. G.; Li, Y. V.; Saint John, D. B.; Jackson, T. N. Ph-Controlled Selective Etching of Al2O3 over Zno. ACS Appl. Mater.
Interfaces 2014, 6, 7028-7031.
51.
Bärtsch, M.; Sarnowska, M.; Krysiak, O.; Willa, C.; Huber, C.; Pillatsch, L.; Reinhard, S.; Niederberger, M. Multicomposite
Nanostructured Hematite -- Titania Photoanodes with Improved Oxygen Evolution: The Role of the Oxygen Evolution Catalyst. ACS Omega
2017, 2, 4531-4539.
TOC Graphic
14
Supporting Information
Table S1. Lattice Parameters of Ca2Nb2O7 (ICSD Coll.Code: 26010), CaNb2O6 (ICSD Coll.Code:15208), and CaNbO2N (ICSD
Coll.Code: 55396).
Material
Crystal
Structure
Space
Group
Ca2Nb2O7
monoclinic
P1121
Lattice Parameters Å
a
7.697
b
13.385
c
5.502
CaNb2O6
orthorhombic
Pbcn
14.926
5.752
5.204
CaNbO2N
orthorhombic
Pnma
5.64051
7.90711
5.55508
Figure S5. SEM images of CNON(Nb)-PC/Co-Pi photoanode (a) before and (b) after Co-Pi photodeposition.
Figure S6. Tauc plots of (a) CNON(Nb)-PC, (b) CNON-PC and (c) CNON-SS.
15
Figure S7. Potentiodynamic scans of CNON(Nb)-PC photoanodes (a) before post-treatment, (b) after post necking with TaCl5 and
annealing in air, and (c) after annealing in NH3 following the post necking procedure in 0.5 M NaOH solution under chopped
illumination with Xe lamp.
Figure S8. XPS spectra of Al-2p for CNON(Nb)-PC/Co-Pi/Al2O3 photoanode as-prepared, soaked in NaOH for 30 min, and after
the photostability test.
16
Figure S9. XPS spectra for Nb 3d peaks for (a) as-prepared CNON(Nb)-PC/Co-Pi and (b) CNON(Nb)-PC/Co-Pi/Al2O3 after PEC
photoanodes.
17
|
1710.03690 | 1 | 1710 | 2017-10-04T02:49:03 | Optical properties of spin coated Cu doped ZnO nanocomposite films | [
"physics.app-ph"
] | Spin coating technique was used to synthesize pure ZnO and Cu doped ZnO films on amorphous and conducting glass substrates. The doped amount of Cu in ZnO was varied up to 5% in atomic percentage. Speed of spin coating system, coating time, initial chemical solution and annealing conditions were varied to optimize the properties of samples. Transmittance of samples was measured for ZnO doped with 1, 2, 3, 4 and 5% of Cu. Absorbance, reflectance and refractive index were derived from the measured transmittance. Film thickness of each film was calculated using the graphs of refractive index versus wavelength. Film thickness varies in a random manner depending on the amount of ZnO or Cu doped ZnO solutions spread on the substrate. The energy gap of each film was calculated using the graph of square of absorption coefficient time photon energy versus photon energy. The calculated energy gap values of Cu doped ZnO film samples decrease with the Cu concentration in ZnO. This means that the conductivity of ZnO can be increased by adding a trace amount of conducting material such as Cu. | physics.app-ph | physics | Optical properties of spin coated Cu doped ZnO nanocomposite films
P. Samarasekara and Udumbara Wijesinghe
Department of Physics, University of Peradeniya, Peradeniya, Sri Lanka
Abstract
Spin coating technique was used to synthesize pure ZnO and Cu doped ZnO films
on amorphous and conducting glass substrates. The doped amount of Cu in ZnO
was varied up to 5% in atomic percentage. Speed of spin coating system, coating
time, initial chemical solution and annealing conditions were varied to optimize
the properties of samples. Transmittance of samples was measured for ZnO doped
with 1, 2, 3, 4 and 5% of Cu. Absorbance, reflectance and refractive index were
derived from the measured transmittance. Film thickness of each film was
calculated using the graphs of refractive index versus wavelength. Film thickness
varies in a random manner depending on the amount of ZnO or Cu doped ZnO
solutions spread on the substrate. The energy gap of each film was calculated
using the graph of square of absorption coefficient time photon energy versus
photon energy. The calculated energy gap values of Cu doped ZnO film samples
decrease with the Cu concentration in ZnO. This means that the conductivity of
ZnO can be increased by adding a trace amount of conducting material such as
Cu.
1. Introduction:
ZnO films are prime candidates of photocells, gas sensors, electronic
devices, optoelectronic devices, acoustic wave devices and piezoelectric devices.
ZnO indicates transparent properties due to its high band gap of 3.2 to 3.3 eV. As
a result, ZnO is used to absorb UV part of the solar spectrum. Most of the oxides
including ZnO are used as gas sensors. The resistivity of undoped ZnO varies
from 10-6 to 106 Ωm depending on the preparation technique. However, the
resistivity can be decreased by doping with a conducting materials or metals. ZnO
films have been synthesized on glass substrates using spray pyrolysis method 1,
sol-gel process 2 and spin coating method 3. Also thin films of ZnO have been
1
prepared using dc and rf sputtering 4 and on Si(111) substrates using pulsed laser
deposition (PLD) 5. In addition, highly aligned ZnO films have been deposited 6.
Stability of cobalt doped ZnO films with deposition temperature has been
investigated 7. Transparent conductive ZnO films doped with Co and In have been
fabricated on glass substrates at 350 oC using ultrasonic spray method 8.
According to X ray diffraction (XRD) patterns, these films have indicated a (002)
preferential direction. After doping with Co, the band energy gap of these films
has increased from 3.25 to 3.36 eV. After doping with In, the band energy gap of
these films has decreased from 3.25 to 3.18 eV. In and Co have been added to
ZnO to enhance the conductive properties of ZnO 8.
Variation of structural and electrical properties with thickness of Ga
doped ZnO films prepared by reactive plasma deposition has been investigated 9.
Li, P and N doped ZnO thin films have been grown using PLD 10. Fourier
transform infrared spectrometer (FTIR) and XRD properties of Al doped ZnO
films deposited on polycrystalline alumina substrates by ultrasonic spray pyrolysis
have been investigated 11. Optical properties and photoconductivity of ZnO thin
films grown by pulsed filtered cathodic arc vacuum technique have been studied
12. Structural and optical properties of ZnO thin films synthesized on (111) CaF2
substrates by magnetron sputtering have been investigated 13. Effect of substrate
temperature on the crystalline properties Al doped ZnO films fabricated on glass
substrates by RF magnetron sputtering method has been studied 14. Low
temperature annealing effect on the structural and optical properties of ZnO films
deposited by PLD has been investigated 15.
Previously ZnO films have been deposited using reactive dc sputtering
method by us 16. Photo-voltaic and absorption properties of ZnO thin films
deposited at different sputtering and annealing conditions were investigated 16. In
this report, the optical properties of spin coated ZnO films have been explained.
The energy gap, optical absorption, and thickness have been calculated for films
fabricated at different rotational speeds of spin coated system for different coating
times and films annealed at different temperatures for different time periods. The
energy gap, optical absorption, and thickness solely depend on the Cu
2
concentrations doped in ZnO films. The spin coating method is a low cost
technique compared to the deposition techniques required vacuum or expensive
equipments.
Low cost P-Cu2O/N-CuO
evaporation method 17. Furthermore, multiwalled carbon nanotubes synthesized
junction was prepared using
thermal
using chemical vapor deposition method was employed to detect H2 and methane
gases 18. Copper oxide was fabricated using reactive dc sputtering by us 19. Energy
gap of semiconductor particles doped with salts were determined by us 20. In
addition, ZnO indicates some magnetic properties. Theoretical studies of magnetic
films have been carried out using Heisenberg Hamiltonian 21-26.
2. Experimental:
Chemicals with purity higher than 98% were used as starting materials.
Copper acetate dehydrate (Cu(CH3COO)2.2H2O), anhydrous ethanol and
monoethanolamine (MEA), potassium iodide (KI), tetrapropylammonium iodide
(Pr4NI) and
iodine were used without grinding. Zinc acetate dihydrate
(Zn(CH3COO)2.2H2O) was grinded before use. All the chemicals except I2 and
ethanol were vacuum dried at 60 oC for 24 hours prior to use.
ZnO
solutions were prepared using zinc acetate dihydrate
(Zn(CH3COO)2.2H2O), anhydrous ethanol and monoethanolamine (MEA) as the
solute, solvent and sol stabilizer, respectively. Zinc acetate was first grinded and
dissolved in a mixture of ethanol and monoethanolamine at room temperature.
The molar ratio of MEA to zinc acetate was kept at 1:1. Five doped solutions were
prepared by adding copper acetate dihydrate (Cu(CH3COO)2.2H2O) to the
mixture with an atomic percentage of Cu varying from 1% to 5%. The resulting
solutions were stirred by a magnetic stirring apparatus at 70 ◦C for an hour.
Finally transparent ZnO solutions were formed. In the sol, the Zn concentration
was 0.5 mol/L. The prepared sols were aged for 24 hours at room temperature.
Then the thin films were prepared by a spin-coating method on glass substrates
which had been pre-cleaned by detergent, and then cleaned in methanol and
3
acetone for 10 min each by using ultrasonic cleaner and then cleaned with
deionized water and dried. The films samples were grown at 2000rpm for 30s.
After coating, the sample was first dried at 200 ◦C for 10min, and then was
annealed at 500 ◦C in ambient atmosphere for an hour.
The optical measurements of the films were carried out using Shimadzu UV
1800 spectrophotometer in the wavelength range from 190 to 900 nm at room
temperature. Samples synthesized on amorphous insulator glass substrates were
used to measure the optical properties.
3. Results and Discussion:
Experimental measurements are usually made
transmittance (
), which is defined as,
in
terms of percentage
T
%
=
I
I
0
×
%100
(1)
where I is the light intensity after it passes through the sample and Io is the initial
light intensity.
All the samples given in this report were prepared at 2000 rpm in 30s and
subsequently annealed at 500 ◦C in air for an hour. After optimizing synthesize
conditions, above conditions were selected to prepare samples. Figure 1 shows the
transmittance curves for undoped ZnO films (solid line) and ZnO doped with 3
(dashed line) and 5% (dotted line) of Cu. The atomic doping percentages of Cu
are given here. Although all the measurements were performed for doping
concentrations of 1, 2, 3, 4 and 5% of Cu, only the curves for 3 and 5% are shown
in this manuscript. The absorption edge is observed around 380 nm. Below the
absorption edge, the transmittance increases with Cu concentration in ZnO thin
film sample. However, just above the transmittance edge, pure ZnO film and film
sample with 3% of Cu indicates the highest and lowest transmittance,
respectively. At longer wavelengths, film sample with 3% of Cu dominates.
Because different materials are capable to absorb different wavelengths, the
dominance of transmittance varies with the wavelength even for the same
concentration of Cu.
4
Figure 1: Transmittance versus wavelength for undoped ZnO and doping
concentrations of 3 and 5% of Cu.
The relationship between absorbance ( ) and transmittance ( ) is given by,
A
−=
log
10
)(
T
−=
log
10
I
I
0
(2)
The curves of absorbance versus wavelength for undoped ZnO (solid line) films
and ZnO doped with 3 (dashed line) and 5% (dotted line) of Cu are given in figure
2. The absorption edge can be observed again around 380 nm. Below this
wavelength, absorption decreases with Cu concentration of ZnO sample. Just
above this wavelength, sample with 3% of Cu indicates the highest absorption.
This variation is obvious, because the absorption is the opposite phenomena of
transmittance. Again the relative variation of absorption between samples with
different concentrations is due to the reason explained previously.
5
Figure 2: Absorbance versus wavelength for undoped ZnO sample and Cu doping
The reflectance (
) was calculated using the following relation
concentrations of 3 and 5%.
−=
(1
×
5.0)
AeT
(3)
R
where R is the reflectance, T is the transmittance and A is the absorbance.
Figure 3 shows the graph between reflectance and wavelength for pure ZnO and
Cu concentrations of 3 and 5%. The curves observed in reflectance graph have
some resemblances to the absorbance curves.
6
Figure 3: Reflectance versus wavelength for undoped ZnO and Cu doped samples.
The refractive index at different wavelengths were calculated using equation,
n
=
+
1
R
−
1
R
5.0
5.0
(4)
The graph between refractive index and the wavelength is given in figure 4 for
pure ZnO and Cu doped samples. Variation of refractive index is similar to the
variation of absorbance and reflectance.
7
Figure 4: Refractive index versus wavelength for undoped ZnO and Cu doped
The thickness of the thin film can be calculated by equation
samples.
=
t
(
n
1
λλ
2
n
1
−
λ
2
(5)
λ
12
)
where λ1 and λ2 are the wavelengths at which two successive maxima or minima
occur, and n1 and n2 are the corresponding refractive indices.
The thicknesses calculated for the films with different Cu concentrations using
equation (5) and figure 4 are given in table 1. Because the thickness of film
depends on the initial amount of solution spread on the substrate, the thickness
varies from film to film in a random manner as given in table 1.
Cu
0%
1%
2%
3%
4%
5%
d/nm
115.0925
108.2954
109.0202
118.8204
110.9055
138.0017
n2
1.2136
1.12466
1.1681
1.17383
1.16073
1.19296
n1
3.96448
3.76367
3.75074
3.50546
3.70123
3.07063
λ1/nm
354.6
359
358.2
360.6
359.1
358.2
λ2/nm
487.1
899.9
899.6
899.4
899.6
899.6
Table 1: Film thicknesses of Cu doped ZnO films.
8
The fundamental absorption, which corresponds to electron excitation from the
valence band to conduction band, can be used to determine the value of the optical
band gap. The absorption coefficient α was determined using the relation
.2α
=
303
×
A
t
(6)
α for each sample at different wavelengths was calculated using figure 2 and table
1. The relationship between the absorption coefficient (α) and the incident photon
energy (hυ) can be written as
να
(7)
5.0)
hB
(
h
=
ν
−
gE
Where,
is a constant,
is the band gap energy of the material. When
is
zero, hυ =Eg from equation (7). Figure 5 shows the graph of (
)2 versus hυ for
films with ZnO (solid line) and Cu concentrations of 3 (dashed line) and 5%
(dotted line). The value of optical band gap was calculated by extrapolating the
)2 versus hυ graph to hυ axis. According to above
straight line portion of (
equation, the value of x axis at intercept of this straight line is the energy gap (Eg).
Figure 5: Graph of (αhν)2 versus photon energy (hν).
9
Energy gaps calculated from figure 5 and equation (7) are given in table 2.
Although only three curves are given in figure 5, the energy gaps calculated for all
the concentrations are given in table 2. Band gap gradually decreases with the
doped amount of Cu. The impurity energy levels and defects always contribute to
the decrease of energy gap. This implies that the addition of Cu enhances the
conductance of the sample as expected. However, this variation of energy gap
(from 3.27 to 3.18 eV) is really small because a trace amount of Cu is added. By
adding a trace amount of Cu, the conductivity of ZnO can be improved without
altering the other properties of ZnO. Similar variation has been observed for
indium doped ZnO films by some other researchers [8]. Electron concentration
dependence of the band gap shift in Cu doped ZnO could be the reason for this
decrease of energy gap. The increase of Urbach energy is the other possibility.
The energy gap of our undoped ZnO film is really close to the standard band gap
value of ZnO (3.2 to 3.3 eV). This confirms the formation of ZnO phase in thin
film. Our synthesized films were apparently transparent. Diffraction peaks
couldn't be observed in XRD patterns of our film samples by persuading that
particles are in nanometer range. Particles sizes less than 5nm can't be detected
using XRD.
Concentration
Band gap energy(eV)
0%
1%
2%
3%
4%
5%
3.27
3.26
3.25
3.24
3.24
3.18
Table 2: Calculated values of energy gap at different Cu concentrations.
10
4. Conclusion:
Films were prepared using a low cost spin coating technique. Below the
absorption edge, the transmittance of samples increases with Cu concentration of
ZnO thin film sample. The absorption edge can be observed around 380 nm.
Absorbance, reflectance and refractive index decrease with Cu concentration of
ZnO sample below this wavelength. However, a systematic variation of
transmittance, absorbance, reflectance or refractive index with wavelength
couldn't be observed above the absorption edge. The thickness of the samples
calculated from refractive index versus wavelength graphs varies from 108.3 to
138 nm depending on the amount of the solution spread on the substrate. Energy
gaps of the samples determined from the graph of (αhν)2 versus photon energy
(hν) gradually decreases with Cu concentration doped with ZnO. The reason for
the variation of the energy gap could be the electron concentration dependence of
the band gap shift.
References:
1. M. Caglar, Y. Caglar and S. Ilican, 2006. The determination of the thickness
and optical constants of the ZnO crystalline thin film by using envelope
method. Journal of Optoelectronics and Advanced Materials 8(4), 1410-1413.
2. H.F. Hussein, Ghufran Mohammad Shabeeb and S. Sh. Hashim, 2011.
Preparation ZnO thin film by using sol-gel processed and determination of
thickness and study optical properties. Journal of Materials and Environmental
Science 2(4), 423-426.
3. S. Ilican, Y. Caglar and M. Caglar, 2008. Preparation and characterization of
ZnO thin films deposited by sol-gel spin coating method. Journal of
Optoelectronics and Advanced Materials 10(10), 2578-2583.
4. A. Mosbah, A. Moustaghfir, S. Abed, N. Bouhssira, M.S. Aida, E. Tomasella
and M. Jacquet, 2005. Comparison of the structural and optical properties of
zinc oxide thin films deposited by dc and rf sputtering and spray pyrolysis.
Surface and Coatings Technology 200, 293-296.
11
5. Wang Zhao-yang, Hu Li-zhong, Zhao Jie, Sun Jie and Wang Zhi-jun, 2005.
Effect of the variation of temperature on the structural and optical properties
of ZnO thin films prepared on Si(111) substrates using PLD. Vacuum 78(1),
53-57.
6. A. Mosbah, S. Abed, N. Bouhssira, M.S. Aida and E. Tomasella, 2006.
Preparation of highly textured surface ZnO thin films. Materials Science and
Engineering B. 129, 144-149.
7. Said Benramache, Boubaker Benhaoua and Foued Chabane, 2012. Effect of
substrate temperature on the stability of transparent conducting cobalt doped
ZnO thin films. Journal of Semiconductors 33(9), 093001–1.
8. Said Benramache, Boubaker Benhaoua and Hamza Bentrah, 2013.Preparation
of transparent conductive ZnO:Co and ZnO: In thin films by ultrasonic spray
method. Journal of Nanostructure in chemistry 3, 54.
9. T. Yamada, T. Nebiki, S. Kishimoto, H. Makino, K. Awai, T. Narusawa and T.
Yamamoto, 2007. Dependence of structural and electrical properties on
thickness of polycrystalline Ga-doped ZnO thin films prepared by reactive
plasma deposition. Superlattices and Microstructures 42, 68-73.
10. J.R. Duclère, M. Novotny, A. Meaney, R. O'Haire, E. McGlynn, M.O. Henry
and J.P. Mosnier, 2005. Properties of Li -, P- and N-doped ZnO thin films
prepared by pulsed laser deposition. Superlattices and Microstructures 38,
397-405.
11. A. Djelloul, M.S. Aida and J. Bougdira, 2010. Photoluminescence, FTIR and
X-ray diffraction studies on undoped and Al-doped ZnO thin films grown on
polycrystalline α–alumina substrates by ultrasonic spray pyrolysis. Journal of
Luminescence 130, 2113-2117.
12. H. Kavak, E.S. Tuzemen, L.N. Ozbayraktar and R. Esen, 2009. Optical and
photoconductivity properties of ZnO thin films grown by pulsed filtered
cathodic vacuum arc deposition. Vacuum 83(3), 540-543.
13. Yinzhen Wang and Benli Chu, 2008. Structural and optical properties of ZnO
thin films on (111) CaF2 substrates grown by magnetron sputtering.
Superlattices and Microstructures 44, 54-61.
12
14. Zhiyun Zhang, Chonggao Bao, Wenjing Yao, Shengqiang Ma, Lili Zhang and
Shuzeng Hou, 2011. Influence of deposition temperature on the crystallinity of
Al-doped ZnO thin films at glass substrates prepared by RF magnetron
sputtering. Superlattices and Microstructures 49, 644-653.
15. B.L. Zhu, X.Z. Zhao, F.H. Su, G.H. Li, X.G. Wu, J. Wu and R. Wu, 2010.
Low temperature annealing effects on the structure and optical properties of
ZnO films grown by pulsed laser deposition. Vacuum 84(11), 1280-1286.
16. P. Samarasekara, A.G.K. Nisantha and A.S. Disanayake, 2002. High Photo-
Voltage Zinc Oxide Thin Films Deposited by DC Sputtering. Chinese Journal
of Physics 40(2), 196-199.
17. P. Samarasekara, 2010. Characterization of Low Cost p-Cu2O/n-CuO
Junction. Georgian Electronic Scientific Journals: Physics 2(4), 3-8.
18. P. Samarasekara, 2009. Hydrogen and Methane Gas Sensors Synthesis of
Multi-Walled Carbon Nanotubes. Chinese Journal of Physics 47(3), 361-369.
19. P. Samarasekara and N.U.S. Yapa, 2007. Effect of sputtering conditions on
the gas sensitivity of Copper Oxide thin films. Sri Lankan Journal of Physics
8, 21-27.
20. K. Tennakone, S.W.M.S. Wickramanayake, P. Samarasekara and, C.A.N.
Fernando, 1987. Doping of Semiconductor Particles with Salts. Physica Status
Solidi (a)104, K57-K60.
21. P. Samarasekara and Udara Saparamadu, 2013. Easy axis orientation of
Barium hexa-ferrite films as explained by spin reorientation. Georgian
Electronic Scientific Journals: Physics 1(9), 10-15.
22. P. Samarasekara and Udara Saparamadu, 2012. Investigation of Spin
Reorientation in Nickel Ferrite Films. Georgian electronic scientific journals:
Physics 1(7): 15-20.
23. P. Samarasekara and N.H.P.M. Gunawardhane, 2011. Explanation of easy axis
orientation of ferromagnetic films using Heisenberg Hamiltonian. Georgian
electronic scientific journals: Physics 2(6): 62-69.
24. P. Samarasekara. 2008. Influence of third order perturbation on Heisenberg
Hamiltonian of thick ferromagnetic films. Electronic Journal of Theoretical
13
Physics 5(17): 227-236.
25. P. Samarasekara and Udara Saparamadu, 2013. In plane oriented Strontium
ferrite thin films described by spin reorientation. Research & Reviews: Journal
of Physics-STM journals 2(2), 12-16.
26. P. Samarasekara, 2008. Four layered ferromagnetic ultra-thin films explained
by second order perturbed Heisenberg Hamiltonian. Ceylon Journal of
Science 14, 11-19
14
|
1810.02000 | 1 | 1810 | 2018-10-03T23:07:35 | 3D woodpile structure tunable plasma photonic crystal | [
"physics.app-ph",
"physics.plasm-ph"
] | A 3D woodpile structure tunable plasma photonic crystal is designed, simulated, and experimentally characterized over the S - X band of the electromagnetic spectrum. The measurements confirm that the electromagnetic response is rich in dynamics. The photonic crystal's reconfigurability, achieved through individual discharge control of the properties of the woodpile plasma columns, offers an unprecedented opportunity to better resolve the interactions between both the Bragg and localized surface plasmon modes which dominate the spectrum at lower frequencies. Both the experiments and simulations reveal evidence for the coupling of Bragg and surface plasmon modes through a Fano resonance. | physics.app-ph | physics | 3D woodpile structure tunable plasma photonic
crystal
B.Wang, J.A. Rodr´ıguez, and M.A. Cappelli
Department of Mechanical Engineering, Stanford University, Stanford, CA
94305-3032
E-mail: [email protected]
Abstract. A 3D woodpile structure tunable plasma photonic crystal is designed,
simulated, and experimentally characterized over the S - X band of the electromagnetic
spectrum. The measurements confirm that the electromagnetic response is rich
in dynamics. The photonic crystal's reconfigurability, achieved through individual
discharge control of the properties of the woodpile plasma columns, offers an
unprecedented opportunity to better resolve the interactions between both the
Bragg and localized surface plasmon modes which dominate the spectrum at lower
frequencies. Both the experiments and simulations reveal evidence for the coupling of
Bragg and surface plasmon modes through a Fano resonance.
8
1
0
2
t
c
O
3
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
1
v
0
0
0
2
0
.
0
1
8
1
:
v
i
X
r
a
3D woodpile structure tunable plasma photonic crystal
2
Photonic crystals (PCs) are artificially-ordered materials designed to transmit or
reflect electromagnetic (EM) waves over a limited range of frequencies as a result of
destructive and constructive Bragg scattering interferences within the periodic structure
[1]. A plasma photonic crystal (PPC) [2, 3, 4, 5, 6] is a PC that uses controllable gaseous
plasma elements to offer a degree of tunability or reconfigurability. The simplest PPCs
consist of one-dimensional (1D) striated plasma-vacuum or plasma-dielectric layers [2, 7].
The most commonly-studied 1D PPCs are laser-produced plasma Bragg gratings[8, 9].
Two-dimensional (2D) and three-dimensional (3D) PPCs are arrays of stand-alone
plasma structures [4, 5] or dielectric/metallic PCs that incorporate repeating plasma
structures. As in passive PCs, PPCs can be functionalized by introducing defects that
break the crystal periodicity. For example, reconfigurable line defects in a 2D PPC have
been shown to result in switchable waveguiding [10]. Passive dielectric or metal PCs can
also be functionalized by incorporating plasma defects to produce novel devices such as
tunable bandpass filters [6] and power limiters [11]. Plasmonic effects in PPCs, such as
the excitation of bulk or localized (surface) plasmons, can further enhance the EM wave
interactions[7, 12], resulting in deep and wide attenuation bands. In 2D PPCs, these
plasmonic effects arise for TE polarization of the incident field (E-field ⊥ to plasma
rods) at frequencies below the plasma frequency, ωp.
In this letter we describe the construction and performance of a 3D PPC consisting
of a woodpile lattice structure formed from intertwined orthogonal nested 2D arrays
constructed from gaseous plasma columns. The woodpile lattice structure allows for the
excitation of both surface plasmon and Bragg scattering modes for arbitrary incident
polarization. The plasma column properties can be individually controlled allowing
reconfigurability of the PC. For example, we can turn off the plasma of all of the columns
oriented in one direction, transforming the structure into a lower dimensionality 2D
array. As described below, we show that there is a synergistic interaction of the two
nested 2D arrays, affected somewhat by the excitation of surface plasmon polaritons.
Figure 1 shows a photograph of the 3D PPC woodpile structure consisting of layers
of 5 x 5 parallel electrical plasma discharge tubes with a lattice spacing of a = 50 mm.
Each neighboring layer is rotated by 90 degrees and spaced a/2 = 25 mm from the
previous layer while offset by a/2 = 25 mm to form the woodpile configuration with a
total of 10 layers, as illustrated in Fig. 2. The discharge tubes are held in place at four
of the crystal boundaries by an acrylic scaffold to form the woodpile structure. The
individually controllable alternating-current (AC) discharge tubes have a 290 mm long
quartz envelope which is 15 mm in outer diameter with a 1 mm wall thickness. The tubes
are filled with argon, with added mercury to a pressure of 250 Pa. The gas temperature
during nominal operation is approximately 330 K, resulting in a mercury vapor partial
pressure of about 3.5 Pa. Each discharge is driven individually by an AC ballast with a
√
peak-to-peak voltage, Vpp = 160V. The voltage waveform of each discharge is triangular
3 V. The
in shape resulting in a root-mean-square (RMS) voltage of VRM S = 80/
ballasts have a variable peak current (also close to triangular) to control the discharge
parameters, ranging from 24.8 mA to 111.1 mA, with a ballast frequency that decreases
3D woodpile structure tunable plasma photonic crystal
3
Figure 1.
arranged in a woodpile configuration.
3D plasma photonic crystal consisting of 50 plasma discharge tubes
linearly from 55.0 kHz to 37.0 kHz for increasing peak current in the range from 24.8
mA to 51.2 mA, and a ballast frequency in the range of 32.2 kHz to 33.8 kHz for peak
discharge currents from 54.4 mA to 111.2 mA. Based on previous studies[6], where these
discharge tubes were used to tune the vacancy defect of a 2D PC, we estimate that the
average plasma density within the discharge is ne(cm−3) ≈ 5 × 109Ip(mA).
We use the ANSYS commercial software package referred to as the High Frequency
Electromagnetic Field Simulator (ANSYS HFSS 16) to compute the electromagnetic
fields and transmission spectra. The simulations are carried out in 3D. The plasma
columns are modeled with a frequency dependent Drude dielectric constant given by
εp = 1 −
ω2
p
(1)
ωp is the plasma frequency, given by ωp =(cid:112)nee2/meεo with ne, e, and me the column
where ω is the EM wave frequency, ν is the electron collisional damping rate, and
ω(ω + iν)
electron number density, electron charge, and electron mass, respectively, and εo is the
vacuum permittivity. As described in our prior study [6], to account for the radially
non-uniform plasma density we assume in the simulations that the plasma electrons are
distributed uniformly over a reduced diameter of 9.2 mm, resulting in a region of electron
3D woodpile structure tunable plasma photonic crystal
4
Figure 2.
(a) Perspective schematic of the 3D woodpile PPC. The solid red and
dashed black rectangles indicate the position of the microwave horn antennas for both
the experiments and simulations. (b) x-z view of the woodpile structure showing the
3D lattice structure. The k-vector of the incident wave is in the y-direction. (c) y-z
view of the woodpile structure depicting the lattice constant a = 50 mm, staggered by
a/2 (25 mm) in the z-direction for each array element in the y axis. The quartz tubes
are represented in teal, while the plasma rods are shown in purple.
density that is twice that of the case where the electrons are distributed uniformly over
the entire inner tube diameter of d = 13 mm. We assume a reasonable value of ν = 1
GHz for the electron collisional damping rate, as this value resulted in good agreement
between the measured and experimental transmission in previous studies[6]. We also
include the presence of the quartz tube with an assumed dielectric constant ε = 3.8. As
illustrated in Fig. 2(a), the source antenna is placed at the the location depicted by the
red rectangle with polarization (E-field) in the z-direction (transverse to the first-layer
plasma columns). The receiver antenna is placed at the depicted location of the black
dashed-line rectangle on the back side of the PPC. Radiative conditions are imposed on
all external boundaries.
In the experiments, a pair of broadband microwave horn antennas (A INFO LB-
20180 2 GHz - 18 GHz) are used as the source and detector, located in the rectangles
shown in Fig. 2(a), spaced 30 cm apart, also with the E field polarized in the z-axis.
(a)(b)(c)zyxxzyz25 mm50 mmE 3D woodpile structure tunable plasma photonic crystal
5
The antennas are connected to a HP 8722D Vector Network Analyzer to measure the s-
parameters, particularly S21, which represents the EM transmission through the crystal.
The measured transmission reported below was recorded with an integration time of 5
ms per frequency point in the scan, with a source power set at -5 dBm.
The use of plasmas allows for localized surface plasmon (LSP) excitation at the
plasma - quartz/air interface at frequencies below the plasma frequency when the E-field
is in a plane perpendicular to the plasma columns. For 2D PPCs, LSPs are excited for
TE polarizations only, with Bragg scattering modes expected for both polarizations. As
we show below, this 3D PPC exhibits characteristics attributable to both LSP and Bragg
scattering, and the attenuation is not simply the linear combination of that associated
with the interwoven 5 x 5 orthogonal 2D PPCs, but instead, exhibits synergistic effects
arising from the interplay between the EM activity contributed by both lattices in this
woodpile configuration.
Experimental and simulated transmission spectra are shown in Fig. 3. With
the plasma discharges turned off (Fig. 3(a)) attenuation is fairly shallow, but there
are discernible band gaps attributable to the quartz throughout the spectrum, most
notably near 3 GHz, with some correspondence to features in the simulations. With
the plasma on and at relatively low discharge currents IP = 24.8 mA (Fig. 3(b)) and
IP = 28.0 mA (Fig. 3(c)), the refractive index of the plasma is still close to unity and
the higher frequency regions of the spectrum show little change in comparison to the
plasma off case, indicating that the presence of the plasma has but a small effect on
the quartz photonic bands in this region, nor do we see any new plasma-specific bands.
However, there is a marked difference at frequencies below the plasma frequency. For
example, in Fig. 3(b), we see the emergence of what we attribute to an LSP band
at a frequency below the lowest quartz Bragg gap recorded in the experiment.
In
cylindrical plasma columns, when the incident wavelengths are much larger than the
column radius, we expect the LSP resonance to be at approximately ωLSP = ωp√
[13].
The location of the expected LSP in the figures are based on the approximate relation
between ne and Ip given earlier. The approximate location of the PPC Bragg resonance
(BG) identified in the figures is based on the location of the quartz gap, as there is no
definitive means of isolating that gap from the effect of the LSP feature. At the slightly
higher discharge current (Fig. 3(c)), the LSP resonance may be nearly coincident with
the PPC Bragg gap and a complex spectrum emerges, suggestive of a lattice (Fano)
resonance [12], with a second peak emerging near 4 GHz. The simulated spectra for
these intermediate discharge current cases (ne = 2 × 1011cm−3, or ωp = 4 GHz and
ne = 3.8 × 1011cm−3, or ωp = 5.5 GHz, respectively) capture the experimental spectra
reasonably well, most notably, the emergence of a second peak (in Fig. 3(c)) which may
be due to the spectral splitting of coincident coupled resonators generally seen in Fano
resonances. At the highest current density case (Fig. 3(d)), this coupled resonance is
quite pronounced as indicated by the depth of the attenuation and a splitting of the
measured spectrum. For comparison, we include a simulated spectrum for a plasma
frequency ωp = 7 GHz (ne = 6 × 1011cm−3), although it is apparent that the predicted
2
3D woodpile structure tunable plasma photonic crystal
6
(a)(b)(c)(d)abLSPLSPQBGBGBGLSPBG 3D woodpile structure tunable plasma photonic crystal
7
Figure 4. Experimental transmission (S21) through the 3D woodpile plasma photonic
crystal for all plasma columns turned on, horizontal (x-aligned) plasmas turned on,
vertical (z-aligned) plasmas turned on, and sum of horizontal and vertical on tube
configurations with discharge current IP = 34.4 mA.
spectrum has the peak frequencies further apart than what is measured, in part due
to an assumed plasma density that is perhaps larger than in the experiments. The
shifted experimental transmission spectra compared to simulations in this region of the
spectrum is not surprising, as the simulations are highly idealized with the assumption
of uniform plasma columns separated from the quartz wall by vacuum. More surprising
is that for this higher current density case, we see the emergence of fairly strong Bragg
gaps at high frequency that are not fully captured by the simulations.
We plot the transmitted spectra for a discharge current of IP = 34.4 mA in Fig. 4.
Here we highlight three cases: (a) the case where the PPC is completely activated (the
horizontal, or x-aligned and vertical, or z-aligned plasma tubes in Fig. 2(a)); (b) only the
horizontal tubes are active; and (c) only the vertical tubes are active. For comparison
to case (a), we also plot the linear sum of the transmission of (b) and (c). When only
the vertical (z-aligned) tubes are active we do not expect LSP excitation. The single
attenuation band seen at low frequencies near 3.5 GHz is therefore the Bragg gap of
the 2D hexagonal array with TM polarization. There appear to be weak Bragg gaps at
frequencies > 7 GHz. When only the horizontal (x-aligned) tubes are active we see a split
spectrum at low frequency for the 2D array with TE polarization, suggesting coupled
3D woodpile structure tunable plasma photonic crystal
8
Figure 5. Simulated E-fields (magnitude) for the 3D woodpile photonic crystal with
ωp = 7 GHz, γp = 1 GHz at f = 2.775 GHz. The fields are shown for the y-z and y-x
planes passing through the center of the PPC.
LSP and Bragg resonances. There are no discernible features at high frequencies. Above
frequencies above 3 GHz, the sum of the transmission measured for the individual 2D
PPCs is much less than that measured when all plasma columns are ignited indicating
that there is synergistic destructive interference between the two nested 2D arrays.
Below 3 GHz, it seems that a linear summation captures the measured spectrum well,
which is dominated by the 2D array with TE excitation, likely as a result of a strong LSP
attenuation. At high frequencies, the attenuation is dominated by the 2D (TM) array,
but the measured attenuation is still less than that predicted by the linear summation
of the individual spectra.
Some insight is gained from the simulations, particularly at lower frequencies, by
examining the behavior of the predicted electric fields. To orient the reader, a 3D
schematic of the PPC with superimposed E-field magnitudes (at an instant in time) is
shown in Fig. 5 for ωp = 7 GHz, γp = 1 GHz at f = 2.775 GHz. The fields are shown
for the y-z and y-x planes passing through the center of the PPC. The y-z plane in Fig.
5 is the plane where the vector E-field is displayed in Fig. 6. In Fig. 6(a), the vector
fields are shown for the low frequency feature labeled as "a" in Fig. 3(d), i.e., f = 2.775
zyxE 3D woodpile structure tunable plasma photonic crystal
9
Figure 6. Simulated y-z cross section E vector field for the 3D woodpile photonic
crystal with ωp = 7 GHz, γp = 1 GHz at (a) f = 2.775 GHz (b) f = 4.350 GHz.
9080706050403020100100E Field V/myzE9080706050403020100100E Field V/myzEf = 2.775 GHzf = 4.350 GHz(a)(b) 3D woodpile structure tunable plasma photonic crystal
10
GHz. The attenuation of the incident fields confirm the presence of Bragg interferences,
and E-field phase sweeps of the region incident on two of the plasma columns with axes
⊥ E (red and black dashed boxes) indicate the excitation of localized surface plasmons
that are in phase with the incident Bragg fields. A close-up of these regions confirm
that these are surface plasmons, with local E-fields that are amplified to levels above
that of the background. At f = 4.35 GHz (Fig. 6(b)), corresponding to the higher
frequency feature in Fig. 3(d), again we see Bragg attenuation, and the LSPs around
the same two plasma columns are even more pronounced. The presence of the LSPs
at two nearby frequencies where there appear to be strong Bragg gap attenuations
and the phase synchronization in time are consistent with a lattice or Fano resonance.
Not apparent in these figures but evident in the simulated E-field phase sweeps are
suggestions that the incident Bragg fields also couple into longitudinal surface plasmon
waves that propagate along the z-aligned plasma columns.
The measurements and simulations described here, of a 3D plasma photonic
crystal with a woodpile configuration, present an electromagnetic system that is rich
in dynamics and an opportunity for reconfigurability that is not generally found in
microwave plasma photonic crystals. We see compelling evidence for the coupling of
Bragg and LSP modes through a Fano resonance. The design allows for individual
control of each element in the 3D crystal. Future research will examine the propagation
of waves along complex defects (associated with non-activated plasma elements) that
extend in three dimensions.
Acknowledgments
This research was supported by a Multidisciplinary University Research Initiative from
the Air Force Office of Scientific Research, with Dr. Mitat Birkan as the program
manager. J.A.R. acknowledges a summer fellowship provided by the Stanford University
School of Engineering.
References
[1] Joannopoulos J J D, Johnson S, Winn J N J and Meade R R D 2008 Photonic crystals: molding the
flow of light 2nd ed (Princeton, New Jersey: Princeton University Press) ISBN 9780691124568
[2] Hitoshi H and Mase A 2004 Journal of Plasma and Fusion Research 80 89 -- 90 ISSN 0918-7928
[3] Sakai O and Tachibana K 2012 Plasma Sources Science and Technology 21 013001 ISSN 0963-0252
[4] Sakai O, Sakaguchi T and Tachibana K 2007 Journal of Applied Physics 101 ISSN 00218979
[5] Wang B and Cappelli M A 2016 Applied Physics Letters 108 161101 URL http://dx.doi.org/
10.1063/1.4946805
[6] Wang B and Cappelli M A 2015 Applied Physics Letters 107 ISSN 00036951
[7] Wang B, Righetti F and Cappelli M 2018 Physics of Plasmas 25 031902
[8] Shi L, Li W, Wang Y, Lu X, Zeng H et al. 2011 Physical review letters 107 095004
[9] Liu F, Yuan S, He B, Nan J, Jiang M, Khan A Q, Yu J, Zeng H et al. 2017 Optics Express 25
22303 -- 22311
[10] Wang B and Cappelli M A 2016 AIP Advances 6 ISSN 21583226 URL http://dx.doi.org/10.
1063/1.4954668
3D woodpile structure tunable plasma photonic crystal
11
[11] Greg´orio J, Parsons S and Hopwood J 2017 Plasma Sources Science and Technology 26 ISSN
13616595
[12] Righetti F, Wang B and Cappelli M 2018 arXiv preprint arXiv:1808.00610
[13] Pitarke J, Silkin V, Chulkov E and Echenique P 2006 Reports on progress in physics 70 1
|
1909.00707 | 3 | 1909 | 2019-12-10T03:45:06 | Fabrication of the impedance-matched Josephson parametric amplifier and the study of the gain profile | [
"physics.app-ph"
] | We designed and fabricated an impedance matched Josephson junction parametric amplifier (JPA) working in the flux-pump mode for the broadband amplification of microwave signals. We developed a very simple fabrication method suitable for a small lab. We studied the phase response, as well as the gain, as a function of frequency and pump power at various pump frequencies. The phase response can be explained with the behavior of the non-linear Duffing oscillator. The observed decrease of the resonance frequency as the pump power increases, as well as the emergency of an unstable bifurcation zone, are the characteristic non-linear behavior of the Duffing oscillator. The gain profile in the stable zone can be explained with a model adapted from the theoretical model for the two-dimensional gain profile of an impedance-matched current-pumped JPA. With an appropriate environmental impedance, the theoretical model captures the features and morphology of the gain profile, such as the emergence of a gain hot zone with two branches around the resonance frequency of the JPA. Based on the gain profile, we propose that the best working zone is the merging point of the two branches of the gain hot zone before the emergence of the bifurcation zone, which gives a large bandwidth and a good gain. Over 17dB gain with a bandwidth larger than 300MHz was observed. The impedance matched JPA is used in our superconducting quantum computers for improving the fast readout fidelity of the transmon qubits. | physics.app-ph | physics | TAS-2019-0211
1
Fabrication of the impedance-matched Josephson
parametric amplifier and the study of the gain
profile
Rui Yang and Hui Deng
9
1
0
2
c
e
D
0
1
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
3
v
7
0
7
0
0
.
9
0
9
1
:
v
i
X
r
a
Abstract -- We designed and fabricated an impedance-matched
Josephson parametric amplifier (JPA) working in the flux-pump
mode for the broadband amplification of microwave signals.
We developed a simple fabrication method suitable for a small
cleanroom. We studied the phase response, as well as the gain,
as a function of signal frequency and pump power at various
pump frequencies. The phase response can be explained with
the behavior of the non-linear Duffing oscillator. The observed
decrease of the resonance frequency as the pump power increases,
as well as the emergence of an unstable bifurcation zone, are
the characteristic non-linear behavior of the Duffing oscillator.
The features of the gain profile in the stable zone can be
explained with a model adapted from the theoretical model
for the two-dimensional gain profile of an impedance-matched,
current-pumped JPA. Over 17 dB gain with a bandwidth more
than 300 MHz (centred around 6.45 GHz) was observed. Our
impedance-matched JPA is used for improving the fast readout
fidelity of transmon qubits.
Index Terms -- Josephson junction parametric amplifier, quan-
tum computer, nanofabrication, gain profile
I. INTRODUCTION
Q UANTUM computers hold the promise for solving
many problems that can not be solved efficiently with
conventional computers[1], [2], [3], [4]. It can provide an
efficient way to simulate quantum mechanical systems[1].
In some applications, quantum algorithms also show great
advantage over conventional algorithms[2], [3], [4]. The proof-
of-concept demonstration of quantum computation and quan-
tum simulation on small scale quantum computers has been
realized[5], [6], [7], [8], [9], [10], [11]. The promising prospect
of building a large quantum computer of practical use draws
the attention of many institutions. Many countries join the race
towards a practical quantum computer. Within the last several
years, progress in the study of superconducting quantum
computers is impressive. For example, Google built a 72-
qubit quantum computer; IBM built quantum processors with
up to 50 qubits and put several quantum processors on the
cloud. Intel also built quantum processors with more than
Manuscript created September 2019. revised November 2019. This work
was supported by the University of Science and Technology of China.
Rui Yang is with Hefei National Laboratory for Physical Sciences at
Microscale and Department of Modern Physics, University of Science and
Technology of China, Hefei, Anhui 230026, China, and also with the Shanghai
Branch, CAS Center for Excellence and Synergetic Innovation Center in
Quantum Information and Quantum Physics, University of Science and
Technology of China, Shanghai 201315, China, e-mail:[email protected]
Hui Deng is with Hefei National Laboratory for Physical Sciences at
Microscale and Department of Modern Physics, University of Science and
Technology of China, Hefei, Anhui 230026, China, e-mail:[email protected]
50 qubits. A key issue in the development of a practical
multi-qubit quantum computer is the high-fidelity readout of
the signals from multiple qubits. A high distinguishability
between the qubit states is crucial for various operations on
the quantum computer, such as running quantum algorithm,
error correction and quantum feedback experiments[12], [13],
[14], [15], [16]. The microwave signals carrying information
of the qubits are very weak. Amplification is thus required.
The commercial microwave HEMT amplifier adds lots of
noise during amplification, the equivalent noise temperature is
several Kelvins[17], [18], thus degrading the distinguishability
between quantum states. To improve the readout fidelity, an
amplifier with the lowest possible added noise is favourable.
Josephson parametric amplifier (JPA) is the most popular
system that can provide such a low noise level[18], [19], [20].
To read many qubits simultaneously, a large bandwidth for the
Josephson parametric amplifier is also required. The simplest
JPA is a non-linear resonator composed of a superconducting
quantum interference device (SQUID) loop shunted by a big
capacitor. However, the bandwidth of such a resonant structure
is quite small, thus making it a narrow-band JPA. A way to
increase the bandwidth of this type of JPA is to incorporate a
gradual impedance transformer into the microwave feedline,
thus lowering the quality factor of the resonant structure
without introducing too much reflection[21]. The gain profile
of such an impedance-matched JPA differs from that of a
narrow-band JPA. The characterization and understanding of
this gain profile can help in choosing the best working zone
for the qubit readout task.
In this work, we first designed and fabricated an impedance-
matched Josephson parametric amplifier working in the flux-
pump mode[22]. The fabrication process we developed sim-
plifies the fabrication of the impedance transformer part. The
process is suitable for a small cleanroom without etching fa-
cilities for the dielectrics. We studied the phase and amplitude
of the microwave scattering parameter, as well as the gain,
as a function of signal frequency and pump power, at various
pump frequencies. The phase response reflects the behavior of
the non-linear Duffing oscillator. The observed decrease of the
resonance frequency as a function of the pump power, as well
as the emergence of an unstable bifurcation zone (which also
has an impact on the amplitude response), are characteristic
for a non-linear Duffing oscillator. The observed amplitude or
gain response has two zones. When the pump power is below
some critical level, the system is in the stable zone. When
the pump power is larger than a critical value, the system
TAS-2019-0211
2
entering the bifurcation zone, the amplitude or gain response
is unstable and a sudden change usually occurs. We adapted a
theoretical model[23] for the two-dimensional (2d) gain profile
of an impedance-matched current-pumped JPA to explain the
observed 2d gain profile in the stable zone. The theoretical
model[23] was originally used for describing the gain profile
of JPAs working in the current-pump mode, which indicates
that the environmental impedance can influence the details of
the gain profile of a JPA. With an appropriate environmental
impedance, the theoretical model captures the features of our
data, such as the emergence of a gain hot zone with two
branches around the resonance frequency of the JPA. Based
on the gain profile, we propose that the best working zone is
the merging point of the gain hot zone before the emergence
of the bifurcation zone, which gives a large bandwidth and
good gain. Over 17 dB gain with a bandwidth larger than 300
MHz centred around 6.45 GHz was observed. The impedance
matched JPA is used in our experiments for improving the fast
readout fidelity of the transmon qubits.
II. DESIGN AND DEVICE FABRICATION
The impedance-matched JPA was designed and fabricated at
the Nanofabrication Facilities at Institute of Physics in Beijing,
University of Science and Technology of China in Hefei, and
National Center for Nanoscience and Technology in Beijing.
The impedance transformer is a compact Klopfenstein filter
which gradually changes the impedance of the transmission
line from 50 Ω to a smaller value (about 15 Ω,
limited
mainly by the dielectric constant of the crossover capacitors
and the size of the JPA),
thus increasing the bandwidth
as well as the saturation power of the JPA to the desired
values, since both the bandwidth and the saturation power
are inversely proportional
to the impedance seen by the
nonlinear resonator composed of the Josephson junction and
the shunt capacitor[21]. The large shunt capacitor is designed
to be around 4 pF, the inductance of the Josephson junction
is designed to be around 70 pH, the corresponding room
temperature junction resistance is around 62 Ω. The design
of the impedance transformer and the circuit illustration can
be found in Fig. 1.
The device fabrication involves many processes. In the first
step, a thin Al film 100 nm thick was deposited on a cleaned
silicon wafer (with a 1 micron thick SiO2 layer). In the second
step, a layer of photoresist S1805 was spin-coated and baked.
Then, lithography was performed by a laser writer to form the
etch mask. After that, wet etching was performed to define
the co-planar wave-guide (CPW) by removing some Al from
the designed area. In the third step, photoresist LOR5A and
S1805 were spin-coated and lithography by a laser writer was
performed again to define the regions for the dielectrics. Then,
a layer of 200 nm thick CaF2 layer was deposited and liftoff
was performed to form the dielectric layer. In the fourth step,
LOR5A and S1805 were spin-coated again and lithography by
a laser writer was performed again to define the regions for
the top electrodes of the shunt capacitor and the crossovers on
the impedance transformer. Then, a layer of Al was deposited
again to form the top electrodes. There is no etching for the
(a)
(b)
(a) The design of the impedance transformer (the meandering line
Fig. 1.
represents the CPW transmission line, the magenta structures represent the
small crossovers). The density of the small crossovers modified the impedance
along the line. The denser the crossovers, the smaller the impedance of the
segment of transmission line will be. (b) The circuit schematic of the JPA
(the inset is a photo of our JPA). The circle with two crosses represents the
SQUID loop with two Josephson junctions, shunted by a capacitor. The line
connected with 'Bias Tee' represents the flux bias line, which tunes the flux
penetrating the SQUID loop. In addition to the DC bias Ibias, a RF bias
serving as a pump tone is also applied on this line with the help of a Bias
Tee. ωp is the frequency of the pump tone. ωs is the frequency of the signal
to be amplified. ωi is the frequency of the idle tone generated during the
amplification. The three frequencies satisfy ωp = ωs + ωi. In the operation
mode, ωs = ωi, thus ωp = 2ωs.
dielectrics in our fabrication steps. This simplified fabrication
process saves the steps of etching dielectrics. The Josephson
junction region is defined by the electron beam lithography on
a MAA/PMMA ebeam resist mask. The Josephson junctions
were deposited in a deposition machine with the double angle
evaporation method[24].
III. MODULATION CURVES
The phase angle of the microwave reflected from the
JPA can be modulated by the flux penetrating the SQUID
loop. Fig. 2 shows the phase of the reflected microwave as
a function of the microwave frequency and flux bias. The
observed phase modulation can be modelled by angle =
), with Lj(φ, Ic) =
arctan(
((1−ω2Lj (φ,Ic)C)2)2−Z2
1−ω2Lj (φ,Ic)C
2Z0
ωLj (φ,Ic)
0
TAS-2019-0211
3
φ0
2πIc cos(πφ/φ0), where ω is the frequency of the probe tone,
φ0 is the flux quantum (about 2.07 × 10−15 Wb), Z0 ≈ 15
Ω is the transformed characteristic impedance of the trans-
mission line, Ic is the zero-field critical current, φ is the flux
penetrating the SQUID loop, C is the shunt capacitance[25].
The zero-field critical current Ic and shunt capacitance can be
obtained by fitting theory to the observed phase modulation
curves of the JPA. The extracted Ic is about 4.5 µA. The
extracted C is about 3.5 pF. These values are close to the
designed values.
Fig. 2. Flux modulation of the phase of the reflected microwave (phase of the
reflected microwave as a function of the signal frequency and the flux bias,
measured by a vector network analyser). The boundary between the bright
and dark regions indicates the resonance frequencies.
IV. TWO DIMENSIONAL GAIN PROFILE
After studying the phase modulation of the JPA devices
caused by the external flux bias, we turned on the flux pump
and characterized the 2d gain profile (gain as a function of the
signal frequency ωs and pump power at the sample, at various
pump frequencies and flux biases) in detail. An example of
the measured 2d gain profile can be found in Fig. 3(b).
This 2d gain profile shows a phase diagram of the JPA, if we
look at the phase information of the microwave. Take Fig. 3(a)
as an example, at a low pump power, the boundary indicated
by the black curve represents the pump-power dependence of
the resonance frequency of the JPA. We can see the resonance
frequency shifts to lower values as the pump power increases.
This phenomenon can be explained by the non-linearity of the
Duffing oscillator[26], [27]. Above a threshold pump power,
the phase changes suddenly. This phenomenon could also be
explained by the physics of the non-linear Duffing oscillator.
Above some critical value of the pump power, the Duffing
oscillator enters a bifurcation zone. The resonance amplitude
will become multi-valued in this zone[26], [27], leading to the
instability of the system. The bifurcation zone also manifests
itself as a sudden change of the gain in the corresponding
amplitude or gain plot (Fig. 3(b)). Comparing the amplitude
or gain plot (Fig. 3(b)) with the phase plot (Fig. 3(a)), we can
see some gain hot zones in the different regions defined by the
phase response. Further experiments show the gain hot zone
lies in the region between the black curve and the bifurcation
zone is stable (useful for readout purpose of qubit states). The
stable gain hot zone usually occurs in the region not far from
the onset of the bifurcation zone. This is because the cavity
photon number changes abruptly at
the bifurcation point,
thus having the maximum parametric conversion effect[26],
[27]. Therefore, the region between the black curve and the
bifurcation zone is where we should focus to search for the
optimal stable gain hot zone.
Let us focus on the stable gain in the stable region. We can
see the gain hot zone generally tracks the curve representing
the pump-power dependence of the resonance frequency. Thus
the gain hot zones for a higher pump frequency are located at
a smaller pump power, because the resonance frequency in-
creases at a lower pump power. The gain hot zone usually has
two branches. Sometimes they merge into a broad plateau. A
theoretical model[23] for the 2d gain profile of an impedance-
matched current-pumped JPA can be adapted to describe the
observed 2d gain profile of our JPA. The gain profile of
an impedance-matched current-pumped JPA can be calculated
based on the differential equation of the circuit system or the
Hamiltonian of the circuit system. The differential equation
for the current-pump mode reads[23]:
d2δs(t)
dt2 +κ0
dδs(t)
dt
+Ω2
p(1−)(1−
1 −
sin(2Ωdt−2θ))δs(t) = As(t)
where δs is the phase parameter on the Josephson junction,
Ωp is the resonant frequency, Ωd is the drive frequency in
the current-pump mode, As is the signal amplitude, is
controlled by the drive amplitude. More details can be found
in reference[23].
Comparing it to the differential equation of the flux-pump
mode:
dδs(t)
+ Ω2
b(1 − η cos(Ωpumpt))δs(t) = As(t)
d2δs(t)
dt2 + κ0
dt
b = 2Ic×2π
Cφ0
cos( π
φ0
Φb),η = tan( π
φ0
where Ω2
∆Φ, Φb is
the flux bias, ∆Φ is the amplitude of the flux pump, C is
the shunt capacitance and Ic is the critical current. We can
see these two equations are basically identical, with some
replacements: Ωpump = 2Ωd, Ω2
Φb)
and =
p(1 − ) = 2Ic×2π
cos( π
φ0
Φb) π
φ0
Cφ0
1+
tan( π
φ0
1
1
Φb) π
φ0
.
∆Φ
Therefore, the differential equation for the flux-pump mode
is similar to that of the current-pump mode. We just need
to do some substitution, replacing some terms in the para-
metric oscillating term in the current-pump mode with their
counterparts in the flux-pump mode. In other words, there
is some one-to-one correspondence between the terms of the
differential equation in the current-pump mode and the terms
of the differential equation in the flux-pump mode. Therefore,
we can directly apply the gain formula of the current-pump
mode to the flux-pump mode by simply converting the pump
strength in the current-pump mode to the flux pump power in
the flux-pump mode.
TAS-2019-0211
4
The resulted gain is g = 1 − κ1χ112. κ1 is related to
the real part of the environmental admittance and χ11 is an
element of the susceptibility matrix. Details can be found in
reference [23].
With the critical current and shunt capacitance extracted
from the phase modulation curves and an appropriate envi-
ronmental impedance as the input, the theoretical model gives
a 2d gain profile which qualitatively matches with the data.
The calculated gain profile can be found in Fig. 4. The model
captures the features of the data, such as the emergence of
a gain hot zone with two branches around the resonance
frequency of the JPA[23]. The branching behavior in the gain
hot zone depends on the environmental impedance. For a non-
optimal environmental impedance, the gain profile has two
branches, related to the normal modes of the JPA and the
environment[23]. For the optimal environmental impedance,
the two branches merge, giving rise to a broad plateau. The
gain hot zone usually gets disturbed by the emergence of the
bifurcation zone and having a distorted shape extended into
the bifurcation zone.
Fig. 4. Calculated 2d gain profile for the pump frequency around 12.6 GHz
and flux bias at 0.3φ0. The calculated gain as a function of pump strength
and signal frequency qualitatively matches with the observed gain profile in
the stable zone. Since the theory works in the stable zone, this calculation
doesn't capture the abrupt change in the bifurcation zone.
Fig. 3.
(a) An example of the phase response (phase as a function of signal
frequency and pump power at the RF input to the JPA) for pump frequency
around 12.6 GHz and flux bias around 0.3φ0. The black curve indicates
the decrease of the resonance frequency as a function of pump power. The
green dashed line indicates the beginning of the bifurcation zone. (b) The
corresponding amplitude response of (a), the entering of the bifurcation zone
also manifests itself here, the abrupt change when pump power is around -34
dBm indicates the beginning of the bifurcation zone.
Based on the 2d gain profile, we propose that the best
working zone is near the merging point of the two branches
of the gain hot zone before the emergence of the bifurcation
zone, which gives a large bandwidth and good gain. In our
JPA, larger than 17 dB gain with a bandwidth larger than 300
MHz (centred around 6.45 GHz) was observed, see Fig. 5 (the
pump frequency is around 12.9 GHz. The flux bias is around
0.3φ0. The pump power is around -40 dBm at the RF input
to the JPA). The large bandwidth with a decent gain is ideal
for the readout task of the multi-qubit quantum chips.
V. SATURATION POWER AND NOISE TEMPERATURE
Besides the gain profile, we also characterized the saturation
power and noise temperature of our JPA. The gain decreases at
higher signal power. The saturation power can be characterized
by the 1 dB compression point, the value of the signal power
where the gain decreases by 1 dB. The 1 dB compression
power in the middle of our best gain hot zones (Fig. 5) is
about -110 dBm. The noise temperature measures how much
Fig. 5. Bandwidth of the JPA. This figure shows the gain as a function of
the signal frequency. We can see that gain over 17 dB can be reached in a
frequency range of around 300 MHz. The pump frequency is around 12.9
GHz. The flux bias is around 0.3φ0. The pump power is around -40 dBm at
the RF input to the JPA.
noise is added during the amplification process. In our setup,
the noise temperature of our JPA in the region with gain larger
than 15 dB is about 300 mK, which is close to the quantum
limit. More details about the noise temperature of the type of
JPA made by us can be found in the literature[28].
VI. QUBIT READOUT
Our impedance-matched JPA is used in our experiments for
improving the fast readout fidelity of transmon qubits. Fig. 6
shows the improvement in the readout fidelity with the JPA.
Fig. 6(a) shows the fidelity without the JPA in action, Fig. 6(b)
shows the fidelity with the JPA turned on. We can see that our
JPA improved the readout fidelity a lot. In the experiments on
the superconducting qubit chips, up to 6 qubits can be read
simultaneously with the JPA[29]. The large bandwidth with
TAS-2019-0211
5
a decent gain and a decent saturation power is ideal for the
readout task of our multi-qubit experiments.
(a)
(b)
Fig. 6.
(a) Readout fidelity with JPA off. The fidelity for the ground state
and excited state are about 0.83 and 0.82, respectively. (b) Readout fidelity
with JPA on. The fidelity for the ground state and excited state are now about
0.99 and 0.97, respectively.
VII. CONCLUSIONS
In summary, we designed and fabricated an impedance-
matched Josephson parametric amplifier working in the flux-
pump mode. The simple fabrication process is suitable for a
small cleanroom without the etching facilities for dielectrics.
We studied the 2d gain profile in detail. We adapted a
theoretical model for the 2d gain profile of an impedance-
matched current-pumped JPA to describe the observed 2d
gain profile of our JPA. With an appropriate environmental
impedance, the theoretical model captures the features of the
gain profile, such as the emergence of a gain hot zone with
two branches around the resonance frequency of the JPA. Non-
linear behaviour was also observed, such as the shift of the
resonance frequency at a higher pump power (Duffing shift),
as well as the emergence of the bifurcation zones (unstable
regions with a sudden drop of the gain) at a higher pump
power. Based on the gain profile, we propose that the best
working zone is the branching point of the gain hot zone in
the stable zone right before the emergence of the bifurcation
zone, which gives a large bandwidth and good gain. Gain Over
17 dB with a bandwidth larger than 300 MHz (centred around
6.45 GHz) was observed. The impedance-matched JPA is used
in our experiments for improving the fast readout fidelity of
the transmon qubits.
ACKNOWLEDGMENT
This work was partially carried out at the USTC Center for
Micro and Nanoscale Research and Fabrication. This research
was supported by the National Basic Research Program (973)
of China (Grant No. 2017YFA0304300), the Chinese Academy
of Sciences, Anhui Initiative in Quantum Information Tech-
nologies, Technology Committee of Shanghai Municipality,
NSFC (Grants No. 11574380, No. 11774406, No. U1530401).
National Key Research and Development Program of China
(Grant No. 2016YFA0302104, No. 2016YFA0300600), Strate-
gic Priority Research Program of Chinese Academy of Sci-
ences (Grant No. XDB28000000), China Postdoctoral Science
Foundation (Grant No. 2018M640055), and Beijing Science
Foundation (Grant No. Y18G07). The authors also thank
QuantumCTek Co., Ltd. for supporting the fabrication and the
maintenance of room temperature electronics.
REFERENCES
[1] R. Feynman, "Simulating physics with computers," International Jour-
nal of Theoretical Physics, vol. 21, no. 467, 1982.
[2] P. W. Shor, "Algorithmic number theory," in First Inter- national
Symposium ANTS-I, 1994, p. 289.
[3] L. K. Grover, "Quantum mechanics helps in searching for a needle in
a haystack," Physics Review Letter, vol. 79, no. 325, 1997.
[4] L. Grover, in Proc. 28th Annual ACM Symposium on the Theory of
Computing (STOC), 1996, p. 212.
[5] R. Barends, J. Kelly, A. Megrant, D. Sank, E. Jeffrey, Y. Chen, Y. Yin,
B. Chiaro, J. Mutus, and C. Neill, "Superconducting quantum circuits
at the surface code threshold for fault tolerance," Nature, vol. 508, no.
500, 2014.
[6] e. a. L. M. K. Vandersypen, "Experimental realization of shors quantum
factoring algorithm using nuclear magnetic resonance," Nature, vol. 414,
no. 883, 2001.
[7] e. a. E. Lucero, "Computing prime factors with a josephson phase qubit
quantum processor," Nature Physics, vol. 8, no. 719, 2012.
[8] J. Jones, M. Mosca, and R. Hansen, "Implementation of a quantum
search algorithm on a quantum computer," Nature, vol. 393, no. 334,
1998.
[9] I. L. Chuang, N. Gershenfeld, and M. Kubinec, "Experimental imple-
mentation of fast quantum searching," Phys. Rev. Lett., vol. 80, no. 3408,
1998.
[10] V. Havlek, A. D. Crcoles, K. Temme, A. W. Harrow, A. Kandala,
J. M. Chow, and J. M. Gambetta, "Supervised learning with quantum-
enhanced feature spaces," Nature, vol. 567, no. 209, 2019.
[11] R. Ma, B. Saxberg, C. Owens, N. Leung, Y. Lu, J. Simon, and
insulator of photons,"
D. Schuster, "A dissipatively stabilized mott
Nature, vol. 566, no. 51, 2019.
[12] D. Riste, M. P. D. Silva, C. A. Ryan, A. W. Cross, A. D. Corcoles, J. A.
Smolin, J. M. Gambetta, J. M. Chow, and B. R. Johnson, "Demonstration
of quantum advantage in machine learning," npj Quantum Information,
vol. 3, no. 16, 2017.
[13] N. Ofek, R. H. A. Petrenko, P. Reinhold, Z. Leghtas, B. Vlastakis, Y. Liu,
L. Frunzio, S. Girvin, and L. Jiang, "Extending the lifetime of a quantum
bit with error correction in superconducting circuits," Nature, vol. 536,
no. 441, 2016.
[14] R. Barends, J. Kelly, A. Megrant, A. Veitia, D. Sank, E. Jef-
frey, T. C. White, J. Mutus, A. G. Fowler, B. Campbell, Y. Chen,
Z. Chen, B. Chiaro, A. Dunsworth, C. Neill, P. OMalley, P. Roushan,
A. Vainsencher, J. Wenner, A. N. Korotkov, A. N. Cleland, and J. M.
Martinis, "Superconducting quantum circuits at the surface code thresh-
old for fault tolerance," Nature, vol. 508, no. 500, 2014.
TAS-2019-0211
6
[15] D. Riste, M. Dukalski, C. Watson, G. D. Lange, M. Tiggelman, Y. M.
Blanter, K. W. Lehnert, R. Schouten, , and L. DiCarlo, "Deterministic
entanglement of superconducting qubits by parity measurement and
feedback," Nature, vol. 502, no. 350, 2013.
[16] R. Vijay, C. Macklin, D. Slichter, S. Weber, K. Murch, R. Naik, A. N.
Korotkov, , and I. Siddiqi, "Stabilizing rabi oscillations in a supercon-
ducting qubit using quantum feedback," Nature, vol. 490, no. 77, 2012.
[17] L. Ranzani, M. Bal, K. C. Fong, G. Ribeill, X. Wu, J. Long, H.-S. Ku,
R. P. Erickson, D. Pappas, and T. A. Ohki, "Kinetic inductance traveling-
wave amplifiers for multiplexed qubit readout," Applied Physics Letter,
vol. 113, no. 242602, 2018.
[18] D. T. Sank, "Fast, accurate state measurement
in superconducting
qubits," Ph.D. dissertation, University of California in Santa Barbara,
California, 2014.
[19] J. Y. Mutus, T. C. White, E. Jeffrey, D. Sank, R. Barends, J. Bochmann,
Y. Chen, Z. Chen, B. Chiaro, A. Dunsworth, J. Kelly, A. Megrant,
C. Neill, P. J. J. OMalley, P. Roushan, A. Vainsencher, J. Wenner,
I. Siddiqi, R. Vijay, A. N. Cleland, and J. M. Martinis, "Design and
characterization of a lumped element single-ended superconducting
microwave parametric amplifier with on-chip flux bias line," Applied
Physics Letter, vol. 103, no. 122602, 2013.
[20] C. M. Caves, "Quantum limits on noise in linear amplifiers," Physical
Review D, vol. 26, no. 1817, 1982.
[21] J. Y. Mutus, T. C. White, R. Barends, Y. Chen, Z. Chen, B. Chiaro,
A. Dunsworth, E. Jeffrey, J. Kelly, A. Megrant, C. Neill, P. J. J. OMalley,
P. Roushan, D. Sank, A. Vainsencher, J. Wenner, K. M. Sundqvist,
A. N. Cleland, and J. M. Martinis, "Strong environmental coupling in a
josephson parametric amplifier," Applied Physics Letters, vol. 104, no.
263513, 2014.
[22] T. Yamamoto, K. Inomata, M. Watanabe, K. Matsuba, T. Miyazaki,
W. D. Oliver, Y. Nakamura, and J. S. Tsai, "Flux-driven josephson
parametric amplifier," Applied Physics Letters, vol. 93, no. 042510,
2008.
[23] T. Roy, S. Kundu, M. Chand, A. M. Vadiraj, A. Ranadive, N. Nehra,
J. A. M.P. Patankar, A. A. Clerk, and R.Vijay, "Broadband parametric
amplification with impedance engineering: Beyond the gain-bandwidth
product," Applied Physics Letters, vol. 107, no. 262601, 2015.
[24] G. Dolan, "Supervised learning with quantum-enhanced feature spaces,"
Applied Physics Letter, vol. 31, no. 337, 1977.
[25] B. Foxen, J. Mutus, E. Lucero, E. Jeffrey, D. Sank, R. Barends, K. Arya,
B. Burkett, Y. Chen, Z. Chen, B. Chiaro, A. Dunsworth, A. Fowler,
C. Gidney, M. Giustina, R. Graff, T. Huang, J. Kelly, P. Klimov,
A. Megrant, O. Naaman, M. Neeley, C. Neill, C. Quintana, P. Roushan,
A. Vainsencher, J. Wenne, T. White, and J. M. Martinis, "High speed
flux sampling for tunable superconducting qubits with an embedded
cryogenic transducer," Superconductor Science and Technology, vol. 32,
p. 1, 2018.
[26] R. Vijay, "Josephson bifurcation amplifier: Amplifying quantum signals
using a dynamical bifurcation," Ph.D. dissertation, Yale University, New
Haven, Connecticut, May 2008.
[27] M. A. Castellanos-Beltran, "Development of a josephson parametric
amplifier for the preparation and detection of nonclassical states of
microwave fields," Ph.D. dissertation, University of Colorado, Colorado,
2010.
[28] K. Q. Huang, Q. J. Guo, C. Song, Y. R. Zheng, H. Deng, Y. L. Wu, Y. R.
Jin, X. B. Zhu, and D. N. Zheng, "Fabrication and characterization of
ultra-low noise narrow and wide band josephson parametric amplifiers,"
Chinese Physics B, vol. 26, no. 094203, 2017.
[29] Y. S. Ye, Z. Y. Ge, Y. L. Wu, S. Y. Wang, M. Gong, Y. R. Zhang, Q. L.
Zhu, R. Yang, S. W. Li, F. T. Liang, J. L., Y. Xu, C. Guo, L. H. Sun,
C. Cheng, N. S. Ma, Z. Y. Meng, H. Deng, H. Rong, C. Y. Lu, C. Z.
Peng, H. Fan, X. B. Zhu, and J. W. Pan, "Propagation and localization
of collective excitations on a 24-qubit superconducting processor," Phys.
Rev. Lett., vol. 123, p. 050502, 2019.
|
1905.06039 | 1 | 1905 | 2019-05-15T09:07:15 | A pseudo-capacitive chalcogenide-based electrode with dense 1-dimensional nanoarrays for enhanced energy density in asymmetric supercapacitors | [
"physics.app-ph",
"cond-mat.mtrl-sci"
] | To achieve the further development of supercapacitors (SCs), which have intensively received attention as a next-generation energy storage system, the rational design of active electrode materials with electrochemically more favorable structure is one of the most important factors to improve the SC performance with high specific energy and power density. We propose and successfully grow copper sulfide (CuS) nanowires (NWs) as a chalcogenide-based electrode material directly on a Cu mesh current collector using the combination of a facile liquid-solid chemical oxidation process and an anion exchange reaction. We found that the as-prepared CuS NWs have well-arrayed structures with nanosized crystal grains, a high aspect ratio and density, as well as a good mechanical and electrical contact to the Cu mesh. The obtained CuS NW based electrodes, with additional binder- and conductive material-free, exhibit a much higher areal capacitance of 378.0 mF/cm2 and excellent cyclability of an approximately 90.2 percentage retention during 2000 charge/discharge cycles due to their unique structural, electrical, and electrochemical properties. Furthermore, for practical SC applications, an asymmetric supercapacitor is fabricated using active carbon as an anode and CuS NWs as a cathode, and exhibits the good capacitance retention of 91% during 2000 charge/discharge processes and the excellent volumetric energy density of 1.11 mW h/cm3 compared to other reported pseudo-capacitive SCs. | physics.app-ph | physics | A pseudo-capacitive chalcogenide-based electrode
with dense 1-dimensional nanoarrays for enhanced energy density
in asymmetric supercapacitors
Young-Woo Leea,1, Byung-Sung Kima,1, Jong Honga, Juwon Leea, Sangyeon Paka,
Hyeon-Sik Jangb, Dongmok Whangb, SeungNam Chaa,*, Jung Inn Sohna,* and Jong Min Kima,c
a Department of Engineering Science, University of Oxford, Oxford OX1 3PJ, United Kingdom.
b School of Advanced Materials Science and Engineering, SKKU Advanced Institute of
Nanotechnology (SAINT), Sungkyunkwan University, Gyeonggi-Do 16419, Republic of Korea.
c Electrical Engineering Division, Department of Engineering, University of Cambridge, Cambridge
CB3 0FA, United Kingdom.
1 These authors contributed equally to this work.
* Corresponding author. . Tel: +44-1865-273912. Fax: +44-1865-273010.
E-mail address: [email protected]; [email protected].
ABSTRACT
To achieve the further development of supercapacitors (SCs), which have intensively
received attention as a next-generation energy storage system, the rational design of active electrode
materials with electrochemically more favorable structure is one of the most important factors to
improve the SC performance with high specific energy and power density. We propose and
successfully grow copper sulfide (CuS) nanowires (NWs) as a chalcogenide-based electrode material
directly on a Cu mesh current collector using the combination of a facile liquid-solid chemical
oxidation process and an anion exchange reaction. We found that the as-prepared CuS NWs have
well-arrayed structures with nanosized crystal grains, a high aspect ratio and density, as well as a
good mechanical and electrical contact to the Cu mesh. The obtained CuS NW based electrodes, with
additional binder- and conductive material-free, exhibit a much higher areal capacitance of 378.0 mF
cm−2 and excellent cyclability of an approximately 90.2% retention during 2000 charge/discharge
cycles due to their unique structural, electrical, and electrochemical properties. Furthermore, for
practical SC applications, an asymmetric supercapacitor is fabricated using active carbon as an anode
and CuS NWs as a cathode, and exhibits the good capacitance retention of 91% during 2000
charge/discharge processes and the excellent volumetric energy density of 1.11 mW h cm-3 compared
to other reported pseudo-capacitive SCs.
Keywords: Copper chalcogenide, Nanowire, Nano-sized grain, Pseudocapacitance, Asymmetric
supercapacitor
INTRODUCTION
Supercapacitors (SCs), such as pseudocapacitor and electric double-layer capacitors (EDLCs)
that can bridge the potential gap between conventional rechargeable batteries and transitional
electrostatic capacitors, have been considered as a promising building block for next generation
energy storage systems that require the high specific energy and power density for wide-ranging
energy/power supplies, electric vehicles, and portable devices.1-3 Pseudocapacitors, eminently, can
achieve a high specific capacitance compared to EDLCs because they are basically operated by
electrochemical Faradaic redox reactions, leading to much greater charge storage through the
chemical intercalation/deintercalation process of anion/cation ions into electrodes.4,5 Accordingly,
high capacitive electrodes for pseudocapacitors, coupled with both high energy and power densities,
have been desired and investigated by focusing primarily on transition metal based oxide/hydroxide
materials (i.e., MOx, M(OH)x, and M1M2Ox; M=Co, Ni, Cu, W, Mo, etc.)6-10 owing to their richer
redox chemical valences as well as on their architecture engineering11-13 to enhance high surface
area/electrolyte contact areas and to induce fast charge transfer rates.
Copper oxide/hydroxide-based materials are one of the attractive pseudo-electrodes to
enhance a charge storage ability in the pseudocapacitors due to their two-electron redox reactions.14,15
In addition, in order to reach a high areal capacitance and good cyclability of pseudo-electrodes,
several strategies have been developed for the production of architecture-controlled, nano-sized, and
complex materials to induce more intercalation active sites, effortless charge transfer pathway, and
improved electrical conductivity.16-18 Despite many efforts for the high electrochemical performance
of the oxide/hydroxide-based pseudo-electrodes, there still remain the challenges, i.e., low areal
capacitance, poor rate performance, unacceptable energy density, and unstable cyclability because of
their low intrinsic electrical conductivity and low structural stability during pseudo-charge/discharge
processes.19,20
In this regard, copper chalcogenide based materials have been recently spotlighted, due to
their high intrinsic electrical conductivity and fast electron/charge transfer rates as well as two
electron redox reactions. Recently, Lei et al. demonstrated that copper chalcogenide/polypyrrole
prepared by an in-situ oxidative polymerization approach exhibits a superior specific capacity of
171.2 mF cm−2 and excellent cycling stability.21 Also, Huang et al. reported that the copper
chalcogenide nanosheets synthesized using an one-step solvothermal process show the enhanced
areal capacitance.22 However, fundamental researches to enhance pseudo-capacitive performances
with copper chalcogenide based materials are still in the beginning stage and need much attention for
achieving the high areal capacitance and energy density in their practical SC applications.
Herein, we present a facile synthesis approach to develop the dense arrays of copper sulphide
(CuS) nanowires (NWs) with nanosized crystal grains with high grain boundaries (GBs), as attractive
copper chalcogenides materials for SCs, directly on the Cu mesh current collector and investigate the
interrelations between their structural and electrochemical properties in pseudocapacitors. We also
propose that superior electrochemical pseudo-capacitive behavior of the well-defined CuS NWs
based electrodes could be attributed to the synergistic effects that might be explained with the
following major viewpoints: 1) the directly grown CuS NWs on the Cu mesh, without the need of
conducting additives and binder materials, can provide stable adhesion, low contact resistance, and
the fast charge transfer channels, thereby can induce the high areal capacitance and volumetric energy
density as well as acceptable stability; 2) the CuS nanostructures consisted by nanosized grains with
high GBs can induce large contact area for electrolytes, thus facilitating the opened diffusion path
favorably for OH− ions and providing the fast charge transfer rates; 3) the CuS phase with the
excellent electrical conductivity and good crystallinity can induce superior electrochemical
cyclability and high structural stability. In addition, we design the asymmetric supercapacitor (ASC)
with active carbon (AC) as the anodic material and CuS NWs as the cathodic material to demonstrate
comparable volumetric energy density of 1.11 mAh cm−3 and good cyclability of 91% during 2000
charge/discharge processes to those of commercial SCs and other reported pseudo-capacitive SCs.
RESULTS AND DISCUSSION
To obtain highly uniform CuS NWs directly grown on a Cu mesh, we used a sequential two-
step synthetic scheme involving (1) the solution growth of Cu(OH)2 and (2) the sulfurization process
for CuS, which are schematically illustrated in Fig. 1a. First, Cu(OH)2 NWs were directly grown on
the Cu mesh using a liquid-solid chemical oxidation process at room temperature for 60 min. Here,
it should be noted that, the piece of an immersed Cu mesh not only provides Cu sources for the
formation of Cu(OH)2 NWs but also acts as the current collector, which allows to achieve improved
adhesion, structural stability and contact resistance at the interface between the electrode material and
the current collector. The obtained Cu(OH)2 NWs presented a high crystallinity (Fig. 1b) and all X-
ray diffraction (XRD) patterns were well indexed to the crystal phase of Cu(OH)2 (JCPDS No. 13-
0420). Subsequently, CuS NWs were then fabricated via the dip-coating of as-prepared Cu(OH)2
NWs surface with thiourea as sulfur sources, followed by the sulfurization process. As a result,
Cu(OH)2 phases were perfectly converted to CuS phases as shown by XRD patterns (JCPDS No. 06-
0464). Furthermore, to investigate the morphology of CuS NWs, we carried out a field emission
scanning electron microscope (FE-SEM) analysis. Figs. 1c and 1d clearly show that dense CuS NWs
were well arrayed and separated from each 1-D nanostructure with a high aspect ratio on the Cu mesh.
Moreover, there was no observation of any structural agglomeration and collapse resulting from the
sulfurization process involving sulfur anion exchange reactions, as exhibited in Fig. 1d.
To further identify the detailed crystal structure and elemental distribution of CuS NWs, we
performed transmission electron microscopy (TEM) examinations. As shown in Fig. 2a, the CuS NW
exhibited a well-defined 1-dimensional (D) nanostructure with the homogeneous elemental
distributions of Cu and S atoms. In particular, the high-resolution TEM image reveals that the CuS
NW is composed of polycrystalline small grains with the size of approximately 5~6 nm, providing
the high grain boundaries, and the interplanar spacing of single crystal grains is ~ 0.28 nm
corresponding to {103} facets in the hexagonal crystal structure of CuS, in Fig. 2b. This finding
suggests that the well-arrayed 1-D nanostructures with nanosized grains and high GBs can provide
the fast charge transfer channels and can also facilitate the opened diffusion path of OH− ions for
superior electrochemical reactions. In addition to the crystal and geometrical structure, since the
chemical state of active materials would be an important key factor playing a crucial role in affecting
electrochemically pseudo capacitive behavior, the X-ray photoelectron spectroscopy (XPS) analysis
was performed to determine the surface chemical state in CuS NWs. We clearly observed Cu 2p3/2
peaks located at 932.8 and 933.8 eV being assigned to the metallic Cu and Cu-S state, respectively,
revealing the presence of Cu2+ states, as show in Fig. 2d. Moreover, we found that the peak of S 2p3/2
with the binding energy at 162.0 eV corresponds to the Cu-S bond, further supporting the evidence
for the Cu2+ state (Fig. 2e). A comparison of structural and chemical analyses results implies that
directly grown CuS NWs on the Cu mesh can exhibit the enhanced pseudo-capacitive behavior during
the charge/discharge process for the excellent electrochemical performance due to their highly dense
array of the 1-D nanostructures with nanosized grains and high electrical conductivity as well as Cu2+
states, allowing for superior electrochemical reactions by providing favorable pseudo-active sites.
The pseudo-behavior properties of CuS NWs were investigated through cycling voltammetry
(CV) and galvanostatic charge/discharge analyses using a three-electrode system with CuS NWs on
the Cu mesh directly used as a working electrode. Fig. 3a presents CV curves obtained from CuS
NWs in different upper potential windows from 0.2 to 0.5 V at a scan rate of 5 mV s−1. It is shown
that the shape of all CV curves of CuS NWs differs distinctly from the typical rectangular shape of
double layer capacitances, indicating that the electrochemical capacitive behavior is mainly governed
by the redox reactions. Explicitly, when the upper potential limit was increased from 0.2 to 0.5 V, we
observed a pair of the redox peaks, originating from the reversible Faradic redox reaction.14,15,23,24
Additionally, good symmetric curves indicate the excellent redox-reversibility of CuS NWs. Fig. 3b
shows comparative CV results of CuS NWs and Cu(OH)2 NWs. The enclosed area of CuS NWs is
approximately 2.3 times bigger than that of the Cu(OH)2 NWs, indicating that the CuS NWs exhibit
the largely enhanced pseudo-capacitance after the sulfurization of as-grown Cu(OH)2 NWs. To
further demonstrate the superior electrochemical capacitive performance of CuS NWs based
electrodes, the areal capacitance was calculated from galvanostatic charge/discharge curves measured
in a potential window of 0 to 0.5 V at different current densities ranging from 2 to 20 mA cm−2. At a
discharging current density of 2 mA cm−2 (Fig. 3c), the areal capacitance of CuS NWs (378.0 mF
cm−2) is 2.2 times higher than that of Cu(OH)2 NWs (172.4 mF cm−2), which is in good agreement
with CV results. Furthermore, even at a high charge/discharge rate of 20 mA cm−2, the CuS NWs
exhibited a superior capacitance retention of 71.2 % as shown in Fig. 3d. Here, note that to the best
of our knowledge, these areal capacity and retention performance are excellent values in comparison
with those of Cu(OH)2 NWs and other reported Cu-based pseudo-capacitive electrodes,21,22,24 as
summarized in Fig. 3d. These results are attributed to the structural synergistic effects, as
schematically illustrated in Fig. 3e: (1) the direct growth of active materials from core metal sources
can minimize interfacial resistant and facilitate charge transfer at the interface between the Cu current
collector and CuS NWs; (2) the dense CuS NWs with well-ordered 1-D nanostructures and the
relatively high conductive CuS phase structure can induce the fast charge/electron transfer rate via
the efficient electron channel; (3) the nanosized crystal grains with high GBs distributed along NWs
can provide the large contact area of electrolyte and the opened diffusion path of OH− ions.
The cyclic performance is also an important factor to evaluate the ability of a SC. Fig. 3f
shows that the CuS NWs based electrode exhibited the superior cyclic performance up to 2000
charge/discharge cycles at a current density of 10 mA cm−2. The reversible capacitance retention (90.2%
of the initial maximum capacitance) of CuS NWs is substantially much better than that of previously
reported Cu-based composites, such as the CuS@PPy composite (88% retention up to 1000 cycles at
1 A g−1),21 the CuS nanosheets (75.4% retention up to 500 cycles at 1 A g −1),22 the CuS embedded
within porous octahedral carbon (approximately 80% retention up to 2000 cycles at 5 mV s−1),25 and
the CuO nanoflowers (84% retention up to 2000 cycles at 100 mV s−1).26 To clearly understand the
electrochemical and structural cycling stability of our CuS NWs, electrochemical impedance
spectroscopy (EIS) was used and Nyquist plots were obtained in the frequency range of 10 mHz to
100 kHz before and after 2000 cycling tests in Fig. 3g. Before the cycling test, the charge transfer
resistance (Rct) of CuS NWs was determined to be 0.11 Ω, which may be attributed to the low contact
resistance and fast charge transfer rate resulting from directly grown CuS NWs on the Cu mesh and
the well-arrayed 1-D nanostructures with nanosized grains. Furthermore, after a 2000 cycling test,
the CuS exhibited a slight increase in the Rct, but still maintained a very low value. The structural
stability of CuS NWs was further confirmed by Raman analyses before and after the cycling test, as
shown in Fig. 3h. Initially, as expected, the two characteristic Lorentzian peaks were observed around
~260.4 and ~468.8 cm−1 corresponding to two different vibrational (stretching) modes related to the
covalent S-S bond and Cu-S bond, respectively. After a cycling test, noticeably the Raman spectrum
of CuS NWs exhibited that the two dominant peaks related to a CuS phase remain unchanged,
confirming that the CuS NWs have a superior structural stability during the charge/discharge process.
However, additional peaks with the relatively very weak intensity were also observed near ~ 288.0,
335.2, and 617.7 cm−1 related to a CuO phase, which might be partially formed via the intercalation
process of OH− ions during the galvanostatic charge/discharge process. Thus, we believe that the
remarkably superior pseudo-behavior properties of the electrode, i.e., high areal capacitance and
excellent cycling performance, are mainly attributed to the unique features of the CuS NWs with
superior electrical conductivity, favorable structural architecture and phase, and electrochemical
stability.
To further investigate practical possibility for SC applications on the basis of previous
electrochemical results and discussion, we fabricated an ACS using the AC as an anodic material and
the CuS NWs as a cathodic material (inset in Fig. 4e) by considering the charge balance between two
electrodes to reach the ideal energy storage performance of an ACS.27 For the comparison and
estimation of operating potential ranges, CV curves of the AC and the CuS NWs were first
characterized (Fig. 4a). From the sum of the potential ranges of these two electrodes, the potential
window of the ACS was estimated to be 1.5 V. Fig. 4b presents CV curves of the AC//CuS ASC in
different upper potential windows from 0.6 to 1.5 V at a scan rate of 50 mV s−1, indicating that this
ACS system can be operated up to 1.5 V under stable electrochemical behavior. The charge/discharge
curves of the AC//CuS ASC are shown in Fig. 4c. At a current density of 2 mA cm−2, the volumetric
capacitance of the AC//CuS ASC reached 3.54 F cm−3, which is much higher capacitance in
comparison with other reported ASCs.28-30 Fig. 4d shows the Ragone plots of the AC//CuS ASC,
revealing the energy and power density calculated from the discharge curves. Notably, the AC//CuS
ASC exhibited the higher volumetric energy density of 1.11 mWh cm−3 and the enhanced volumetric
power density of 0.36 W cm−3 in comparison with those of commercial energy storage devices and
other reported pseudo-capacitive SCs.31-36 In addition, even after 2000 charge/discharge cycles at a
current density of 10 mA cm−2 (Fig. 4e), the AC//CuS ASC showed a good cyclability with the high
capacitance retention of 91.0 % and Coulombic efficiency of 97.7 %.
CONCLUSIONS
In summary, we demonstrated the CuS NW arrays directly grown on the Cu mesh
successfully prepared via a facile liquid-solid chemical oxidation process and an anion exchange
reaction for the enhanced pseudo-capacitive behavior properties in SCs. The obtained CuS NWs
exhibited a superior areal capacitance of 378.0 mF cm−2 at a current density of 2 mA cm−2 and good
rate performance as well as long-term electrochemical and structural stabilities during the
galvanostatic charge/discharge process, compared to other reported Cu-based SCs. The improved
electrochemical pseudo-capacitive behavior properties of the CuS NWs can be attributed to unique
electrical and structural properties of the CuS NWs grown directly on Cu mesh current collector,
providing synergistic benefits such as high electrical conductivity, good contact resistance, densely
well-arrayed 1-D nanostructures with nanosized grains and high GBs, and the stable and favorite
structure of a CuS phase. Furthermore, the AC//CuS ASC with a wide potential window of 1.5 V
exhibited the enhanced volumetric energy density of 1.11 mWh cm−3 , the good cyclability of 91%
up to 2000 cycles, and the high Coulombic efficiency of 97.7 %. Thus, it is expected that the CuS
NWs directly grown on a Cu mesh current collector will be extensively utilized as highly stable and
efficient pseudo-capacitive electrodes in SCs.
EXPERIMENTAL
Fabrication of 1-D Cu(OH)2 and CuS NWs
Copper hydroxide nanowires (Cu(OH)2 NWs) were synthesized directly on the Cu mesh by using an
aqueous solution of sodium hydroxide (NaOH, 97%) and ammonium persulfate ((NH4)2S2O8, 98%).
The solution was prepared by mixing 20 mL of a 10 M NaOH solution, 10 mL of a 1 M (NH4)2S2O8
solution, and 22.5 mL of DI water. A pre-cleaned piece of a Cu mesh (1 x 1 cm2 with thickness of
0.0267 cm, Alfa Aesar) was subsequently immersed into the mixed solution at room temperature.
After the synthesis of Cu(OH)2 NWs for 1h, the Cu mesh sample was taken out of the solution and
then rinsed with DI water, followed by dried on a hot plate at 60 °C for 15 min. A light blue color of
the mesh was observed, indicating that highly dense Cu(OH)2 NWs were synthesized. In order to
convert as-grown Cu(OH)2 NWs to CuS NWs, a 0.2 M thiourea (CH4N2S, 99.0%) solution was
dropped on the surface of Cu(OH)2 NWs and then blown dry with nitrogen gas. After heating on the
hot plate at 150 °C, the color of the sample was changed from a blue to dark brown color. It indicates
that CuS NWs were completely transformed from Cu(OH)2 NWs.
Electrochemical Characterization
The electrochemical properties of the as-prepared Cu(OH)2 and CuS NWs directly grown on the Cu
mesh were measured in a three-electrode system, consisting of as-prepared CuS NWs electrodes as a
working electrode, Pt wire as a counter electrode, and Ag/AgCl (in saturated 3 M KCl) as a reference
electrode, in order to analyze CV, galvanostatic charge/discharge, and EIS behavior using a
potentiostat (PGSTAT302N, Metrohm, Autolab). For the fabrication of anode electrodes in the
AC//CuS ASC, the slurry was prepared by mixing the active carbon as an active material,
poly(vinylidene difluoride) as a binder, Ketjen black as a conductive material, and then was coated
onto the compressed nickel foam as a current collector. All electrochemical results of the AC//CuS
ASC were obtained using a two electrode system under 1.0 M KOH solution at room temperature.
Acknowledgments
This research was supported by the Industrial Fundamental Technology Development Program
(10052745, Development of nano-sized (100nm) manganese ceramic material for high voltage
pseudo-capacitor) funded by the Ministry of Trade, Industry and Energy (MOTIE) of Korea, and the
European Research Council under the European Union's Seventh Framework Programme (FP/2007-
2013) / Grant Agreement no. 685758, Project '1D-NEON'.
References
1 M. Zhi, C. Xiang, J. Li, M. Li and N. Wu, Nanoscale, 2013, 5, 72-88.
2
3
4
5
6
7
8
9
G. Wang, L. Zhang and J. Zhang, Chem. Soc. Rev., 2012, 41, 797-828.
P. Simon, Y. Gogotsi and B. Dunn, Science, 2014, 343, 1210-1211.
R. A. Fisher, M. R. Watt and W. J. Ready, ECS J. Solid State Sci, Technol., 2013, 2, M3170-
3177.
V. Auqustyn, P. Simon and B. Dunn, Energy Environ, Sci., 2014, 7, 1597-1614
D. Chen, Q. Wang, R. Wang and G. Shen, J. Mater. Chem. A, 2015, 3, 10158-10173.
S. Lee, Y.-W. Lee, D.-H. Kwak, M.-C. Kim, J.-Y. Lee, D.-M. Kim and K.-W. Park, Ceram. Int.,
2015, 41, 4989-4995.
Y. Zhang, L. Li, H. Su, W. Huang and X. Dong, J. Mater. Chem. A, 2015, 3, 43-59.
K.-H. Lee, Y.-W. Lee, A-R. Ko, G. Cao and K.-W. Park, J. Am. Ceram. Soc., 2013, 96, 37-39.
10 Y. Zhu, X. Ji, Z. Wu, W. Song, H. Hou, Z. Wu, X. He, Q. Chen and C. E. Banks, J. Power
Sources, 2014, 267, 888-900.
11 G. Zhang and X. W. Lou, Sci, Rep., 2013, 3, 1470
12 W. Tang, L. Liu, S. Tian, L. Li, Y. Yue, Y. Wu and K. Zhu, Chem. Commun., 2011, 47, 10058-
10060.
13 M. Yu, W. Qiu, F. Wang, T. Zhai, P. Fang, X. Lu and Y. Tong, J. Mater. Chem. A, 2015, 3, 15792-
15823
14 Y.-K. Hsu, Y.-C. Chen and Y.-G. Lin, J. Electroanal. Chem., 2012, 673, 43-47.
15
J. Chen, J. Xu, S. Zhou, N. Zhao and C.-P. Wong, J. Mater. Chem. A, 2015, 3, 17385-17391.
16 P. Xu, K. Ye, M. Du, J. Liu, K. Cheng, J. Yin, G. Wang and D. Cao, RSC Adv., 2015, 5, 36656-
36664.
17 V. D. Patake, S. S. Joshi, C. D. Lokhande, O.-S. Joo, Mater. Chem. Phys., 2009, 114, 6-9.
18 Y. Lu, X. Liu, K. Qiu, J. Cheng, W. Wang, H. Yan, C. Tang, J.-K. Kim and Y. Luo, ACS Appl.
Mater. Interfaces, 2015, 7, 9682-9690.
19 A. Pendashteh, M. F. Mousavi and M. S. Rahmanifar, Electrochim. Acta, 2013, 88, 347-357.
20 A. Pramanik, S. Maiti and S. Mahanty, Dalton Trans., 2015, 44, 14604-14612.
21 H. Peng, G. Ma, K. Sun, J. Mu, H. Wang and Z. Lei, J. Mater. Chem. A, 2014, 2, 3303-3307.
22 X. Dong, K. Wang, C. Zhao, X. Qian, S. Chen, Z. Li, H. Liu and S. Dou, J. Alloy. Compd., 2014,
586, 745-753.
23 T. Zhu, B. Xia, L. Zhou and X. W. Lou, J. Mater. Chem., 2012, 22, 7851-7855.
24 Y.-K. Hsu, Y.-C. Chen and Y.-G. Lin, Electrochim. Acta, 2014, 139, 401-407.
25 R. Wu, D. P. Wang, V. Kumar, K. Zhou, A. W. K. Law, P. S. Lee, J. Lou and Z. Chen, Chem.
Commun., 2015, 51, 3109
26 S. K. Shinde, D. P. Dubal, G. S. Ghodake and V. J. Fulari, RSC Adv., 2015, 5, 4443-4447.
27 K. H. Lee, Y.-W. Lee, S. W. Lee, J. S. Ha, S.-S. Lee and J. G. Son, Sci. Rep., 2015, 5, 13696.
28 X. Lu, M. Yu, T. Zhai, G. Wang, S. Xie, T. Liu, C. Liang, Y. Tong and Y. Li, Nano Lett., 2013,
13, 2628-2633.
29 T. Zhai, F. Wang, M. Yu, S. Xie, C. Liang, C. Li, F. Xiao, R. Tang, Q. Wu, X. Lu and Y. Tong,
Nanoscale, 2013, 5, 6790-6796.
30 X. Lu, M. Yu, G. Wang, T. Zhai, S. Xie, Y. Ling, Y. Tong and Y. Li, Adv. Mater., 2013, 25, 267-
272.
31 D. Yu, K. Goh, H. Wang, L. Wei, W. Jiang, Q. Zhang, L. Dai and Y. Chen, Nat. Nanotechnol.,
2014, 9, 555-562.
32 W. Zilong, Z. Zhu, J. Qiu and S. Yang, J. Mater. Chem. C, 2014, 2, 1331-1336.
33 X. Xiao, T. Li, P. Yang, Y. Gao, H. Jin, W. Ni, W. Zhan, X. Zhang, Y. Cao, J. Zhong, L. Gong,
W.-C. Yen, W. Mai, J. Chen, K. Huo, Y.-L. Chueh, Z. L. Wang and J. Zhou, ACS Nano, 2012,
6, 9200-9206.
34 Z.-S. Wu, W. Ren, D.-W. Wang, F. Li, B. Liu and H.-M. Cheng, ACS Nano, 2010, 4, 5835-5842.
35 X. Lu, D. Zheng, T. Zhai, Z. Liu, Y. Huang, S. Xie and Y. Tong, Energy Environ. Sci., 2011, 4,
2915-2921.
36 W. Chen, R. B. Rakhi and H. N. Alshareef, J. Mater. Chem., 2012, 22, 14394-14402.
Figure
Fig. 1. (a) Schematic illustration of the synthesis process of CuS NWs. (b) XRD patterns of the as-
prepared Cu(OH)2 and CuS NWs. (c) A FE-SEM image of CuS NWs. (d) A high-magnification SEM
image of CuS NWs. The inset indicates the high-magnification SEM image of Cu(OH)2 NWs.
Fig. 2. (a) TEM (left) and corresponding EDX element mapping images (right) for the Cu (red) and
S (green) in the CuS NW. (b) High-resolution TEM images of the CuS NW. The inset indicates the
FFT pattern of a CuS NW. XPS (c) Cu 2p and (d) S 2p spectra of the CuS NWs.
Fig. 3. (a) CVs of CuS NWs in different upper potential windows from 0.2 to 0.5 V. (b) CVs of CuS
NWs and Cu(OH)2 NWs at a scan rate of 50 mV s−1. (c) Galvanostatic charge/discharge curves of
CuS NWs and Cu(OH)2 NWs at a current density of 2 mA cm−1. (d) Specific capacitances of CuS
NWs and Cu(OH)2 NWs at different constant current densities in comparison with other reported Cu-
based electrodes. (e) Schematic illustration related to pseudo-behavior properties of OH− ions in the
CuS NW. (f) Cycle performance of CuS NWs at a current density of 10 mA cm−2 up to 2000 cycling
charge/discharge tests. (g) Nyquist plots and (h) Raman spectra of the CuS NWs before and after
cycling tests. Insets in Fig. 3g indicate the expanded Nyquist plots and the equivalent circuit.
Fig. 4. (a) Comparative CVs of AC and CuS NWs at 50 mV s−1. (b) CVs of the AC//CuS ASC in
different upper potential windows from 0.6 to 1.5 V at a scan rate of 50 mV s−1. (c) Galvanostatic
charge/discharge curves of the AC//CuS ASC under various different current densities. (d) Ragone
plots of the AC//CuS ASC and recently reported other pseudo-capacitive SCs. (e) Cycle performance
of the AC//CuS ASC at a current density of 10 mA cm−2 up to 2000 cycling charge/discharge tests.
|
1807.10811 | 1 | 1807 | 2018-06-25T09:52:25 | Modelling of standard and specialty fibre-based systems using finite element methods | [
"physics.app-ph"
] | We report on the investigation of an approach for modelling light transmission through systems consisting of several jointed optical fibres, in which the analytical modelling of the waveguides was replaced by Finite Element Modelling (FEM) simulations. To validate this approach we first performed FEM analysis of standard fibres and used this to evaluate the coupling efficiency between two singlemode fibres under different conditions. The results of these simulations were successfully compared with those obtained using classical analytical approaches, by demonstrating a maximum loss deviation of about 0.4 %. Further, we performed other more complex simulations that we compared again to the analytical models. FEM simulations allow addressing any type of guiding structure, without limitations on the complexity of the geometrical waveguide cross section and involved materials. We propose as example of application the modelling of the light transmitted through a system made of a hollow core photonic crystal fibre spliced between two singlemode standard optical fibres, and qualitatively compare the results of the simulation with experimental results. | physics.app-ph | physics | Modelling of standard and specialty fibre-based systems using finite
element methods
Natascia Castagna*a, Jacques Morela, Luc Testab, Sven Burgerc,d
aFederal Institute of Metrology METAS, Lindenweg 50, 3003 Bern-Wabern, Switzerland; bEcole
Polytechnique Fédérale de Lausanne, EPFL PH D2 375 Station 3, 1015 Lausanne, Switzerland;
cJCMwave GmbH, Bolivarallee 22, 14050 Berlin, Germany; dZuse Institute Berlin, Takustrasse 7,
14195 Berlin, Germany
This is a draft manuscript of a paper published in Proc. SPIE (Proceedings Volume 10683, Fiber Lasers and Glass
Photonics: Materials through Applications, Event: SPIE Photonics Europe, 2018).
The full citation is: N. Castagna, J. Morel, L. Testa, S. Burger, Proc. SPIE 10683, 1068336 (2018)
The digital object identifier (DOI) is: 10.1117/12.2307372
One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of
any material in this publication for a fee or for commercial purposes, or modification of the contents of the publication
are prohibited.
*[email protected]; phone 0041 583870643; www.metas.ch
ABSTRACT
We report on the investigation of an approach for modelling light transmission through systems consisting of several
jointed optical fibres, in which the analytical modelling of the waveguides was replaced by Finite Element Modelling
(FEM) simulations. To validate this approach we first performed FEM analysis of standard fibres and used this to
evaluate the coupling efficiency between two singlemode fibres under different conditions. The results of these
simulations were successfully compared with those obtained using classical analytical approaches, by demonstrating a
maximum loss deviation of about 0.4 %. Further, we performed other more complex simulations that we compared again
to the analytical models. FEM simulations allow addressing any type of guiding structure, without limitations on the
complexity of the geometrical waveguide cross section and involved materials. We propose as example of application
the modelling of the light transmitted through a system made of a hollow core photonic crystal fibre spliced between two
singlemode standard optical fibres, and qualitatively compare the results of the simulation with experimental results.
Keywords: Finite element modelling, modal distribution, specialty fibres
1. INTRODUCTION
Important quantities like the insertion loss between jointed multimode fibres or the bandwidth depend on how the
different guided modes are populated and propagate all along the fibre. This strongly depends on how the light source is
coupled into the fibre, and on how the modal selection occurs at the interface between different sections of concatenated
fibres. A modelling of these effects is usually performed using analytical approaches to quantify the modal distribution
and then to calculate the overlap integrals. These techniques are very well applicable when considering simple structures,
but may become impractical when dealing with more complex waveguides, like micro-structured fibres. In this work we
investigate a different approach, in which the analytical modelling of the waveguides was replaced by FEM simulations,
allowing addressing any type of guiding structure, without limitation in its complexity. To validate this approach, we
performed different FEM analysis of standard fibres and the results of these simulations were successfully compared
with those obtained using classical analytical approaches. First, we evaluated the coupling efficiency between two
singlemode fibres under different conditions, namely the misalignment of the fibre optical axis with respect to the
direction of propagation of the light source; the results are presented in Section 2. Then, in light of the results obtained
with this first test, we performed a more complex simulation, for which we were inspired by an analytical work
published by Mafi et al. in 20111. In their work the authors make use of a multimode fibre as connector between two
singlemode fibres of different mode-field diameters to optimize the mode matching and thus to reduce the coupling
losses; we present in Section 3 the comparison between their analytical models and our FEM simulations. Finally in
Section 4 we propose, as example of application, the modelling of the light transmitted through a system made of a
hollow core photonic crystal fibre spliced between two singlemode standard optical fibres, and qualitatively compare the
results of the simulation with experimental results.
2. COUPLING EFFICIENCY BETWEEN TWO SINGLEMODE FIBRES
The aim of this simulation was to model the coupling efficiency between two singlemode fibres. We considered two
singlemode fibres, whose optical axis were misaligned, by applying either an angular, or a lateral offset. The two cases
are schematised in Figure 1.
Figure 1. Angular (a) and lateral (b) misalignments between two fibres.
The FEM software package JCMsuite2 was used to model the fibres, to calculate the guided eigenmodes as well as the
overlap integral between the two fibres. The parameters used in the simulation were as follows:
Core diameter φc / µm
Cladding diameter φcl / µm
Core refractive index nc
Cladding refractive index ncl
Wavelength λ/ nm
9
120
1.452
1.443
1550
2.1 Angular misalignment
We considered in this first example an angular misalignment between the two optical fibre axis, see Figure 1a). The
result of the simulation is shown in Figure 2, together with the result obtained with an analytical approach for the same
configuration and parameters. The analytical model was developed by RP Photonics3. The vertical axis shows the
normalised coupling efficiency.
Figure 2. Coupling efficiency between two identical singlemode fibres with an angular misalignment between the two
optical axes. In blue the result of the FEM modelling and in red the result of an analytical approach.
The agreement between the two approaches is high, with a maximum deviation of about 0.4 %.
2.2 Lateral misalignment
As a second example, the influence of a lateral misalignment of the two optical axes along the x-axis to the coupling
efficiency was investigated (see Figure 1b). Once again the FEM simulations were compared with the results of an
analytical approach and are shown in In Figure 3.
Figure 3. Coupling efficiency between two identical singlemode fibres with a lateral misalignment along the x-axis. In blue,
the results of our FEM modelling and in red the results of an analytical approach.
The agreement between the two evaluation methods is better than 0.4%.
3. LOW-LOSS COUPLING BETWEEN FIBRES: ANALYTICAL VS FEM ANALYSIS
In light of these results, we investigated the performances of a more complex system, inspired by the analytical work
published by Mafi et al.1. The aim of that work was to demonstrate that the coupling losses between two singlemode
fibres with different mode-field diameters may be reduced by inserting between them a piece of gradient-index
multimode fibre having a well adapted length (see Figure 4). They developed for that purpose an analytical model, which
allowed determining the coupling efficiency and its dependency with the length of the multimode fibre.
Figure 4. Scheme of the optical system used to reduce the coupling losses between two singlemode fibres (Fibre 1 and Fibre
3) with dissimilar mode-field diameters. Fibre 2 is a multimode fibre, which acts as a mode filed diameter adapter. d1, d2 and
d3 are the respective core diameters of the three fibres.
The coupling efficiency oscillates, according to the cited paper, with a period length Z given by:
(1)
(2)
Z
=
⋅
π
2/2d
2
∆⋅
,
where ∆ is the index step of the gradient index fibre defined by:
2
cl
n
2
.
2
n −
c
n
c
=∆
Considering the same parameters as those used by Mafi et al.1, namely d1 = d3 = 8.2 µm, d2 = 50 µm, and taking nc=1.452
and ncl=1.443 for the index of refractions of the core and of the cladding of fibres 1 and 2, a period length of Z = 0.7 mm
was calculated using Eq. (1). This same structure was modelled using the FEM software package JCMsuite2 to calculate
the eigenmodes of the multimode fibre and the transmission through the three fibres with the overlap integrals. Figure 5
shows the resulting transmission through the optical system as a function of the length of the multimode fibre. The
periodical dependence of the transmission to the fibre length is clearly visible, whit a period length of
ZFEM = 0.67 mm, which is in a very good agreement with the analytical solution.
Figure 5. Transmission through the three fibres system shown in Figure 4 as a function of the length of the multimode fibre.
4. APPLICATION TO SPECIALTY FIBRES
We propose as example of application the modelling of light transmission through a setup including a photonic crystal
fibre (PCF) which is spliced to a standard single mode fibre (SMF) on both sides, in the same setup as depicted in Fig. 4.
We have performed experimental transmission measurements of this setup with the PCF filled with acetylene. Figure 6A
shows a microscopic image of the PCF cross section before splicing, with a core diameter of about 7.3 µm. The SMF
core diameter is 8.2 µm. Figure 6B shows the measured transmission as function of the frequency of the exciting laser.
The wavelength is centred around 1542.4 nm and scanned in a range of about 3.5 GHz (few pm). The measured
transmission signal exhibits an intensity modulation with a period of about 0.5 GHz superposed to the acetylene Doppler
absorption. Our numerical setup relates to this example, however, it does not accurately model the experimental setup
nor the presence of the gas inside the PCF, which would be beyond the scope of this demonstrational study. Figure 6C
shows the FEM triangular mesh4 discretizing the cross section of a hollow core PCF with a central core of 7 missing
cladding pores, with a cladding with a periodicity length of 2.52 µm, a pore strut width of 80 nm, corner rounding radius
of 390 nm and 7 rings of cladding pores. A slight asymmetry of the geometrical setup is assumed (ellipticity of 2%)
which yields a small difference of the propagation constants of the two (nearly degenerate) fundamental modes. The
computed intensity distribution of one of these is visualized in Fig. 6D. For the corresponding FEM computation, higher-
order finite elements (p=3) are used5. In our computational setup, we first compute the modes of the involved fibres
(Fibre 1, 2, 3), then we compute the overlap integrals (which is done in a postprocess, utilizing the higher-order FEM
discretization of the field distributions). From application of the respective transfer matrices to the incoming fibre mode,
we compute the total power transmission through the setup. The transmission spectrum for a corresponding frequency
scan is shown in Fig. 6E. Please note, that here only the two fundamental modes contribute: in the numerical study, all
investigated higher-order propagation modes of the structure exhibited damping losses at least one order of magnitude
larger than the damping losses of the fundamental modes, we have therefore only considered the fundamental mode
propagation in the results shown in Fig. 6E.
Figure 6. A: Microscopic image of the investigated PCF. B: Detector signal proportional to the transmission through the
PCF shown in A, spliced in between two standard SMF and filled with acetylene. C: Geometrical model of the cross section
of a HCPCF (light grey: air, dark grey: silica) with triangular mesh for the FEM discretization. D: Visualization of the
electric field intensity distribution of a fundamental mode in the PCF (red: high intensity, blue: low intensity) with
superimposed cladding structure for visualization purposes (grey). E: Simulated transmission through a SMF -- HCPCF --
SMF setup.
5. CONCLUSION
FEM modelling techniques can be advantageously used to evaluate the transmission properties of complex waveguide-
based systems, especially in situations where the derivation of analytical solutions is getting too complicated, as it is the
case when using specialty or photonic crystal fibres. Another interesting domain of application is related to the
metrology of multimode waveguides, especially towards the definition of optimum templates for the modal distribution
in multimode fibres, to ensure repeatable power and loss measurements.
6. ACKNOWLEDGMENTS
This project has received funding from the EMPIR programme co-financed by the Participating States and from the
European Union's Horizon 2020 research and innovation programme under grant agreement number 14IND13
(PhotInd).
REFERENCES
[1] Mafi, A., Hofmann, P., Salvin, C.J. and Schülzgen, A., "Low-loss coupling between two single-mode optical
fibers with different mode-field diameters using a graded-index multimode optical fiber", Opt. Lett. 36(18),
3596-3598 (2011).
[2] JCMsuite, https://www.jcmwave.com.
[3] RP Photonics, https://www.rp-photonics.com.
[4] Pomplun, J., Zschiedrich, L., Klose, R., Schmidt, F., Burger,S., "Finite Element simulation of radiation losses in
photonic crystal fibers", Phys. stat. sol. (a) 204, 3822-3837 (2007).
[5] Burger, S., Zschiedrich, L., Pomplun, J., Herrmann, S., Schmidt, F., "Hp-finite element method for simulating
light scattering from complex 3D structures", Proc. SPIE 9424, 94240Z (2015).
|
1907.13072 | 1 | 1907 | 2019-07-30T16:57:28 | Optical properties of mist CVD grown $\alpha$-Ga$_2$O$_3$ | [
"physics.app-ph",
"cond-mat.mtrl-sci"
] | We report on the study of optical properties of mist CVD grown alpha Gallium oxide with the observation of excitonic absorption in spectral responsivity measurements. 163 nm of Gallium oxide was grown on sapphire using Gallium acetylacetonate as the starting solution at a substrate temperature of 450 deg C. The film was found to be crystalline and of alpha phase with an on axis full width at half maximum of 92 arcsec as confirmed from X ray diffraction scans. The Taucs plot extracted from absorption spectroscopy exhibited two transitions in the UV regime at 5.3 eV and 5.6 eV, corresponding to excitonic absorption and direct band to band transition respectively. The binding energy of exciton was extracted to be 114 meV from spectral responsivity measurements. Further, metal semiconductor metal photodetectors with lateral inter digitated geometry were fabricated on the film. A sharp band edge was observed at 230 nm in the spectral response with peak responsivity of around 1 Amperes per Watt at a bias of 20 V. The UV to visible rejection ratio was found to be around 100 while the dark current was measured to be around 0.1 nA. | physics.app-ph | physics | Optical properties of mist CVD grown α-Ga2O3
Usman Ul Muazzam1, Prasad chavan1, Srinivasan Raghavan1, R. Muralidharan1, Digbijoy N
Nath1
1 Centre for Nano Science and Engineering, Indian Institute of Science, 560012, Bangalore, India
We report on the study of optical properties of mist CVD grown α-Ga2O3 with the observation of excitonic
absorption in spectral responsivity measurements. 163 nm of Ga2O3 was grown on sapphire using Ga-(acac)3 as
the starting solution at a substrate temperature of 450°C. The film was found to be crystalline and of α-phase with
an on-axis full width at half maximum (FWHM) of 92 arcsec as confirmed from X-ray diffraction scans. The
Tauc's plot extracted from absorption spectroscopy exhibited two transitions in the UV regime at 5.3 eV and 5.6
eV, corresponding to excitonic absorption and direct band-to-band transition respectively. The binding energy of
exciton was extracted to be 114 meV from spectral responsivity (S.R) measurements. Further, metal-
semiconductor-metal (MSM) photodetectors (PD) with lateral inter-digitated geometry were fabricated on the
film. A sharp band edge was observed at 230 nm (~ 5.6 eV) in the spectral response with peak responsivity of ~1
A/W at a bias of 20 V. The UV to visible rejection ratio was found to be ~ 100 while the dark current was measured
to be ~ 0.1 nA.
The various polymorphs of Ga2O3, with their wide band gap of 4.6-5.3 eV eV1,2, have attracted attention of the
device community for their promises in the areas of high-power switching3, deep-UV optoelectronics4, gas
sensors5, high-temperature and transparent electronics6. UV-C photodetectors for instance, are useful in UV
astronomy, bio-medical and forensic applications, and for missile plume detection in the strategic sector7 -- 9.
Although the β phase is the most stable among the five different polymorphs (α, β, γ, ε, δ) of Ga 2O3 and thus has
been the most widely investigated, there has been an increasing interest in α-Ga2O3 in recent times. It has a
corundum crystal structure and has been predicted to have the highest bandgap (~ 5.3 eV10) among all the
polymorphs of Ga2O3. This makes α-Ga2O3 an attractive candidate for ultra-high breakdown transistors and deep-
UV opto-electronics at sub-240 nm wavelengths. The growth of α-Ga2O3, which requires relatively low
temperatures (430oC -- 470oC) 10, has been reported using approaches such as atomic layer deposition (ALD), mist
chemical vapor deposition and molecular beam epitaxy11,12. Although there is a report on the demonstration of
13 , the investigation of the growth as well as
field effect transistor (FET) based on mist CVD grown α- Ga2O3
structural, optical and electrical transport properties of this emerging polymorph of Ga2O3 is still at an embryonic
stage. In this letter, we report on the study of optical -- in particular excitonic - properties of mist CVD grown α-
Ga2O3 with a subsequent realization of a high-responsivity solar blind UV-C photodetectors.
The mist CVD system used for the growths has been developed in-house and consists of two parts, the reactor
and the mist generator. A volume of 0.33 mole 5N pure gallium acetylacetonate Ga(acac)3 dissolved in de-ionised
water (DI) was used as the source of gallium precursor. Small amount (0.1 ml) of HCl was added to ensure the
complete dissolution of Ga(acac)3 in DI water. This solution was then ultra-sonicated at a frequency of 1.6 MHz
using the mist-generator. The generated mist was directed to the deposition zone using N2 (500 sccm) as carrier
gas. c-plane sapphire wafer of 2-inch diameter was diced into 1 cm x 1 cm pieces and were solvent cleaned in
acetone, isopropyl alcohol and rinsed with DI water. For each growth run, a piece of sapphire was placed inside
the deposition zone using a quartz tube with diameter of 40 mm. The growth was carried out for one hour at a
temperature of 450 °C and at atmospheric pressure.
The XRD scans were carried out using a Rigaku SmartLab, Cu-Ka radiation X-ray diffraction system. The film
grown on sapphire was confirmed to be α-Ga2O3 from θ-2θ scan. Figure1 shows (0006) reflection of α-Ga2O3,
and the inset to figure1 shows the symmetric rocking curve with an FWHM of 92.2 arcsec indicative of a low
screw dislocation density in the epi-layer. The surface morphology of the as-deposited film was studied using
atomic force microscope (Dimension ICON, Bruker) and the rms roughness was found to be 2 nm as shown in
Figure 2(a), which confirms the smoothness of the film. The film thickness was found to be 163 nm from
ellipsometry measurements. Figure 2(b) shows the image of the as-grown film as obtained from scanning electron
microscope (GEMINI Ultra 55, FE-SEM, Carl Zeiss), indicating that the layer is continuous and uniform.
a) Corresponding author email: [email protected]
[email protected]
Absorption measurement was done using UV-Vis setup (UV-3600, UV-VIS-NIR spectrophotometer,
Shimadzu). The Tauc's plot (figure 3 ) exhibited a distinct kink in addition to the primary absorption edge. The
first edge at 5.3 eV corresponds to excitonic transition while the kink with sharp transition corresponds to band-
to-band absorption at 5.6 eV when extrapolated linearly to intersect the x-axis and indicates the band gap of α-
Ga2O3.
Photodetectors with metal semiconductor metal (MSM) layout in an interdigitated geometry were fabricated
on the as-grown α -Ga2O3 sample using standard i-line lithography process. The device schematic is shown in Fig.
4(a). Ni (20nm)/Au (100nm) stack was deposited using sputtering to form Schottky contact. Each MSM detector
as shown in Fig. 4(b), comprised of seventeen pairs of interdigitated fingers where each finger had a width of 4
µm and the finger spacing was 6 µm. The active area of each device was 260 µm x 300 µm.
Spectral responsivity (SR) measurement was done using a quantum efficiency setup, the details of which are
reported elsewhere14. The SR spectra exhibited a primary peak at 230 nm corresponding to band-to-band
absorption while an excitonic peak could be observed at 235 nm. The binding energy of exciton, estimated from
the difference between the two peaks, was found to be 114 meV, which is in close agreement with earlier reported
values15. This is also the first report of observation of excitonic peak in spectral response of any polymorph of
gallium oxide.
Raman spectra was recorded in the backscattering geometry using 532 nm laser, the light was then collected
using 100x objective in backscattered geometry and analysed using LabRAM HR, Horiba spectrometer. Figure 5.
shows Raman spectra of α -Ga2O3. The corundum structure of α-Ga2O3 belongs to -3m (D3d) point group and R-
6) space group. According to group theory analysis the irreducible representation of zone-centre optical
3c (D3d
mode is:
Г = 2A1g + 2A1u + 3A2g + 2A2u + 5Eg + 4Eu (1)
In addition, unit cell of α -Ga2O3 is centrosymmetric thus all vibrations that are Raman allowed are infrared
forbidden and vice-versa. The A1g and Eg are Raman active, A2u and Eu are infrared active, and A1u and A2g
vibrations are neither Raman nor infrared active. The spectrum shows A1g(LO) phonon mode at 216 cm-1. This
mode is attributed to Ga atoms vibrating against each other along c-axis. The high frequency Eg mode at 430 cm-
1 is due to lighter O atom vibrations perpendicular to c-axis16. Low intensity of Raman modes of ⍺-Ga2O3 may be
due to thinner sample.
In most of the oxide semiconductors, the excitonic binding energy is larger than the Rydberg exciton effective
energy which is given by:
Eexo = Ry
𝜇
𝑚𝑜
𝜖𝑠
(
𝜖𝑜
2 (2)
)
1
𝜇
=
1
𝑚𝑒
+
1
𝑚ℎ
Here, Ry is the Rydberg energy which has value of 13.6 eV, 𝜇 is the reduced mass of exciton, mo is free electron
mass, ϵs the static dielectric constant of Ga2O3 which is 1017. Since mh >> me, μ is approximately taken as me
18. From equation (1), the excitonic binding energy is found to be around 37.53 meV which is
which is 0.276mo
underestimated because we have not considered the interaction between LO phonons and excitons. Polar materials
have more than one atom per unit cell having non-zero Born effective charge; thus, atomic displacement
corresponding to polar LO phonons can give rise to microscopic Born effective electric fields at long wavelengths
leading to strong coupling between LO phonons and excitons19. Interaction between polar optical phonons and
excitons can be described using Frohlich coupling constant, which is given by20:
αF =
𝑞2
8𝜋𝜖𝑜ℏ
√
2𝑚𝑐
ℏ𝜔𝑜
(
1
𝜖∞
−
1
𝜖𝑠
) (3)
Using equations (2) and (3) 𝐸𝑒𝑥 can be estimated to be 91.71meV.
This value is very close to the value estimated from spectral responsivity. To estimate the dissociation field for
exciton, polaron (coupling of LO phonon with electron) radius was calculated using21:
ap = √
2ℏ
𝑚𝜔𝐿𝑂
(4)
a) Corresponding author email: [email protected]
[email protected]
where ap is Polaron radius. From equation (4), ap was found to be 32.2 Å, which corresponded to a dissociation
field of 0.354 MV/cm.
Figure6 (a) shows the variation of responsivity with wavelength at different voltages on linear scale (5 V, 10
V, 15 and 20 V). Inset to figure 6(a) shows the same in log scale. The peak responsivity was measured to be 0.95
A/W at 230 nm at a bias of 20 V. The UV to visible rejection ratio was calculated by dividing the responsivity
value at 230 nm by that at 400 nm, and was found to be> 102 at 20 V.
Figure 6(b) shows the current-voltage (I-V) characteristic of the detectors under dark and under illumination
at 230 nm. The photo current was found to be 85 nA while the dark current was measured to be 137 pA at an
applied bias of 20 V, indicating a photo-to-dark current ratio exceeding two orders of magnitude.
Figure 6(c) shows the variation of peak responsivity with applied voltage at an illumination of 230 nm. The
peak responsivity was found to increase with an increase in applied voltage.
The theoretical value of responsivity at 230 nm, assuming a quantum efficiency of 100%, is 185 mA/W. This
is much smaller than the peak responsivity value of 518 mA/W at 5 V (at 230 nm) measured in this work, even at
a relatively low bias of 5 V, implying that there is gain in the devices22 -- 24. This gain could be because of oxygen
vacancies25,which act as hole trapping centres in the bulk of the semiconductor which could leads to photoinduced
barrier lowering22 resulting in an increase in transit time.
In conclusion, we have reported on the study of growth and photo-response properties of mist CVD grown α-
Ga2O3 on c-plane sapphire. Solar blind deep-UV photodetectors realized on these samples exhibited high
responsivity of 0.5 A/W at 5 V bias with a sharp peak at 5.5 eV, low dark current of ~ pA and UV-to-visible
rejection ratio exceeding two orders of magnitude. We reported the first observation of excitonic peak in spectral
responsivity with an excitonic binding energy of 114 meV. This work is expected to aid further in the
understanding of optical properties of α-Ga2O3 Towards realizing high-performance deep-UV optoelectronics
based on gallium oxide.
This work was supported in part by the Ministry of Electronics and Information Technology (MeitY),
and in part by the DST through the NNetRA.
References
1 T. Wang, W. Li, C. Ni, and A. Janotti, Phys. Rev. Appl. 10, 1 (2018).
2 Y. Chen, X. Xia, H. Liang, Q. Abbas, Y. Liu, and G. Du, Cryst. Growth Des. 18, 1147 (2018).
3 M. Higashiwaki, K. Sasaki, H. Murakami, Y. Kumagai, A. Koukitu, A. Kuramata, T. Masui, and S. Yamakoshi,
Semicond. Sci. Technol. 31, 34001 (2016).
4 A. Singh Pratiyush, S. Krishnamoorthy, S. Vishnu Solanke, Z. Xia, R. Muralidharan, S. Rajan, and D.N. Nath,
Appl. Phys. Lett. 110, 1 (2017).
5 A. Trinchi, W. Wlodarski, and Y.X. Li, Sensors Actuators, B Chem. 100, 94 (2004).
6 M. Orita, H. Ohta, M. Hirano, and H. Hosono, Appl. Phys. Lett. 77, 4166 (2000).
7 L. Sang, M. Liao, and M. Sumiya, Sensors (Switzerland) 13, 10482 (2013).
8 E. Monroy, F. Omnès, and F. Calle, Semicond. Sci. Technol. 18, (2003).
9 M. Razeghi, Proc. IEEE 90, 1006 (2002).
10 D. Shinohara and S. Fujita, Jpn. J. Appl. Phys. 47, 7311 (2008).
11 S.H. Lee, K.M. Lee, Y.-B. Kim, Y.-J. Moon, S. Bin Kim, D. Bae, T.J. Kim, Y.D. Kim, S.-K. Kim, and S.W.
Lee, J. Alloys Compd. 780, 400 (2018).
12 X. Chen, Y. Xu, D. Zhou, S. Yang, F.F. Ren, H. Lu, K. Tang, S. Gu, R. Zhang, Y. Zheng, and J. Ye, ACS Appl.
Mater. Interfaces 9, 36997 (2017).
13 G.T. Dang, T. Kawaharamura, M. Furuta, and M.W. Allen, IEEE Trans. Electron Devices 62, 3640 (2015).
14 P. Jaiswal, U. Ul Muazzam, A.S. Pratiyush, N. Mohan, S. Raghavan, R. Muralidharan, S.A. Shivashankar, and
D.N. Nath, Appl. Phys. Lett. 112, (2018).
a) Corresponding author email: [email protected]
[email protected]
15 A. Segura, L. Artús, R. Cuscó, R. Goldhahn, and M. Feneberg, Phys. Rev. Mater. 1, 024604 (2017).
16 R. Cuscó, N. Domènech-Amador, T. Hatakeyama, T. Yamaguchi, T. Honda, and L. Artús, J. Appl. Phys. 117,
(2015).
17 M. Higashiwaki, K. Sasaki, A. Kuramata, T. Masui, and S. Yamakoshi, Phys. Status Solidi Appl. Mater. Sci.
211, 21 (2014).
18 H. He, R. Orlando, M.A. Blanco, R. Pandey, E. Amzallag, I. Baraille, and M. Rérat, Phys. Rev. B - Condens.
Matter Mater. Phys. 74, 1 (2006).
19 Peter Y. Yu and M. Cardona, Fundamentals of Semiconductors Physics and Materials Properties, Fourth Edi
(Springer, 2010).
20 N. Tanen, H. (Grace) Xing, Z. Guo, A. Verma, D. Jena, N. Ma, and T. Luo, Appl. Phys. Lett. 109, 212101
(2016).
21 J.T. Devreese, ArXiv:1012.4576 [Cond-Mat.Other] (2015).
22 O. Katz, V. Garber, B. Meyler, G. Bahir, and J. Salzman, Appl. Phys. Lett. 79, 1417 (2001).
23 H. Srour, J.P. Salvestrini, A. Ahaitouf, S. Gautier, T. Moudakir, B. Assouar, M. Abarkan, S. Hamady, and A.
Ougazzaden, Appl. Phys. Lett. 99, 3 (2011).
24 F. Xie, H. Lu, X. Xiu, D. Chen, P. Han, R. Zhang, and Y. Zheng, Solid. State. Electron. 57, 39 (2011).
25 A.M. Armstrong, M.H. Crawford, A. Jayawardena, A. Ahyi, and S. Dhar, J. Appl. Phys. 119, 1 (2016).
Figures
Figure: 1. XRD θ-2θ diffraction pattern of as deposited film of α-Ga2O3, inset shows rocking curve plot of (0006)
peak of ⍺-Ga2o3 with FWHM of 92.2 arcsec.
a) Corresponding author email: [email protected]
[email protected]
20406080101102103104105(0003)a-Al2O3(0006)a-Al2O3Intensity2q (degrees)(0006)a-Ga2O3-200-1000100200Intensity (a.u)w (arcsec)FWHM = 92.2 arcsec
(a)
(b)
200 nm
Figure2. (a)AFM scan image showing R.M.S roughness of 2 nm. (b) SEM micrograph showing smooth
morphology of as deposited film.
Figure3. Tauc's plot showing two transitions.
a) Corresponding author email: [email protected]
[email protected]
234560E+001E+102E+103E+104E+10Eg = 5.53 eV(ahu)2 (eV-2cm-2)hu (eV)Eg = 5.29 eV Ni/Au (20 nm/100 nm)
α - Ga2O3
Sapphire
(a)
(b)
Figure4 (a) Schematic of MSM photodetector (side view). (b) Optical micrograph of MSM photodetector (top
view).
Figure5 (a). Raman spectra of as deposited film. The green and black labels correspond to ⍺-Ga2O3 and Sapphire
peaks respectively.
a) Corresponding author email: [email protected]
[email protected]
100200300400500600700800EgEgEgA1gIntensity (a.u)Raman shift (cm-1)A1gEg (a)
(b)
(c)
Figure6 (a). Shows variation of spectral response with wavelength as a function of voltage in linear scale, inset
shows variation of SR with wavelength as a function of voltage in log scale. Also can be seen U.V-Visible
rejection ratio is > 102. (b) Variation of photocurrent (at 230 nm) and dark current with applied voltage. (c)
Variation of peak SR (at 230 nm) with applied voltage.
a) Corresponding author email: [email protected]
[email protected]
2202402602803003203403603804001x100Spectral Responsivity (A/W)Wavelength (nm) 5V 10V 15V 20V114 meV25030035040010-310-210-1100SR (A/W)Wavelength (nm) 5V 10V 15V 20VLog Scale481216200.60.81.0Peak SR (A/W)VoltagePeak SR @ 230 nm0510152010-1110-1010-910-810-7Current (A)Voltage (V) Photocurrent @ 230 nm Dark Current |
1804.01111 | 1 | 1804 | 2018-04-03T18:07:00 | Designing Thermoplasmonic Properties of Metallic Metasurfaces | [
"physics.app-ph",
"cond-mat.mes-hall"
] | Surface plasmons have been used recently to generate heat nanosources, the intensity of which can be tuned, for example, with the wavelength of the excitation radiation. In this paper, we present versatile analytical and numerical investigations for the three-dimensional computation of the temperature rise in complex planar arrays of metallic particles. In the particular case of elongated particles sustaining transverse and longitudinal plasmon modes, we show a simple temperature rise control of the surrounding medium when turning the incident polarization. This formalism is then used for designing novel thermoplasmonic metasurfaces for the nanoscale remote control of heat flux and temperature gradients. | physics.app-ph | physics | Designing Thermoplasmonic Properties of Metallic
Metasurfaces
Ch. Girard, P. R. Wiecha, A. Cuche, and E. Dujardin
CEMES, University of Toulouse and CNRS (UPR 8011), 29 rue Jeanne Marvig, BP
94347, 31055 Toulouse, France
Abstract. Surface plasmons have been used recently to generate heat nanosources,
the intensity of which can be tuned, for example, with the wavelength of the excitation
radiation. In this paper, we present versatile analytical and numerical investigations
for the three -- dimensional computation of the temperature rise in complex planar
arrays of metallic particles.
In the particular case of elongated particles sustaining
transverse and longitudinal plasmon modes, we show a simple temperature rise control
of the surrounding medium when turning the incident polarization. This formalism is
then used for designing novel thermoplasmonic metasurfaces for the nanoscale remote
control of heat flux and temperature gradients.
PACS numbers: 41.20.-q, 78.20.Bh, 73.20.Mf
8
1
0
2
r
p
A
3
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
1
v
1
1
1
1
0
.
4
0
8
1
:
v
i
X
r
a
Designing Thermoplasmonic Properties of Metallic Metasurfaces
2
1. Introduction
The ability of metal structures to confine the electromagnetic fields gave birth to a
multitude of applications in areas as diverse as biophysics, sensor technology or devices
for fast data processing[1, 2, 3]. The light confinement phenomenon originates in the
surface plasmons (SP) travelling or localized at the surface of these nanostructures.
Most plasmonics applications, based on the engineering of surface plasmons, exploit the
electromagnetic fields produced by the collective electronic oscillations. In particular, SP
engineering has been considered as a viable approach to the coplanar implementation of
high speed, low dissipative information devices using analogical or digital concepts[4, 5].
Very recently, plasmonics has fostered another realm of applications in which dissipative
effects are being advantageously utilized [6, 7, 8, 9, 10, 11, 12, 13, 14, 15]. Indeed, besides
their widely used propensity to enhance and confine the near -- field electromagnetic
intensity, metal particles and nanostructures have revealed a great potential as local
heat sources[6, 7, 11, 16]. A realistic description of such localized dissipation effects
is directly related to the description of the imaginary part of the dynamical response
functions of the nanostructures, such as the dielectric permittivity of the metal (ω), and
the local electric field intensity Il(r, ω) induced inside the metal[7, 11, 17, 18]. While the
dielectric constant only depends on the nature of the metal, the intensity distribution
of the optical electric field induced in the particle is extremely sensitive to the presence
of plasmon resonances occurring in the spectral variation of the local field distribution
Il(r, ω)[7, 17, 19, 20]. These resonances play a crucial role since the amount of heat
tranferred to the particle can be adjusted by tuning the incident wavelength in or out of
the resonance range. Several experimental thermoplasmonic building blocks have indeed
been designed from this concept and realized from colloidal chemistry or sophisticated
lithography processes [13, 14, 16, 21].
In this paper, we propose a flexible analytical scheme [22, 23] completed by
numerical studies [19] to investigate the thermoplasmonic properties of arrays of
individual plasmonic entities with arbitrary shapes sustaining multiple plasmon modes
in the optical range. Arrays of nanostructures are systems of fundamental interest
in plasmonics since they combine the optical properties of individual resonators and
the collective response of the assembly. Such systems have already contributed to major
breakthroughs in several fields in optics like optical sensing, metasurfaces for light phase
and orbital angular momentum control, strong optical coupling, ... [24, 25, 26, 27]. In
a typical configuration, the particle arrangement is supported by a planar dielectric
substrate. Here we apply the well -- established self -- consistent scheme based on the
Green Dyadic Functions (GDF) formalism and compute the local field intensity inside the
particle array by including the coupling with the substrate. The average temperature
in the vicinity of the metallic sructures is then derived from the local field intensity.
As a first step towards realistic configurations, we describe each individual metal
structure as an anisotropic polarizable particle excited by their self -- consistent local
electric field[28]. This first approach consists in gradually developping a quasi -- analytical
Designing Thermoplasmonic Properties of Metallic Metasurfaces
3
description of the calculation that provides an intuitive access to the underlying physical
and thermal mechanisms. This simple analytical description is then complemented by
an adequate discretisation of the particle physical volumes in order to describe arbitrary
geometries[19]. The numerical applications are based on the permittivity of gold taken
from Johnson and Christy data[29]. In the third section, the specific case of periodic
arrays of gold nanorods is investigated and we demonstrate a simple and reversible
control of the temperature rise near the particles when turning the incident polarization.
Applications to the new concept of thermoplasmonic metasurfaces are then discussed
in the two last sections where we demonstrate that our numerical technique is well --
suited for the design of optimized thermoplasmonic meta -- cells, using an evolutionary
optimization (EO) algorithm.
Figure 1. (color online) Perspective view illustrating a periodic assembly of plasmonic
structures fabricated at the surface of an insulating sample.
2. Thermal response of a periodic array of identical metal particles
Let us consider a periodic 2D array of N elongated gold structures arranged in a periodic
way at the surface of a dielectric planar substrate (Fig. (1)). The particle location is
defined by a set of N vectors ri = (Li, Z) (where the two -- dimensional vector Li belongs
to the (XOY ) plane). Unlike what happens with perfectly spherical particles, single
nanorods exhibit extinction spectra with two plasmon bands that correspond to electron
oscillations along their length (low energy longitudinal mode) and across their section
(high energy transverse mode) [30]. This shape effect can be described with a simple
analytical model by using an anisotropic dynamical polarisability [28]. When the long
axis of the particles is aligned along (OY) axis as shown in figure (1), the polarizability
tensor is diagonal and reads:
α⊥(ω0)
0
0
α(ω0) =
,
0
α(cid:107)(ω0)
0
0
0
α⊥(ω0)
where the two independent components α⊥(ω0) and α(cid:107)(ω0) can be described by the
formula associated with a prolate ellipsoid [28].
(1)
nZYX1n2 Designing Thermoplasmonic Properties of Metallic Metasurfaces
4
2.1. Local field calculation
When a monochromatic electromagnetic plane wave of frequency ω0 and electric field
amplitude E0 hits the interface between environment (media 1) and the glass substrate
(media 2) at normal incidence and interacts with the metallic particles, the optical
electric field can be written:
E0(r, t) =
1
2
{E0(r, ω0) exp(iω0t) + C.C.} ,
(2)
where E0(r, ω0) (with r = (x, y, z)) represents its Fourier amplitude:
E0(r, ω0) = E0[exp(−in1k0z) + R exp(in1k0z)] ,
(3)
in which k0 is the wave vector modulus in vacuum and R = (n1 − n2)/(n1 + n2) is the
Fresnel reflection coefficient expressed with the optical indices of surrounding medium
(n1) and dielectric substrate (n2), respectively. The polarization of the incident wave,
associated with the direction of the vector E0, can be materialized by the angle θ between
E0 and the (OX) axis:
E0 = E0(cos(θ), sin(θ), 0) .
(4)
2.1.1. Self -- consistency The local fields E(ri, ω0) induced at the center of the particles
by the illumination field verify a set of N coupled equations that can be condensed as
follows:
E0(ω0) = M(ω0) · E(ω0)
(5)
where E0(ω0) is the input supervector that contains the N incident fields at the particle
locations:
E0(ω0) = {E0(r1, ω0), ...., E0(ri, ω0), ...} ,
(6)
and E(ω0) is the output supervector that contains the local field values:
E(ω0) = {E(r1, ω0), ...., E(ri, ω0), ...} .
(7)
For N polarizable particles, the (3N × 3N ) coupling matrix M(ω0) has a very simple
form given by:
M(ω0) = I − A(ω0) ,
where I is the identity matrix and:
A11(ω0)
,
A1j(ω0)
....
....
....
....
....
....
AN N (ω0)
(8)
(9)
....
....
Aij(ω0)
....
....
....
....
A(ω0) =
is composed of N 2 submatrices defined from the particle polarizabilities and the field --
propagators S(ri, rj, ω0) between two particles locations, ri and rj:
Aij(ω0) = S(ri, rj, ω0) · α(ω0)
(10)
Designing Thermoplasmonic Properties of Metallic Metasurfaces
5
2.1.2. Weak coupling between individual metallic nanotructures The calculation of the
local field in the particle array needs the inversion of the matrix M(ω0) shown by
equation (5), which can be considerably simplified depending on the interparticle spacing
D. For example, when the lateral spacing D is of the order of the incident wavelength
λ0 and the thickness of the metal particles is much smaller, the mutual interactions
vanish so that all the tensorial components S(ri, rj, ω0) · α(ω0) (cid:28) 1. This hypothesis,
that corresponds to the first Born approximation (FBA), leads to the simplified relation:
M−1(ω0) = I + A(ω0) + O(A2) ,
(11)
2.2. Dissipated power
During the illumination process, the temperature rises because of the electronic Joule
effect induced inside the metal particles. This dissipative energy channel can be
described by computing the power per unit volume dissipated inside the metal. From
the electric field Ei(r, t) and the induction vector Di(r, t), we can derive the amount of
power dissipated by the ith metallic particle. In CGS electrostatic units, this leads to:
Q(ri) =
1
4π
dr < Ei(r, t) · ∂
∂t
Di(r, t) > ,
(12)
vi
where the integral runs over the particle volume, and the brakets schematize the time
average. Similarly to equation (2), the vectors Ei(r, t) and Di(r, t) can be expressed in
term of their Fourier amplitudes Ei(r, ω0) and Di(r, ω0). After taking the time average,
this transformation leads to:
(cid:90)
(cid:110)
Ei(r, ω0) · D(cid:63)
(cid:111)
iω0
i (r, ω0) · Di(r, ω0)
Q(ri) =
−E(cid:63)
16π
dr ,
1
vi
i (r, ω0)
(cid:90)
(cid:88)
β
(13)
Next, we introduce the constitutive relation between Di(r, ω0) and Ei(r, ω0). Rewriting
this equation in a tensorial form will allow to consider non -- spherical particles:
Di,α(r, ω0) =
α,β(ω0)Ei,β(r, ω0) ,
(14)
where α and β are two cartesian indices. After replacing (14) into (13) and assuming a
diagonal form for α,β(ω0) one gets:
(cid:90)
(cid:110)(cid:88)
(cid:61)α,α(ω0)Ei,β(r, ω0)2(cid:111)
Q(ri) =
ω0
8π
(15)
where the (cid:61) symbol means imaginary part. Equation (15) is general and does not contain
any approximation and applies to arbitrary geometries and any type of materials.
dr ,
vi
α
2.2.1. Dipolar response approximation For metallic particles of small size, the
multipolar contributions higher than the dipolar one can be neglected.
In this case,
Designing Thermoplasmonic Properties of Metallic Metasurfaces
6
the permittivity of the ith particle can be schematized by the following relation in which
δ(r − ri) represents the Dirac δ distribution centered around the particle location ri:
× δ(r − ri) .
0
0
α⊥(ω0)
α(cid:107)(ω0)
α⊥(ω0)
(cid:110)(cid:61)α⊥(ω0)Ei,y(ri, ω0)2
0
0
0
0
(ω0) = 1 + 4π
(16)
From this simplified relation it is straightforward to perform the volume integral found
in equation (15):
ω0
2
Q(ri) =
+(cid:61)α(cid:107)(ω0)(Ei,x(ri, ω0)2 + Ei,z(ri, ω0)2)
(cid:111)
,
(17)
Notably, a more accurate calculation of both the local fields and the successive
field-gradients would require to go beyond the dipolar approximation. However, the
proposed description makes it possible to derive the analytical calculation throughout
its development.
2.2.2. The particular case of a single elongated plasmonic particle When the metallic
pattern simply consists in a single particle located at the position r1, the collective
effects vanish and the matrix M−1(ω0) = I which means that the local electric field
E(r1, ω0) = E0(r1, ω0). This asymptotic case gives rise to a particularly simple equation
that can be obtained by using Eqs. (3) and (4):
Q =
E2
0 ω0
2
{(cid:61)α⊥(ω0) cos2(θ) + (cid:61)α(cid:107)(ω0) sin2(θ)} .
(18)
Here, we explicitly see how tuning the polarization angle θ controls the amount of heat
transferred to the particle. Finally, after introducing the well -- known relation between
electric field amplitude E0 and laser power S0 delivered per unit area [31]:
E2
0 =
8πS0
c
,
one obtains:
Q = 4πS0k0{(cid:61)α⊥(ω0) cos2(θ) + (cid:61)α(cid:107)(ω0) sin2(θ)} .
(19)
(20)
In this last equation, the influence of the relative weights of both transverse and
longitudinal plasmon modes clearly appears through the two components of the nanorod
polarizability. To illustrate how this equation governs the heat absorbed by a single
ellipsoidal gold nanorod, we present in Fig (2) a sequence of four maps Q(λ0, θ) for aspect
ratios η = a/b varying from 1 to 4. The laser power S0 is set at 1 mW/µm2 and the
particle long axis is chosen parallel to the (OY ) cartesian axis. In the (λ0, θ) coordinate
plane, the resulting heat response of the particle varies from polarization -- independent
dissipation maximized at low incident wavelength for the sphere case (characterized by
a single color band), to an oblong domain that is shifted to longer incident wavelength
as η increases and shows an optimum for 90o polarization angle.
Designing Thermoplasmonic Properties of Metallic Metasurfaces
7
Figure 2. (color online) Simulation of four color plots of the heat Q(λ0, θ) dissipated
per time unit as a function of the incident wavelength (500 nm ≤ λ0 ≤ 850 nm) and
the polarization angle (0 ≤ θ ≤ 180o) for a gold prolate ellipsoid of short and long
axis b and a, respectively. In maps (A) to (D), four aspect ratio η = a/b have been
considered: (A) sphere case (η =1) a = b = 10 nm; (B) ellipsoid (η =2) a = 20 nm
and b = 10 nm; (C) ellipsoid (η =3) a = 30 nm and b = 10 nm; and (D) ellipsoid (η
=4) a = 40 nm and b = 10 nm. All the color bars are expressed in nanoWatt (nW).
2.2.3. Volume discretization approach Finding exact solutions of equation (12) for
more realistic situations requires an additional volume discretization procedure of the
source region occupied by the plasmonic particles. Generally, each particle volume Vp is
discretized into np identical elementary volumes vp. Such a procedure converts integrals
into discrete summations. The main analytical steps of this technique are detailed in
reference[19] and lead to:
N(cid:88)
E(rt,i, ω0) = E0(r, ω0) +
×S(rt,i, rp,j, ω0) · E(rp,j, ω0) .
p=1
np(cid:88)
j=1
χp(ω0)
In this expression, the parameters χp(ω0) associated with the elementary volumes vp are
homegeneous to dipolar polarizabilities:
p(ω0) − env(ω0)
χp(ω0) =
4π
vp .
The vectors rp,j and rt,i represent the location of jth and ith discretized cells inside the
pth and tth metallic particles, respectively. Next, the self -- consistent electric field inside
the metal particles is computed. This procedure leads to a system of N × np vectorial
equations with N × np unknown fields E(rp,j, ω0).
(21)
(22)
0306090120150500600700800DCB Incident wavelength (nm) Polarization angle (°)0.03.55006007008000306090120150 Incident wavelength (nm) Polarization angle (°)0355006007008000306090120150 Polarization angle (°)Incident wavelength (nm)0180A5006007008000306090120150 Polarization angle (°) Incident wavelength (nm)0315 Designing Thermoplasmonic Properties of Metallic Metasurfaces
8
For a given metal particle (labelled by the subscript p), the solving procedure
detailed in reference[19] is directly related to the discretization volume vp, which itself
depends on the discretization grid used to mesh the particles. The expressions of the
χp(ω0) coefficients for both cubic and hexagonal compact discretization grids can be
found in table (1) of reference [19].
3. Temperature Profile
Figure 3. (color online) Simulation of three temperature spectra computed above a
set of nine cylindric gold nanorods (25 × 80 nm) illuminated in normal incidence by
a linearly polarized plane wave (see geometry insert). The laser power S0 is fixed at
5 mW/µm2 and the lateral pitches, dx and dy between the rods is 500 nm. The three
spectra correspond to the polarization directions indicated by the red, blue and green
arrows. The computation has been performed by discretization of the nanorod volumes
by N × 333 elementary cells distributed over a hexagonal compact three -- dimensional
mesh.
The steady-state temperature increase, ∆T (λ0, θ), can be deduced from the
(15)), by
distribution of heat Q(rp,j) deposited inside the particle lattice (cf. Eq.
the thermal Poisson equation:
(23)
∆T (λ0, θ) =
1
4πκ
Q(rp,j)
Robs − rp,j
N(cid:88)
np(cid:88)
p=1
j=1
where κ is the environmental thermal conductivity and Robs defines an observation point
located in the vicinity of the sample. In figures (3) and (4), we have used relation (23)
(nm)λxyddT(o) 0 6 7 450 500 550 600 650 700 750 4 3 2 1 5 Designing Thermoplasmonic Properties of Metallic Metasurfaces
9
(color online) 3D color plots of three simulations of the temperature
Figure 4.
distribution above a set of nine gold nanorods corresponding to incident polarization
aligned (A) along Y axis, (B) at 45o and (C) along the X axis (same parameters as
figure (3)). The wavelength, λ0 = 680 nm, is chosen at the center of the longitudinal
plasmon band.
to investigate the photothermal effects induced near an array of nine gold nanorods
deposited on a dielectric surface (see insert of figure (3)). The metal particles are
surrounded by an isotropic medium of refractive index n1 = 1.33, mimicking an aqueous
medium.
The spectral variation of the temperature as a function of the incident wavelength
is presented in figure (3). The temperature shift ∆T (λ0, θ) is computed from equation
(23) at a position Robs = (0, 0, 150nm) which overhangs the central gold pad. The three
polarisation directions considered here, i.e. zero degree (red curve); 45o (blue curve)
and 90o (green curve), demonstrate that the temperature around the metal pattern can
be effectively tuned with a drastic increase observed near the longitudinal resonance.
Thus, by exciting the longitudinal plasmon band (λ = 680 nm) of the sample, the local
temperature can be modulated over one order of magnitude by the simple tuning of
the the field polarization. Obviously, this control is much less effective when exciting
the transverse mode (λ = 530 nm) because of a weaker quality factor.
In addition
this resonance is bound by two isobestic points [32] at 500 and 540 nm, where the
temperature is independent of the polarization (Figure 3).
Finally, figure (4) shows a sequence of three temperature maps resulting from the
monochromatic excitation of the longitudinal band (λ = 680 nm) at normal incidence
YXY(o)(o)(o)TXTABCXYT Designing Thermoplasmonic Properties of Metallic Metasurfaces
10
with a plane wave. All three maps are displayed with the same vertical scale to highlight
the temperature rise occurring when the incident polarization is aligned with the main
axis of the nanorods. A general observation is the strong temperature rise in the direct
vicinity of individual metallic structures and the broader yet less intense temperature
increase over the entire pattern [21].
4. Metasurfaces for Thermoplasmonic control
Figure 5. (color online) Example of thermoplasmonic metasurface able to generate
strong temperature contrast. The unit cell, located inside the dashed white frame,
is a set of four gold nanorods perpendicular to each other. The labels (a) and
(b) represents two consecutive rod orientations, parallel and antiparallel, inside the
pattern. The metasurface is superimposed by a temperature map computed with a
incident polarization aligned along the (a) nanorods.
In the recent years, by designing the surface of some materials at a subwavelength
scale, new applications for optics were highlighted (see for example [33]). These new
structures have been referred to as metasurfaces because they can modify the main
physical characteristics of the incident light. These modifications include, the phase,
the angular momentum, or the light polarization [34], and can generate and exalt
nonlinear phenomena [35, 36, 37]. For example, plasmonic metasurfaces containing
two -- dimensional subwavelength gold patterns have been developed that allow imprinting
arbitrary phase patterns onto a propagating beam [25] and perpendicular gold nanorods
arrays have been used to perform polarization conversion control from the capacitive
coupling to the conductive coupling regimes between the gold entities [34].
Metasurface physics also presents attractive opportunities for the thermoplasmon-
ics. In particular, the design and the juxtaposition of elementary plasmonic cells sen-
sitive to the incident polarization is a good manner to lead to efficient temperature
(a)(b) Designing Thermoplasmonic Properties of Metallic Metasurfaces
11
gradient controls. In figure (5), we consider a simple example of metasurface consisting
of elementary cells containing four nanorods.
In this paving, two consecutive plas-
monic elements are perpendicular relative to each other, so that the excitation of the
transverse and longitudinal modes will move from one structure to its neighbor, when
progressively turning the polarization angle of the incident light. The next figure (6)
displays the temperature map evolution expected around this metasurface geometry.
These computations have been performed by keeping the same parameters as in the
previous section, i.e., 16 cylindric gold nanorods (25 × 80 nm) illuminated in normal
incidence by a linearly polarized plane wave (see geometry insert). The laser power S0 is
fixed at 5 mW/µm2 and the lateral pitches, dx and dy between the rods are 250 nm. A
detailed examination of the maps shows a regular shaping of the temperature distribu-
tion, in a range that varies from 5 degrees for an incident power of 5 mW/µm2, and that
displays a periodic series of hot spots with tunable location by applying a remote control
of the polarization. For example, as shown in the first map of figure (6), a polarization
√
parallel to OX axis yields a hot spot pattern distributed on a square lattice with a side
2 × the lateral nanorod pitch dx. Consequently, by gradually turning the
equal to
incident polarization, we can accurately drag these hot spots, in a controllable manner,
from the parallel to perpendicularly oriented nanorods. In addition, as described in the
Supplemental Information Document (see figures S2 and S3), such a local temperature
gradient tuning leads to the possibility of controlling the heat flux in the vicinity of the
metasurface.
This functionality is particularly attractive for nano-biology manipulations and
applications, where tuning the symmetry of such temperature gradients would generate
complex convection currents in liquids that could be advantageously exploited, for
example, to thermally assisted plasmonic trapping [38, 39], to trigger thermotactic
mobility or behavioral plasticity on demand [40, 41].
5. Evolutionary optimization of optical hybrid material meta -- cells
In order to complete this theoretical paper, we demonstrate that our numerical
technique is well -- suited for the design of optimized thermoplasmonic meta -- cells, using
an evolutionary optimization (EO) algorithm. In the recent past, EO techniques have
been successfully applied on various problems in nano -- optics[42, 43, 44]. We will also
use a multi-material structure in our demonstration, which means that each meta-cell
(or meta-unit) is composed of multiple elements of different materials. Multi-material
systems can be modeled with our approach by using position-dependent dielectric
functions p(ω0) in equation (22) for the meshpoints at rp. A full metasurface would
finally consist of many of those optimized meta-units.
Designing Thermoplasmonic Properties of Metallic Metasurfaces
12
Figure 6. (color online) (A) Top view of the thermoplasmonic metasurface described
in figure (5).
(B) to (G) evolution of the temperature maps when the incident
polarization, represented by a double red arrow, is turned from 0deg to 90deg with
18deg steps.
5.1. Evolutionary optimization of photonic nano-structures
As illustrated in our previous examples (see section (4)), the design of geometric
assemblies of nanostructures starts with the conception of a reference geometry by
intuitive considerations. Via the systematic variation of a few parameters, this reference
systems is then optimized within its possibilities. Such an approach, however, is limited
to rather simple problems and requires a certain degree of understanding and intuition
Co)ABDEFGT∆( Designing Thermoplasmonic Properties of Metallic Metasurfaces
13
for the considered system. In case of complex structures or complicated phenomena,
the intuitive method often fails. To overcome these limitations, we apply in this section
an evolutionary optimization algorithm in order to design a meta -- unit for optimum
nano -- scale heat generation. EO is a heuristic optimization method to find the global
maximum or minimum of complex, possibly non-analytic problems. EO algorithms
use a population of parameter -- sets for the problem, which are driven through a cycle of
reproduction, evaluation and selection, in which weak solutions are iteratively eliminated
and strong parameter-sets are kept. Here, we couple the "jDe" EO algorithm [45]
provided by the "paGMO" toolkit [46] to our volume discretization method for full -- field
electrodynamical simulations. For these simulations, we use our own toolkit "pyGDM"
[47]. More details on the approach can be found in Ref. [44].
(color online) A) sketch of the evolutionary optimization scheme.
Figure 7.
(B)
illustration of the optimization model and problem. A chain of 20 nano-rods on
a glass substrate is searched in order to maximize the temperature increase at a
specific location (x = 4500, y = 0, z = 300) [nm], far from the focal spot of a focused
illumination (at (x = 0, y = 0, z = 0) [nm]). The free parameters of the optimization
are the angles αi as well as the material (either gold or silicon) of each rod. As further
parameter, the spacing D between the rods is optimized by the algorithm.
5.2. Optimization of a chain of nano-rods for localized heating
This problem is inspired by works which have shown that chains of both, dielectric and
metallic nano-particles can effectively guide light along relatively large distances[49, 50].
Here, the aim of the optimization is to find a chain of 20 nano-rods (each rod of size
70×175×140 nm3) which delivers the highest possible temperature increase at a specific
location far from a focused illumination with a wavelength of λ0 = 600 nm. The focal
spot with a beam waist of w0 = 300 nm is set at the origin, centered on the first nano-
rod (see also subplots (ii) in Fig. 8). The rods lie on a glass substrate (n = 1.5, κ = 0.8
W m−1K−1), in water (n = 1.33, κ = 0.6 W m−1K−1). The temperature increase ∆T
is to be maximized at (x = 4500, y = 0, z = 300) [nm]. The free parameters for the
optimization are each rod's rotation angle αi, each rod's material (either gold or silicon,
refractive indices taken from Refs. [29, 48]) and the distance D between the nano-rods
(which are equidistant along the chain). The geometry of the problem is depicted in
xyzfocused illumina-tion on first rodat (0,0,0) [nm]D.....DDαimaximize temperature at:(4500, 0, 300) [in nm]chain of 20 rods, each rod made either from gold or from siliconsame distance Dbetween all rodseach rod: individual rotation angle αi BAselection reproduction evaluation stop-criterion met?quit cycle, takebest solutionrandominitialization Designing Thermoplasmonic Properties of Metallic Metasurfaces
14
Fig. 7B. It is obvious, that a systematic evaluation of all possible solutions is impossible,
considering the 41 free parameters. For the EO algorithm we use a population of 50
individuals, which we evolve for 2000 iterations. On an ordinary office PC (AMD FX-
8350 CPU) one run never took longer than 24 hours. Figure 8 shows the results of the
optimizations, the fitness as a function of the iteration number is shown in the subplots
(i).
In order to verify the convergence, we ran the optimizations several times with
random initial parameters. The different runs yielded similar results, hence we conclude
that the optimizations converged close to the global optimum.
Figure 8. (color online) Results of the EO of a chain for a maximization of (A) the
absolute temperature increase at the target location and (B) the relative temperature
increase at the target position, normalized to ∆T@laser at the location of the focused
illumination. (i) convergence of the optimization. (ii) intensity of the incident field E0,
with focal spot of waist w0 = 300 nm at (0, 0, 0) nm, linearly polarized along OY (λ0
= 600nm). (iii) scattered field intensity on a logarithmic color scale, calculated 50nm
above the chain of nano -- rods (normalized to E02). (iv) mapping of the temperature
increase along the chain. (v) and (vi) zooms around the focal spot and the location
of the ∆T probe. All temperature mappings are calculated at a height of z = 300
nm. White scale bars are 500 nm. The optimization target location is indicated by a
cross -- shaped marker.
the absolute
5.2.1. Maximize absolute temperature increase In the first
temperature difference ∆T is the optimization target. The results are shown in figure 8A.
The EO algorithm chose a spacing D of approximately 250nm. It placed two gold rods
in the beginning of the chain, which are aligned with the incident light's polarization in
order to obtain the strongest response and, consequently, the highest dissipation inside
the metal. Following the first two gold blocks, a chain of silicon rods is guiding light
run,
Amaximize absolute ΔTBmaximize relative temp.: ΔT/ ΔT@laser(ii)(iv)(iv)(v)(vi)(vi)(v)E0ΔT = 2.38 K(ΔT/ΔT@laser = 0.051)ΔT = 0.73 K(ΔT/ΔT@laser = 2.83 )(i)(ii)E0(i)(iii)(iii) Designing Thermoplasmonic Properties of Metallic Metasurfaces
15
towards the target location, where further gold rods dissipate the arriving light into
heat (see the scattered near-field in subplot (iii) of figure 8). The heat radiated from
the two initial blocks locally rises the temperature by more than 45◦. At x = 4.5 µm,
the temperature still is increased by more then 2 degrees. The temperature rise is shown
in subplots (iv)-(vi) of figure 8. The heat generation from the terminal gold elements in
the chain is only contributing weakly to the total temperature increase, adding about
0.25◦.
5.2.2. Maximize normalized temperature increase In a second run, the goal of the
optimization is set to the relative temperature increase ∆T /∆T@laser. We aim to
maximize the temperature at the target location (4500, 0, 300) [nm] normalized to the
temperature at the position of the illumination beam at (0,0,300) [nm]. The results
of this simulation are shown in figure 8B. The spacing D was set by the EO again to
approximately 250 nm, which seems to be ideal for guiding light trough the chain at the
illumination wavelength λ0 =600 nm. However, in contrast to the maximization of the
absolute temperature, the EO algorithm chose silicon for the entire chain and placed
two gold elements only at its very end. The first two nano -- rods were furthermore
rotated perpendicular to the light's polarization in order to minimize their optical
response and hence reduce dissipation in the early chain. If the silicon-rod at (0,0,0)
were oriented along the polarization of E0, it would induce a temperature increase of
≈ 1.4◦ at (0, 0, 300) [nm]. Despite their horizontal orientation, the first two silicon rods
effectively couple light into the chain, through which it is guided towards the gold-rods
at its end. There the light is dissipated inside the metal, which acts as a local heat
source just below the target position. The solution, found by the EO, has a very low
∆T@laser ≈ 0.13◦ at the origin, but increases the temperature at the target location by
∆T ≈ 0.73◦.Compared to ∆Tlaser, this is ×2.8 higher.
6. Conclusion
To conclude, we have presented new geometries based on arrays of subwavelength
metallic nanoparticles organized at the surface of a dielectric substrate that results
in thermoplasmonic metasurfaces. The different kinds of arrays discussed here make
possible an on-demand and spatially controlled rise of temperature by the mean of the
incident wavelength and polarization. This approach, based on localized resonances,
is complementary to high order plasmonic resonances in larger 2D cavities that allow
for an all-optical and polarization dependent control of the temperature landscape in
their vicinity [51]. The formalism used in this work is optimized for the direct space
applications and provides convenient analytical formula, with which the mechanisms for
converting light energy into heat can be intuitively represented for both spherical and
elongated metallic particles. Using anisotropic polarizabilities, our approach reveals
the clear relationship between excitation parameters (laser power, polarization, and
wavelength) and expected thermal effects (heat amount, temperature, ...). Interestingly,
Designing Thermoplasmonic Properties of Metallic Metasurfaces
16
the extension of this formalism by means of a volume discretization procedure of the
metallic structures placed in the vicinity of a solid -- liquid interface, supplies a numerical
test bench for future applications of complex metasurfaces in thermoplasmonics, such
as those encountered in micro-fluidic environments for nano-biology. Finally, we
demonstrated that evolutionary optimization together with nano-optical simulations
allows finding geometric assemblies for complex thermoplasmonic problems. Using an
appropriately formulated problem, a hybrid -- material chain of silicon and gold nano-rods
can be optimized such, that the optical energy, delivered by a focused illumination spot,
is transferred towards a distant location where it is locally transformed into heat.
Designing Thermoplasmonic Properties of Metallic Metasurfaces
17
(Grants ANR-13-BS10-0007-PlaCoRe),
Acknowledgments: This work was
supported by the Agence Nationale
de la Recherche (ANR)
the Programme
Investissements d'Avenir under the program ANR-11-IDEX-0002-02, reference ANR-
10-LABX-0037-NEXT, and the computing center CALMIP in Toulouse.
[1] W. L. Barnes, A. Dereux, and T. W. Ebbesen, Nature 424, 824-830 (2003).
[2] N. J. Halas,S. Lal,W.-S. Chang, S. Link,and P. Nordlander, Chem. Rev. 111, 3913-3961 (2011).
[3] A. G. Brolo, Nature Photonics 6, 709 (2012).
[4] H. Wei, Z. Wang, X. Tian, M. Kall, H. Xu, Nat. Communications 2, 387 (2011).
[5] S. Viarbitskaya, A. Teulle, R. Marty, J. Sharma, C. Girard, A. Arbouet, and E. Dujardin, Nat.
Materials 12, 426 (2013).
[6] A. O. Govorov, W. Zhang, T. Skeini, H. H. Richardson, J. Lee, and N. A. Kotov, Nanoscale Res.
Lett. 1, 84 (2006).
[7] A. O. Govorov and H. H. Richardson, Nanotoday 2, 30 (2007).
[8] H. H. Richardson, M. T. Carlson, P. J. Tandler, P. Hernandez, and A. O. Govorov, NanoLetters,
9, 1139 (2009).
[9] J. Chen, W.-K. Chen, J. Tang, and P. M. Rentzepis, PNAS, 108, 18887 (2011).
[10] H. Ma, P. M. Bendix, and L. B. Oddershede, NanoLetters, 12, 3954 (2012).
[11] G. Baffou and R. Quidant, Laser Photonics Rev. 7, 171 (2013)
[12] A. Cuche, A. Canaguier-Durand, E. Devaux, J. A. Hutchison, C. Genet, and T. W. Ebbesen,
NanoLetters, 13, 4230 (2013).
[13] Z. J. Coppens, W. Li, D. G. Walker, and J. G. Valentine, NanoLetters, 13, 1023 (2013).
[14] B. Desiatov, I. Goykhman, and U. Levy, NanoLetters, 14, 648 (2014).
[15] J. B. Herzog, M. W. Knight, and D. Natelson NanoLetters, 14, 499 (2014).
[16] A. Sanchot, G. Baffou, R. Marty, A. Arbouet, R. Quidant, Ch. Girard, and E. Dujardin ACS
Nano, 6, 3434 (2012).
[17] G. Baffou, R. Quidant, and Ch. Girard, Appl. Phys. Lett., 94, 153109 (2009).
[18] G. Baffou, M. P. Kreuzer , F. Kulzer , and R. Quidant, Optics Express, 17, 3291 (2009).
[19] Ch. Girard, E. Dujardin, G. Baffou, and R. Quidant, New Journal of Physics, 10, 105016 (2008).
[20] M. Essone Mezeme and C Brosseau, Phys. Rev. E 87, 012722 (2013).
[21] G. Baffou, P. Berto, E. Bermudez Urena, R. Quidant, S. Monneret, J. Polleux, and H. Rigneault,
ACS Nano, 7, 6478 (2013).
[22] A. Teulle, R. Marty, Ch. Girard, A. Arbouet, and E. Dujardin, Opt. Communications, 291, 412
(2013).
[23] S. J. Weber, G. Colas des Francs, and Ch. Girard, Phys. Rev. B, 91, 205419 (2015).
[24] C. Enkrich, M. Wegener, S. Linden, S. Burger, L. Zschiedrich, F. Schmidt, J. F. Zhou, Th. Koschny,
and C. M. Soukoulis, Phys. Rev. Lett, 95, 203901 (2005).
[25] N Yu, P. Genevet, M. A. Kats, F. Aieta,J-Ph. Tetienne, F. Capasso, Z. Gaburro, Science, 334,
333 (2011).
[26] N. Lawrence, J. Trevino, and L. Dal Negro, Opt. Lett. 37, 5076 (2012).
[27] J. N. Anker, W. P. Hall, O. Lyandres, N. C. Shah, J. Zhao, and R. P. Van Duyne, Nat. Materials,
7, 442 (2008).
[28] Plasmonic Nanoparticle Networks, Erik Dujardin and Christian Girard, "Handbook of
Nanophysics" (K. Sattler, ed.), Taylor & Francis, London 2010, Chapter 27.
[29] P. B. Johnson and R. W. Christy, Phys. Rev. B 6, 4370 (1972).
[30] S. Link, M. B. Mohamed, and M. A. El -- Sayed, J. Phys. Chem. B, 103, 3073 (1999).
[31] L. D. Landau and E. M. Lifshitz, Field theory, Pergamon press, London, third edition (1960).
[32] K. Metwally, S. Mensah, and G. Baffou, ACS Photonics, 4, 1544 (2017).
[33] N. Yu and F. Capasso, Nat. Materials, 13, 139 (2014).
[34] L.-J. Black, Y. Wang, C. H. de Groot, A. Arbouet, and O. L. Muskens, ACS Nano, 8, 6390 (2014).
[35] H. Linnenbank and S. Linden, Optica, 2, 698 (2015).
[36] A. E. Minovitch, A. E. Miroshnichenko, A. Y. Bykov, T. V. Murzina, D. N. Neshev, and Y. S.
Designing Thermoplasmonic Properties of Metallic Metasurfaces
18
Kivshar, Laser Photonics Rev., 9, 195 (2015).
[37] L.-J. Black, P. R. Wiecha, Y. Wang, C. H. de Groot, V. Paillard, Ch. Girard, O. L. Muskens, and
A. Arbouet, ACS Photonics, 2, 1592 (2015).
[38] J. S. Donner, J. Morales-Dalmau, I. Alda, R. Marty, and R. Quidant, ACS Photonics, 2, 355
(2015).
[39] A. Cuche, O. Mahboub, E. Devaux, C. Genet, and T. W. Ebbesen, Phys. Rev. Lett, 108, 026801
(2012).
[40] I. Mori and Y. Oshima, Nature (London) 376, 344 (1995).
[41] F. N. Hamada, M. Rosenzweig, K. Kang, S. R. Pulver, A. Ghezzi, T. J. Jegla, and P. A. Garrity,
Nature (London) 454, 217 (2008).
[42] T. Feichtner, O. Selig, M. Kiunke, and B. Hecht, Phys. Rev. Lett, 109, 127701 (2012).
[43] C. Forestiere, Y. He, R. Wang, R. Kirby, and L. Dal Negro, ACS Photonics, 3, 68 -- 78 (2016).
[44] P. R. Wiecha, A. Arbouet, C. Girard, A. Lecestre, G. Larrieu, and V. Paillard, Nature
nanotechnology, 12 (2), 163 (2017).
[45] S. M. Islam, S. Das, S. Ghosh, S. Roy, and P. N. Suganthan, IEEE Transactions on Systems, Man,
and Cybernetics, Part B (Cybernetics), 42, 482 -- 500 (2012).
[46] F. Biscani, D. Izzo, and C. H. Yam, arXiv:1004.3824 [cs, math], (2010).
[47] P. R. Wiecha, arXiv:1802.04071 [cond-mat, physics:physics] (2018).
[48] E. D. Palik, Handbook of Optical Constants of Solids, Academic Press, 547 -- 569 (1997).
[49] C. Girard and R. Quidant, Optics Express, 12, 6141 -- 6146 (2004).
[50] R. M. Bakker, Y. F. Yu, R. Paniagua-Dom´ınguez, B. Luk'yanchuk, and A. I. Kuznetsov, Nano
Letters, 17, 3458 -- 3464 (2017).
[51] P. Wiecha, M.-M. Mennemanteuil, D. Khlopin, J. Martin, A. Arbouet, D. G´erard, A. Bouhelier,
J. Plain, and A. Cuche, Phys. Rev. B 96, 035440 (2017).
|
1807.07715 | 1 | 1807 | 2018-07-20T07:01:52 | Controlled Intracellular Delivery of Single Particles in Single Cells by 3D Hollow Nanoelectrodes | [
"physics.app-ph",
"physics.bio-ph",
"physics.optics"
] | We present an electrophoretic platform based on 3D hollow nanoelectrodes capable of controlling and quantifying the intracellular delivery of single nanoparticles in single selected cells by surface-enhanced Raman spectroscopy (SERS). The gold-coated hollow nanoelectrode has a sub-femtoliter inner volume that allows the confinement and enhancement of electromagnetic fields upon laser illumination to distinguish the SERS signals of a single nanoparticle flowing through the nanoelectrode. The tight wrapping of cell membranes around the nanoelectrodes enables effective membrane electroporation such that single gold nanorods are delivered into a living cell with a delivery rate subject to the applied bias voltage. The capability of the 3D hollow nanoelectrodes to porate cells and reveal single emitters from the background under live flow is promising for the analysis of both intracellular delivery and sampling. | physics.app-ph | physics | Controlled Intracellular Delivery of Single Particles in Single Cells by
3D Hollow Nanoelectrodes
Jian-An Huang a, Valeria Caprettini a,b, Yingqi Zhao a, Giovanni Melle a,b, Nicolò Maccaferri a,
Matteo Ardini a, Francesco Tantussi a, Michele Dipalo a, Francesco De Angelis a,*
a Istituto Italiano di Tecnologia, Via Morego 30, 16163 Genova, Italy
b DIBRIS, University of Genoa, Via all'Opera Pia 13, 16145 Genova, Italy
* [email protected].
Supporting Information included
Abstract: We present an electrophoretic platform based on
3D hollow nanoelectrodes capable of controlling and quanti-
fying the intracellular delivery of single nanoparticles in single
selected cells by surface-enhanced Raman spectroscopy
(SERS). The gold-coated hollow nanoelectrode has a sub-
femtoliter inner volume that allows the confinement and en-
hancement of electromagnetic fields upon laser illumination to
distinguish the SERS signals of a single nanoparticle flowing
through the nanoelectrode. The tight wrapping of cell mem-
branes around the nanoelectrodes enables effective mem-
brane electroporation such that single gold nanorods are
delivered into a living cell with a delivery rate subject to the
applied bias voltage. The capability of the 3D hollow
nanoelectrodes to porate cells and reveal single emitters from
the background under live flow is promising for the analysis of
both intracellular delivery and sampling.
Significance: The delivery of molecules into the intracellular
compartment is one of the fundamental requirements of the
current molecular biology. However, the possibility of deliver-
ing a precise number of nano-objects (nanoparticles, proteins,
genic materials) with single-particle resolution is still an open
challenge. Here, we show that single nano-objects can be
delivered into cells cultured in vitro by an electrophoretic ap-
proach based on 3D hollow nanoelectrodes combined with
surface-enhanced Raman scattering, which enables real-time
counting of individual nanoparticles. By a simple refinement of
the platform, it is possible to realize one nanoelectrode per
cell, which would be ideal for use in the emerging field of
single-cell technology and the on-chip analysis of the content
extracted from cells, including proteins, DNA and miRNA.
Figure 1. Schematic representation of the 3D hollow nanoelectrode device for single-particle intracellular delivery.
The cell is tightly wrapped around the gold-coated hollow nanoelectrode and is first electroporated by a pulsed volt-
age. Then, the nanorods originally in the cis chamber are delivered into the cell through the hollow nanoelectrode by
a DC potential between the two Pt wire electrodes. Inset: a laser excites the Raman signals of the delivered nanorods
for counting the number of delivered nanorods.
leads
Introduction
The intracellular delivery of nanoparticles, such as
quantum dots and gold nanoparticles, is widely used
in proteomics, drug delivery, and single-cell studies.1-
10 Nanoparticle endocytosis usually
to
nanoparticle aggregation in endosomal vesicles and
nanoparticle attachment to the cell membrane.11
However, these vesicles can prevent the trapped
nanoparticles from approaching targeted organelles
or molecules.12 Additionally, nanoparticle aggregates
much larger than single nanoparticles could distort
molecular behavior in mechanistic studies, such as
those of intracellular transport by motor proteins.13-20
Therefore, physical delivery methods for injecting
single particles into the cytoplasm of living cells are
highly desired.
Although many different approaches have been
developed for cells cultured in vitro,21 some important
limitations, such as poor dosage control, remain.22
Among them, the development of quantitative meth-
ods has remained difficult. In particular, the possibil-
ity of delivering a precise number of nano-objects
with single-particle resolution is an ongoing issue.
Moreover, the ability to target single selected cells
within a large population would be of additional
value in the emerging field of single-cell biology,
which aims to discover characteristics of individual
cells that are hidden in experiments performed using
large cell numbers.
In recent years, different physical methods based
on nanopores23, 24 and hollow nanotube systems (or
nanostraws)25-27 have been developed as reliable
means of delivery with high cell viability. However,
these methods have not overcome the aforemen-
tioned limitations.
In this work, we show that single nano-objects can
be delivered into cells cultured in vitro by an electro-
phoretic approach combined with surface-enhanced
Raman scattering, which enables the counting of in-
dividual nanoparticles in real time. The method is
based on plasmonic hollow nanotubes that simulta-
neously act as nanoelectrodes for cell electropora-
tion and particle delivery (nanochannels) and as
plasmonic antennas for Raman signal enhancement.
The concept is represented in Figure 1. Hollow nano-
tubes are fabricated on a Si3N4 substrate embedded
in a polydimethylsiloxane (PDMS) chamber to sepa-
rate a trans chamber from a cis chamber. The gold-
coated hollow nanotube interfacing with the cell
membrane acts as a nanoelectrode to generate elec-
tropores by a pulsed voltage. Then, DC potential is
applied to two Pt wire electrodes in both chambers to
deliver nanorods from the cis chamber to the electro-
porated cell through the hollow nanoelectrodes. The
optical energy can be confined inside the hollow
nanotube28 such that upon laser illumination, single
nanorods entering the nanoelectrodes can be well
distinguished from the background in the cis chamber.
Figure 2. SEM images of 3D hollow nanoelectrode
array on Si3N4. Inset: magnified SEM image of a
single nanoelectrode (a). The nanoelectrode has a
hollow volume and is 300 nm in diameter and 2 μm
in length (b). Simulated electromagnetic field inten-
sity distribution of an illuminated nanoelectrode (en-
hancement factor between 5 and 10) (c).
Results and Discussion
We first assessed the performance of the 3D hollow
nanoelectrodes by quantifying single-particle trans-
location through the hollow nanoelectrodes in phos-
phate-buffered saline without cells. Raman-tagged
gold nanorods 25 × 90 nm in size (Supporting Infor-
mation Figure S1) were used for electrophoretic trans-
location. Under 785-nm illumination, single nanorods
exhibited stable Raman spectra (Supporting Informa-
tion Figure S2) in which the signal-to-baseline inten-
sity of the Raman band at 593 cm-1 was used as the
signal for counting the translocated nanorods.
The hollow nanoelectrodes covered by a 30-nm-
thick gold layer had an inner diameter of 300 nm
and a length of 2 μm, as shown in Figure 2a and 2b.
When illuminated with a 785-nm laser, the electro-
magnetic field intensity was enhanced by a factor of
up to 10-fold the intensity of the incident field, as
shown in Figure 2c, and as previously demon-
strated.28
The inner volume of the nanotube was approxi-
mately 0.14 fL. However, the experimental detection
volume is expected to be even smaller because, as
shown in Figure 2c, the plasmonic field is accumu-
lated in a total volume that is smaller than that of the
nanotube. By collecting the Raman signal using an
objective with a high numerical aperture (NA = 1)
focused at the nanotube tip, we detected only the
nanorods translocating into the nanotube, whereas
the nanorods dispersed in solution in the cis chamber
contributed to a very low background noise level.
To allow a single nanorod to translocate through
the nanotube (i.e., to prevent the coincidence of two
particles in the same time window), the concentration
of nanorods dispersed in solution must be carefully
adjusted.29 We used Poisson statistics to calculate the
optimal concentration (see Methods for details).30
According to the calculations, a concentration of 1011
particles per mL leads to 0.014 nanorods diffusing
inside the hollow nanoelectrode, on average. The
probability of 0, 1 and 2 nanorods diffusing in the
hollow nanoelectrode was 0.986, 1.38 × 10-2 and
9.66 × 10-5, respectively. Since the probability of
having 2 nanorods passing through the nanochannel
simultaneously was very low (<10-4), we assumed
that the recorded Raman signal was always due to a
single nanorod.
Electrophoretic Translocation
DC voltages ranging from -0.5 to -2 V were used
for the electrophoretic translocation since the nano-
rods were negatively charged. The DC potential was
applied between two Pt wire electrodes that were
separated by a distance of approximately 15 mm in
the cis and trans chambers. Single-particle transloca-
tions were demonstrated as bursts in a time trace of
the intensity changes of the nanorod Raman band at
Figure 3. Electrophoretic translocation of nanorods through the hollow nanoelectrodes without cells. (a) Time traces of
electrophoretic translocation at a DC bias of -2 V (red curve) and 0 V (black curve, diffusion regime). Inset: schematic of
the electrophoretic translocation without cells. (b) Measured event rates of nanorod translocation; a burst with a signal-to-
noise intensity ratio no less than 3 was considered an event. (c) Probability of the coincidence of 2 nanorods in flow during
electrophoretic translocation.
593 cm-1, as shown in Figure 3a.
reports of single-particle
A burst with a signal-to-noise intensity ratio no less
than 3 was considered a single-particle translocation
event. In the case of spontaneous diffusion (no bias
applied), only 2 events occurred within 30 s. In con-
trast, the number of events increased significantly at
a bias of -2 V. Thus, the event rate increased with the
applied potential, indicating that more single nano-
rods were translocated in a certain time window
(Figure 3b). The average translocation time obtained
was 57 ms at a bias of -2 V (Supporting Information
Figure S3). Although the translocation time depended
on many parameters, such as the ratio of the nanorod
size to the nanoelectrode diameter, the translocation
time obtained is on the same order as those from
previous
translocation
through nanopores,31-34 confirming that the observed
events were related to single-particle translocation.
A critical parameter for ensuring single-particle
translocation is the low probability of the coincidence
of more than one particle in flow. Unlike the diffusion
case, the probability calculated by the Poisson statis-
tics for particles in flow35 considers the irreversible
flow of the nanoparticles with a flow rate (or event
rate) and an exposure time as the detection time
window (see Methods for details). At an exposure
time of 10 ms and a measured event rate of 32 per
minute, the probability of translocating 2 nanorods
simultaneously at a bias of -2 V was calculated as
1.46 × 10-5. The probabilities at other bias voltages
were even smaller, as shown in Figure 3c. Thus, the
electrophoretic translocation of a single nanorod was
confirmed as the most likely cause for event detection.
Compared with particle diffusion, electrophoresis
prevents translocated nanorods from returning to the
nanoelectrodes. Moreover, this approach provides
more cognizant control of the translocation rate. Here,
the electrophoretic voltage of our hollow nanoelec-
trode system was optimized with the nanorod con-
centration for the efficient intracellular delivery of
single nanorods, as shown below.
Intracellular Delivery
To demonstrate intracellular delivery, NIH-3T3 cells
were cultured in the trans chamber to allow cell
growth on the hollow nanoelectrodes with tight
membrane wrapping (Figure 4). Together with two Pt
wire electrodes for translocating the nanorods, a ca-
Figure 4. Cross-sectional SEM image of a cell cul-
tured on the nanoelectrodes (a). A magnified SEM
image showing that the cell membrane is tightly
wrapped around the nanoelectrode (b).
ble was connected to the gold layer of the hollow
nanoelectrodes for cell membrane electroporation.
The membrane was porated by applying a peak-to-
peak pulsed voltage of 3 V for 10 s with pulse length
of 100 μs and a frequency of 20 Hz between the Pt
wire electrode in the trans chamber and the hollow
nanoelectrodes. After the electropores were gener-
ated in the cell membrane, electrophoretic delivery
of the nanorods was conducted with DC voltage (-
0.5 to -2 V) between the two Pt wire electrodes in
the trans and cis chambers. Gold nanorods with 10 x
40 nm in size were used to facilitate delivery through
the small electropores. A time trace of the electropo-
ration and delivery exhibited delivery events at a
bias of -2 V with an event rate of 4 min-1 after mem-
brane electroporation, as shown in Figure 5a.
That fact that no bursts were observed from a bias
of -0.5 to -1.5 V suggests that the electroporated cell
membrane presented many barriers to the electro-
phoretic delivery. One such barrier could be that the
resistance of the cell membrane decreased the elec-
trophoretic voltage, even if the cell membrane was
electroporated.36 In addition, the transient electro-
pores continuously shrink after electroporation,37, 38
leading to a low event rate. Such an event rate corre-
sponds to the probability of coincidence in flow as
small as 10-7, ensuring single-particle intracellular
delivery.
To assess the nanorod delivery, we examined the
cells by Raman mapping and analyzing the Raman
band at 593 cm-1. The distribution and transport of
the nanorods are shown in merged images in Figure
5b and 5c, in which the colored dots indicate the
signal-to-baseline Raman intensity of the nanorods.
In previous reports on the endocytic uptake of
nanoparticles, the Raman signals of intracellular
nanoparticles were colocalized with black dots in
bright-field optical images that corresponded to in-
travesicular nanoparticle aggregates.39, 40 Our case
is in strong contrast with these reports because no
black dots were observed in the bright-field images
overlapping the colored dots. Since it takes at least 2
hours for motor proteins to capture and aggregate
single intracellular nanoparticles,41, 42 the colored
dots should correspond to single nanorods, which
could not be resolved by the bright-field 60× objec-
tive.
The number of colored dots in the Raman maps that
represent delivered nanorods can hardly be equal to
the number of bursts of the time trace. On the one
hand, the Raman mapping was performed spot-by-
spot by mechanically moving the sample stage. As
each spot required 300 ms to map, the whole map-
ping process was usually completed in 3 to 5 minutes,
which was too slow to trace the real-time distribution
of the nanorods. On the other hand, the delivered
nanorods could have moved from the focal plane to
the upper interior of the cell during the Raman map-
ping. Nevertheless, the temporal limitation due to the
stage-scanning Raman microscope can be readily
overcome using a laser-scanning Raman micro-
scope.43, 44
The continuous colocalization of the colored dots
with some nanoelectrode positions in the Raman
maps suggests that the nanorods were accumulated
inside the nanoelectrodes. This accumulation could
have occurred because the electropores were too
small to allow nanorod passage. When the accumu-
lated nanorods were clogged inside the nanoelec-
trode, they could be moved back into the cis cham-
ber by applying a positive DC bias (Supporting In-
formation Figure S4). This control of reversible nano-
rod movement through the nanoelectrodes could be
a way for the hollow nanoelectrodes to extract ob-
jects from porated cells and perhaps be applied for
intracellular sampling.45
Figure 5. A time trace of the electrophoretic intracel-
lular delivery of nanorods at a bias of -2 V after
electroporation (a). Optical images merged with
Raman maps of the nanorods. White dotted circles
are the locations of the nanoelectrodes. White ar-
rows indicate the nanorods delivered into one cell.
The colored dots represent the Raman intensity of the
nanorods, as indicated by the colored bars. Intracel-
lular delivery of the nanorods after 5 min (b) and 10
min (c). All scale bars are 10 μm.
Conclusion
We demonstrated the electrophoretic intracellular
delivery of nanorods with the capability of controlling
and detecting single events, thus providing a method
for accurate quantitative delivery. The platform is
based on multifunctional plasmonic hollow nano-
tubes that can work as i) nanoelectrodes for cell
electroporation, ii) nanochannels for electrophoretic
delivery, and iii) plasmonic antennas to enhance the
optical signals of nano-objects translocating through
the channel. The tight wrapping of the cell mem-
brane around the nanoelectrodes allowed the gen-
eration of electropores large enough to allow nano-
rod passage while still preserving the membrane ad-
hesion after electroporation and preventing nanorod
leakage. By a simple refinement of the platform, it is
possible to realize one nanoelectrode per cell for the
discrimination of cells that have and have not re-
ceived a nano-object, which would be ideal for use
in the emerging field of single-cell technology. Finally,
as hollow nanotubes have been demonstrated to ex-
tract cytosolic context from living cells,45 in the future,
such a platform could be used for the real-time moni-
toring and on-chip analysis of the content extracted
from cells, including proteins, DNA and miRNA.46
Methods
Materials. Raman-tagged gold nanorods dis-
persed in deionized water were purchased from
Nanopartz Inc. (Loveland, CO, USA) with Nile blue A
(NBA) as the Raman reporter and stabilized by car-
boxyl groups. The gold nanorods were either 10 × 40
nm or 25 × 90 nm in size, with transverse plasmonic
resonance at a wavelength of 510 nm and longitudi-
nal plasmonic resonance at a wavelength of 780 nm.
The zeta potential at pH = 7 and the concentration of
the 10 × 40-nm gold nanorods was -18 mV and 4 ×
1013 particles per mL, respectively, and that of the 25
× 90-nm nanorods was -15 mV and 4 × 1012 parti-
cles per mL, respectively.
Device fabrication. To fabricate the 3D hollow
nanoelectrodes, S1813 photoresist (Shipley) was spin-
coated on a 1 x 1-cm Si3N4 membrane at 4000 rpm
for 1 min and soft baked at 95°C for 5 min. After
sputtering a 7-nm-thick titanium and a 20-nm-thick
gold layer on the back of the Si3N4 membrane, fo-
cused ion beam milling (FIB, FEI Helios NanoLab 650
DualBbeam) at a voltage of 30 keV and a current
from 0.23 to 2.5 nA was used to drill hole arrays in
the back of the Ti/Au-coated Si3N4 sample. Different
FIB currents correspond to different nanotube inner
diameters. Then, the sample was ashed by oxygen
plasma at 100 W for 2 min to smooth the photoresist
and was then developed in acetone for 2 min to form
polymer nanotube arrays. Then, the nanotube arrays
were thinned down by oxygen plasma at 100 W for
2 min. An alumina layer of 5 nm was deposited on
the back of the sample by atomic layer deposition
(Oxford Instruments) to neutralize the surface charge.
After being coated with a 7-nm-thick Ti layer and a
30-nm-thick gold layer by sputtering at a 45° tilt an-
gle with rotation to ensure uniform coating, the sam-
ple was annealed on a hot plate at 200°C in the air
for 1 hour and allowed to cool naturally. The as-
made nanoelectrodes were attached with a cable by
silver paste and embedded in a microfluidic chamber
made from polydimethylsiloxane (PDMS, Dow Corn-
ing SYLGARD 184 silicone elastomer) at 60°C for
approximately 40 min.
Cell culture. We used NIH-3T3 cells for the deliv-
ery experiments. Before seeding the cells, the devices
were irradiated with UV rays for 30 min in a laminar-
flow hood to sterilize them. The devices were treated
O.N. with complete DMEM to saturate the PDMS
chamber. Then, NIH-3T3 cells were seeded on the
devices at a concentration of 0.8 × 104 cells/cm2 and
incubated at 37°C in a 5% CO2 atmosphere for 24 h
in DMEM with 1% pen/strep antibiotic and 10% fetal
bovine serum (Sigma Aldrich) before the experiments
were performed.
Raman measurements. Raman measurements were
obtained by a Renishaw inVia Raman spectrometer
with a Nikon 60 × water immersion objective with a
1.0 NA delivering a 785-nm laser with a power of
approximately 3.27 mW. Intracellular nanoparticle
delivery was measured using an Andor EMCCD
camera (DU970P-BVF) integrated into the Renishaw
spectrometer with an exposure time of 10 ms. Cell
mapping was conducted with the Renishaw CCD
camera at an exposure time of 300 ms and a step of
1 μm.
Poisson statistics for particle diffusion. When an
average of <N> nanoparticles are diffusing in a
given volume, the probability of having m nanoparti-
cles at any time in the volume can be calculated by
the Poisson statistics for diffusion:30
Photostable Biological Probes. Journal of the American Chemical
Society 2013, 135 (22), 8350-8356.
5.
He, Y.; Kang, Z.-H.; Li, Q.-S.; Tsang, C. H. A.; Fan, C.-H.;
Lee, S.-T., Ultrastable, Highly Fluorescent, and Water-Dispersed
Silicon-Based Nanospheres as Cellular Probes. Angewandte Che-
mie-International Edition 2009, 48 (1), 128-132.
Poisson statistics
(1)
flow. When
nanoparticles are in flow in a given volume at a rate
of c, the probability of having n nanoparticles in the
volume at time Δt can be calculated by the Poisson
statistics in flow:35
for particles
in
(2)
Acknowledgments
We thank Dr. Xavier Zambrana Puyalto and Dr.
Rosario Capozza for their valuable discussions. The
research leading to these results was funded by the
European Research Council under the European Un-
ion's Seventh Framework Programme
(FP/2007-
2013)/ERC Grant Agreement no. [616213] and CoG:
Neuro-Plasmonics.
Author Contributions
F.D.A. conceived and supervised the work. J.A.H.
and M.A. prepared the nanoparticles. J.A.H., V.C.
and Y.Z. designed and fabricated the devices. V.C.
and G.M. cultured the cells. J.A.H., V.C., F.T. and
M.D. designed the optical-electrical setup. J.A.H.
and V.C. performed the Raman measurements. J.A.H.
and Y.Z. analyzed the data. N.M. performed the
electromagnetic simulations. All authors contributed
to the manuscript preparation.
References
1.
Doane, T. L.; Burda, C., The unique role of nanoparticles in
nanomedicine: imaging, drug delivery and therapy. Chemical Society
Reviews 2012, 41 (7), 2885-2911.
2.
Kang, B.; Afifi, M. M.; Austin, L. A.; El-Sayed, M. A., Ex-
ploiting the Nanoparticle Plasmon Effect: Observing Drug Delivery
Dynamics in Single Cells via Raman/Fluorescence Imaging Spectros-
copy. Acs Nano 2013, 7 (8), 7420-7427.
3.
Zheng, X. T.; Li, C. M., Single cell analysis at the nano-
scale. Chemical Society Reviews 2012, 41 (6), 2061-2071.
4.
Zhong, Y.; Peng, F.; Bao, F.; Wang, S.; Ji, X.; Yang, L.; Su,
Y.; Lee, S.-T.; He, Y., Large-Scale Aqueous Synthesis of Fluorescent
and Biocompatible Silicon Nanoparticles and Their Use as Highly
6.
Zimmerman, J. F.; Parameswaran, R.; Murray, G.; Wang,
Y.; Burke, M.; Tian, B., Cellular uptake and dynamics of unlabeled
freestanding silicon nanowires. Science Advances 2016, 2 (12).
7.
Leijten, J.; Khademhosseini, A., From Nano to Macro: Mul-
tiscale Materials for Improved Stem Cell Culturing and Analysis. Cell
Stem Cell 2016, 18 (1), 20-24.
8.
Xi, W.; Schmidt, C. K.; Sanchez, S.; Gracias, D. H.;
Carazo-Salas, R. E.; Jackson, S. P.; Schmidt, O. G., Rolled-up Func-
tionalized Nanomembranes as Three-Dimensional Cavities for Single
Cell Studies. Nano Letters 2014, 14 (8), 4197-4204.
9.
Hartmann, R.; Weidenbach, M.; Neubauer, M.; Fery, A.;
Parak, W. J., Stiffness-Dependent In Vitro Uptake and Lysosomal
Acidification of Colloidal Particles. Angewandte Chemie-International
Edition 2015, 54 (4), 1365-1368.
10.
Chen, X.; Zhang, W., Diamond nanostructures for drug de-
livery, bioimaging, and biosensing. Chemical Society Reviews 2017,
46 (3), 734-760.
11.
Chou, L. Y. T.; Ming, K.; Chan, W. C. W., Strategies for
the intracellular delivery of nanoparticles. Chemical Society Reviews
2011, 40 (1), 233-245.
12.
Gilleron, J.; Querbes, W.; Zeigerer, A.; Borodovsky, A.;
Marsico, G.; Schubert, U.; Manygoats, K.; Seifert, S.; Andree, C.;
Stoeter, M.; Epstein-Barash, H.; Zhang, L.; Koteliansky, V.; Fitzgerald,
K.; Fava, E.; Bickle, M.; Kalaidzidis, Y.; Akinc, A.; Maier, M.; Zerial,
M., Image-based analysis of lipid nanoparticle-mediated siRNA
delivery, intracellular trafficking and endosomal escape. Nature Bio-
technology 2013, 31 (7), 638-U102.
13.
Cui, B. X.; Wu, C. B.; Chen, L.; Ramirez, A.; Bearer, E. L.;
Li, W. P.; Mobley, W. C.; Chu, S., One at a time, live tracking of
NGF axonal transport using quantum dots. Proceedings of the Na-
tional Academy of Sciences of the United States of America 2007,
104 (34), 13666-13671.
14.
Zhang, K.; Osakada, Y.; Vrljic, M.; Chen, L. A.; Mudra-
kola, H. V.; Cui, B. X., Single-molecule imaging of NGF axonal
transport in microfluidic devices. Lab on a Chip 2010, 10 (19), 2566-
2573.
15.
Courty, S.; Luccardini, C.; Bellaiche, Y.; Cappello, G.; Da-
han, M., Tracking individual kinesin motors in living cells using single
quantum-dot imaging. Nano Letters 2006, 6 (7), 1491-1495.
16.
Pinaud, F.; Clarke, S.; Sittner, A.; Dahan, M., Probing cel-
lular events, one quantum dot at a time. Nature Methods 2010, 7 (4),
275-285.
17.
Leduc, C.; Si, S.; Gautier, J. J.; Gao, Z. H.; Shibu, E. S.;
Gautreau, A.; Giannone, G.; Cognet, L.; Lounis, B., Single-molecule
imaging in live cell using gold nanoparticles. In Biophysical Methods
in Cell Biology, Paluch, E. K., Ed. 2015; Vol. 125, pp 13-27.
18.
Cognet, L.; Tardin, C.; Boyer, D.; Choquet, D.; Tamarat, P.;
Lounis, B., Single metallic nanoparticle imaging for protein detection
in cells. Proceedings of the National Academy of Sciences of the
United States of America 2003, 100 (20), 11350-11355.
19.
Li, M.; Lohmuller, T.; Feldmann, J., Optical Injection of
Gold Nanoparticles into Living Cells. Nano Letters 2015, 15 (1), 770-
775.
20.
Kang, J. W.; So, P. T. C.; Dasari, R. R.; Lim, D. K., High
Resolution Live Cell Raman Imaging Using Subcellular Organelle-
Targeting SERS-Sensitive Gold Nanoparticles with Highly Narrow
Intra-Nanogap. Nano Letters 2015, 15 (3), 1766-1772.
21.
Yan, L.; Zhang, J. F.; Lee, C. S.; Chen, X. F., Micro- and
Nanotechnologies for Intracellular Delivery. Small 2014, 10 (22),
4487-4504.
22.
Stewart, M. P.; Sharei, A.; Ding, X. Y.; Sahay, G.; Langer,
R.; Jensen, K. F., In vitro and ex vivo strategies for intracellular deliv-
ery. Nature 2016, 538 (7624), 183-192.
23.
Kurz, V.; Tanaka, T.; Timp, G., Single Cell Transfection
with Single Molecule Resolution Using a Synthetic Nanopore. Nano
Letters 2014, 14 (2), 604-611.
24.
Chang, L. Q.; Bertani, P.; Gallego-Perez, D.; Yang, Z. G.;
Chen, F.; Chiang, C. L.; Malkoc, V.; Kuang, T. R.; Gao, K. L.; Lee, L.
J.; Lu, W., 3D nanochannel electroporation for high-throughput cell
transfection with high uniformity and dosage control. Nanoscale 2016,
8 (1), 243-252.
25. Messina, G. C.; Dipalo, M.; La Rocca, R.; Zilio, P.; Capret-
tini, V.; Zaccaria, R. P.; Toma, A.; Tantussi, F.; Berdondini, L.; De
Angelis, F., Spatially, Temporally, and Quantitatively Controlled
Delivery of Broad Range of Molecules into Selected Cells through
Plasmonic Nanotubes. Advanced Materials 2015, 27 (44), 7145-+.
26.
Caprettini, V.; Cerea, A.; Melle, G.; Lovato, L.; Capozza,
R.; Huang, J.-A.; Tantussi, F.; Dipalo, M.; De Angelis, F., Soft elec-
troporation for delivering molecules into tightly adherent mammalian
cells through 3D hollow nanoelectrodes. Scientific Reports 2017, 7.
27.
Xie, X.; Xu, A. M.; Leal-Ortiz, S.; Cao, Y. H.; Garner, C.
C.; Melosh, N. A., Nanostraw-Electroporation System for Highly
Efficient Intracellular Delivery and Transfection. Acs Nano 2013, 7 (5),
4351-4358.
28. De Angelis, F.; Malerba, M.; Patrini, M.; Miele, E.; Das, G.;
Toma, A.; Zaccaria, R. P.; Di Fabrizio, E., 3D Hollow Nanostructures
as Building Blocks for Multifunctional Plasmonics. Nano Letters 2013,
13 (8), 3553-3558.
29. Goddard, G.; Brown, L. O.; Habbersett, R.; Brady, C. I.;
Martin, J. C.; Graves, S. W.; Freyer, J. P.; Doorn, S. K., High-
Resolution Spectral Analysis of Individual SERS-Active Nanoparticles
in Flow. Journal of the American Chemical Society 2010, 132 (17),
6081-6090.
30.
Hill, E. K.; de Mello, A. J., Single-molecule detection using
confocal fluorescence detection: Assessment of optical probe volumes.
Analyst 2000, 125 (6), 1033-1036.
31.
Liu, L.; Kong, J.; Xie, X.; Wu, H.; Ye, X.; Zhao, Z.; Wang,
translocation through a solid-state
L.; Liu, Q., Gold nanorod
nanopore. Chinese Science Bulletin 2014, 59 (7), 598-605.
32.
Angeli, E.; Volpe, A.; Fanzio, P.; Repetto, L.; Firpo, G.;
Guida, P.; Lo Savio, R.; Wanunu, M.; Valbusa, U., Simultaneous
Electro-Optical Tracking for Nanoparticle Recognition and Counting.
Nano Letters 2015, 15 (9), 5696-5701.
33. Cecchini, M. P.; Wiener, A.; Turek, V. A.; Chon, H.; Lee, S.;
Ivanov, A. P.; McComb, D. W.; Choo, J.; Albrecht, T.; Maier, S. A.;
Edel, J. B., Rapid Ultrasensitive Single Particle Surface-Enhanced
Raman Spectroscopy Using Metallic Nanopores. Nano Letters 2013,
13 (10), 4602-4609.
34. Qiu, Y. H.; Siwy, Z., Probing charges on solid-liquid inter-
faces with the resistive-pulse technique. Nanoscale 2017, 9 (36),
13527-13537.
35.
Keij, J. F.; Vanrotterdam, A.; Groenewegen, A. C.;
Stokdijk, W.; Visser, J. W. M., COINCIDENCE IN HIGH-SPEED
FLOW-CYTOMETRY - MODELS AND MEASUREMENTS. Cytometry
1991, 12 (5), 398-404.
36.
Boukany, P. E.; Morss, A.; Liao, W.-c.; Henslee, B.; Jung,
H.; Zhang, X.; Yu, B.; Wang, X.; Wu, Y.; Li, L.; Gao, K.; Hu, X.; Zhao,
X.; Hemminger, O.; Lu, W.; Lafyatis, G. P.; Lee, L. J., Nanochannel
electroporation delivers precise amounts of biomolecules into living
cells. Nature Nanotechnology 2011, 6 (11), 747-754.
37. Demiryurek, Y.; Nickaeen, M.; Zheng, M. D.; Yu, M.; Zahn,
J. D.; Shreiber, D. I.; Lin, H.; Shan, J. W., Transport, resealing, and
re-poration dynamics of two-pulse electroporation-mediated mo-
lecular delivery. Biochimica Et Biophysica Acta-Biomembranes 2015,
1848 (8), 1706-1714.
38.
Teissie, J.; Golzio, M.; Rols, M. P., Mechanisms of cell
membrane electropermeabilization: A minireview of our present (lack
of ?) knowledge. Biochimica Et Biophysica Acta-General Subjects
2005, 1724 (3), 270-280.
39.
Huefner, A.; Kuan, W. L.; Muller, K. H.; Skepper, J. N.;
Barker, R. A.; Mahajan, S., Characterization and Visualization of
Vesicles in the Endo-Lysosomal Pathway with Surface-Enhanced
Raman Spectroscopy and Chemometrics. Acs Nano 2016, 10 (1),
307-316.
40.
Huefner, A.; Kuan, W. L.; Barker, R. A.; Mahajan, S., Intra-
cellular SERS Nanoprobes For Distinction Of Different Neuronal Cell
Types. Nano Letters 2013, 13 (6), 2463-2470.
41.
Sirimuthu, N. M. S.; Syme, C. D.; Cooper, J. M., Investiga-
tion of the stability of labelled nanoparticles for SE(R)RS measurements
in cells. Chemical Communications 2011, 47 (14), 4099-4101.
42.
Liu, M.; Li, Q.; Liang, L.; Li, J.; Wang, K.; Li, J.; Lv, M.;
Chen, N.; Song, H.; Lee, J.; Shi, J.; Wang, L.; Lal, R.; Fan, C., Real-
time visualization of clustering and intracellular transport of gold
nanoparticles by correlative imaging. Nature Communications 2017,
8.
43.
Bohndiek, S. E.; Wagadarikar, A.; Zavaleta, C. L.; Van de
Sompel, D.; Garai, E.; Jokerst, J. V.; Yazdanfar, S.; Gambhir, S. S., A
small animal Raman instrument for rapid, wide-area, spectroscopic
imaging. Proceedings of the National Academy of Sciences of the
United States of America 2013, 110 (30), 12408-12413.
44.
Kang, J. W.; Nguyen, F. T.; Lue, N.; Dasari, R. R.; Heller, D.
A., Measuring Uptake Dynamics of Multiple Identifiable Carbon
Nanotube Species via High-Speed Confocal Raman Imaging of Live
Cells. Nano Letters 2012, 12 (12), 6170-6174.
45.
Cao, Y.; Hjort, M.; Chen, H.; Birey, F.; Leal-Ortiz, S. A.;
Han, C. M.; Santiago, J. G.; Pasca, S. P.; Wu, J. C.; Melosh, N. A.,
Nondestructive nanostraw intracellular sampling for longitudinal cell
monitoring. Proceedings of the National Academy of Sciences of the
United States of America 2017, 114 (10), E1866-E1874.
46.
Higgins, S. G.; Stevens, M. M., Extracting the contents of
living cells. Science 2017, 356 (6336), 379-380.
Supporting Information
Controlled Intracellular Delivery of Single Particles
in Single Cells by 3D Hollow Nanoelectrodes
Jian-An Huang a, Valeria Caprettini a,b, Yingqi Zhao a, Giovanni Melle a,b, Nicolò Maccaferri a, Matteo
Ardinia, Francesco Tantussi a, Michele Dipalo a, Francesco De Angelis a,*
a Istituto Italiano di Tecnologia, Via Morego 30, 16163 Genova, Italy
b DIBRIS, University of Genoa, Via all'Opera Pia 13, 16145 Genova, Italy
* [email protected]
Supplementary Figures
Figure S1. SEM images of the Raman-tagged gold nanorods with sizes of 25 × 90 nm (left) and 10 × 40
nm (right).
10
Figure S2. Raman spectrum of the Raman tag, Nile Blue A, adsorbed on the gold nanorods in which the
593 cm-1 and 633 cm-1 bands are selected for evaluation of the nanorod aggregation (Inset). Time
traces of signal-to-baseline intensity of different Raman peaks and baseline (700 cm-1) of 0.1 nanorod
on average diffusing in a detection volume (Φ 1.2 × 5 µm) of the 60 × water immersion objective with
N.A. = 1.0. (a). The corresponding correlation of the Raman peaks between 593 cm-1 and 663 cm-1 (b)
or between 593 cm-1 and the baseline at 700 cm-1 (c) in which the R2 is the coefficient of correlation
and <N> is the average number of nanoparticle in the detection volume. The high correlations between
the Raman peaks or baseline indicate that no nanorod aggregation exists and the spectra of single nano-
rod in flow are stable. 1, 2
11
Figure S3. Time traces of the nanorods (25 × 90 nm) translocating through the nanoelectrodes without
cells under different electrophoresis bias voltage; Inset is the definition of the translocation time of a
burst (a). Histograms of the corresponding translocation times fitted by lognormal probability functions
(red curves) to extract average translocation times: 77 ms at -1.5 V bias (b) and 57 ms at -2 V bias (c),
respectively. The SEM image in the Inset of (b) indicates the translocated nanorods. The bursts in the
histograms are selected with thresholds: intensity signal-to-noise ratio no less than 3 and the transloca-
tion times limited from 10 to 500 ms.
12
Figure S4. The time trace of nanorods (25 × 90 nm) clogging in a nanoelectrode. The nanorods were
repelled by positive bias electrophoresis at different voltages. The fact that intensity decreased to near
zero suggested that the nanorods were cleared out from the nanoelectrodes.
References
1.
K., High‐Resolution Spectral Analysis of Individual SERS‐Active Nanoparticles in Flow. Journal of the American
Chemical Society 2010, 132 (17), 6081‐6090.
2.
Solution‐Based Characterization of Optimized SERS Nanoparticle Substrates. Journal of the American Chemical
Society 2009, 131 (1), 162‐169.
Goddard, G.; Brown, L. O.; Habbersett, R.; Brady, C. I.; Martin, J. C.; Graves, S. W.; Freyer, J. P.; Doorn, S.
Laurence, T. A.; Braun, G.; Talley, C.; Schwartzberg, A.; Moskovits, M.; Reich, N.; Huser, T., Rapid,
13
|
1901.01387 | 1 | 1901 | 2019-01-05T09:17:27 | Chiral plasmonic nanocrystals for generation of hot electrons: towards polarization-sensitive photochemistry | [
"physics.app-ph",
"cond-mat.mtrl-sci",
"physics.chem-ph"
] | The use of biomaterials - with techniques such as DNA-directed assembly or bio-directed synthesis - can surpass top-down fabrication techniques in creating plasmonic superstructures, in terms of spatial resolution, range of functionality and fabrication speed. Particularly, by enabling a very precise placement of nanoparticles in a bio-assembled complex or a controlled bio-directed shaping of single nanoparticles, plasmonic nanocrystals can show remarkably strong circular dichroism (CD) signals. Here we show that chiral bio-plasmonic assemblies and nanocrystals can enable polarization-sensitive photochemistry based on the generation of energetic (hot) electrons. It is now established that hot plasmonic electrons can induce surface photochemistry or even reshape plasmonic nanocrystals. Here we show that merging chiral plasmonic nanocrystal systems and the hot-election generation effect offers unique possibilities in photochemistry - such as polarization-sensitive photochemistry promoting nonchiral molecular reactions, chiral photo-induced growth of a colloid at the atomic level and chiral photochemical destruction of chiral nanocrystals. Regarding practical applications, our study suggests interesting opportunities in polarization-sensitive photochemistry, chiral recognition or separation, and in promoting chiral crystal growth at the nanoscale. | physics.app-ph | physics | Chiral plasmonic nanocrystals for generation of hot
electrons: towards polarization-sensitive photochemistry
Tianji Liu,1,2 † Lucas V. Besteiro,1,3 † Tim Liedl4, Miguel A. Correa-Duarte5,
Zhiming Wang,*1 and Alexander Govorov*1,2
1Institute of Fundamental and Frontier Sciences, University of Electronic Science and Technology
of China, Chengdu 610054, China
2Department of Physics and Astronomy, Ohio University, Athens, Ohio 45701, United States
3 Centre Énergie Matériaux et Télécommunications, Institut National de la Recherche Scientifique,
1650 Boul. Lionel Boulet, Varennes, QC J3X 1S2, Canada
4Fakult𝑎(cid:4663) t f𝑢(cid:4663) r Physik and Center for Nanoscience, Ludwig-Maximilians-Universt𝑎(cid:4663) t M𝑢(cid:4663) nchen,
Geschwister-Scholl-Platz 1, 80539 Munich, Germany
5Department of Physical Chemistry, Center for Biomedical Research (CINBIO), Southern Galicia
Institute of Health Research (IISGS), and Biomedical Research, Networking Center for Mental
Health (CIBERSAM), Universidade de Vigo, 36310 Vigo, Spain
† Equal contributors
1
Abstract: The use of biomaterials - with techniques such as DNA-directed assembly or bio-
directed synthesis - can surpass top-down fabrication techniques in creating plasmonic
superstructures, in terms of spatial resolution, range of functionality and fabrication speed.
Particularly, by enabling a very precise placement of nanoparticles in a bio-assembled complex or
a controlled bio-directed shaping of single nanoparticles, plasmonic nanocrystals can show
remarkably strong circular dichroism (CD) signals. Here we show that chiral bio-plasmonic
assemblies can enable polarization-sensitive photochemistry based on the generation of energetic
(hot) electrons. It is now established that hot plasmonic electrons can induce surface
photochemistry or even reshape plasmonic nanocrystals. Here we show that merging chiral
plasmonic nanocrystal systems and the hot-election generation effect offers unique possibilities in
photochemistry - such as polarization-sensitive photochemistry promoting nonchiral molecular
reactions, chiral photo-induced growth of a colloid at the atomic level and chiral photochemical
destruction of chiral nanocrystals. In contrast, for chiral molecular systems the equivalent of the
described effects is challenging to be observed because molecular species exhibit typically very
small CD signals. Moreover, we compare our findings with traditional chiral photochemistry at
the molecular level, identifying new, different regimes for chiral photochemistry with possibilities
that are unique for plasmonic colloidal systems. In this study, we bring together the concept of
hot-electron generation and the field of chiral colloidal plasmonics. Using chiral plasmonic
nanorod complexes as a model system, we demonstrate remarkably strong CD in both optical
extinction and generation rates of hot electrons. Studying the regime of steady-state excitation, we
discuss the influence of geometrical and material parameters on the chiral effects involved in the
generation of hot electrons. Optical chirality and the chiral hot-electron response in the nanorod
dimers result from complex inter-particle interactions, which can appear in the weak coupling
2
regime or in the form of Rabi splitting. Regarding practical applications, our study suggests
interesting opportunities in polarization-sensitive photochemistry, chiral recognition or separation,
and in promoting chiral crystal growth at the nanoscale.
KEYWORDS: hot electrons, chiral plasmonics, circular dichroism, photochemistry
Introduction. In materials with a large number of mobile electrons, such as metals, these charge
carriers can be displaced by external electric fields, and we refer to the resonant modes excited
in such a way as plasmons. The field of plasmonics has grown rapidly in the past decades,
motivated by the capabilities that they afford us in manipulating light in the nanoscale. By
constructing plasmonic systems in the nanometer scale we are, effectively, employing antennas
that couple strongly with electromagnetic radiation at frequencies up to the UV spectral range 1
and localize its radiant energy. In doing so we can manipulate the propagation of light at this
scale and create novel optical effects, 2 -- 6 enhance secondary such as Raman scattering 7,8 or
efficiently promote the conversion of light into other forms of energy. 9 -- 11
Among the topics where the use of plasmonic nanoparticles has extended our scientific
purview has been the study of optical chirality. Plasmonic nanoparticles and assemblies have
proven useful not only in facilitating strategies to measure the chirality of molecular analytes by
enhancing and shifting their optical circular dichroism (CD) signal,12 but also in creating artificial
chiral systems with a strong and controllable differential optical response to left and right
3
circularly polarized light (CPL). This latter aspect has been naturally developed in the creation
of patterned surfaces through top-down fabrication methods,13 -- 17 but self-assembly fabrication
techniques have also opened the door to the creation of complex colloidal plasmonic bio-
assemblies with a tailored chiral response.13,18 -- 25
Regarding the localization and conversion of radiant energy, plasmonic nanoparticles have
been used in a variety of ways, including the enhancement of photovoltaic devices through
different mechanisms, 10,14,26,27 driving photocatalytic processes 28 -- 38 or heating in the nanoscale.
17,39 -- 42 All of these disparate applications take advantage of different physical processes, but they
have in common their exploitation of a fundamental property of plasmonic excitation, the
enhancement of the electric field inside and around the plasmonic nanostructures. Here we would
like to highlight in particular the application of plasmonic enhancement to act as a photocatalyst,
driving chemical reactions of current scientific and technological relevance such as water
splitting, 43 H2 dissociation 28 or CO2 reduction. 44 Two ways in which plasmonic nanoparticles
can serve that purpose is by providing additional energy for the reaction by heating their
surrounding medium, 17,39 -- 42 and to generate and donate excited (hot) charge carriers to the
reacting species. 28 -- 38 The injection of hot electrons (HEs) as a photocatalytic process is very
favorable from an energetic standpoint, because the excitation energy required to transfer an
electron from the metal Fermi sea to the molecular excited state is lower than direct optical
transitions in the molecule, which extends the usable spectral range of solar radiation.
Furthermore, HE generation can be boosted through techniques that enhance the induced field in
the plasmonic nanoparticle, such as the creation of hot spots. 45 -- 49
The confluence of the fields of chiral engineering and energy conversion has the potential to
develop new scientific and technological opportunities, both in terms of allowing us additional
4
control parameters over plasmonic-driven processes and outlining new detection techniques for
nanoscale chirality. The current literature shows successful attempts in using planar chiral
metamaterials in connection with HE injection in semiconductors, 14,15 photochemistry 16 and
photoheating. 17,40
In this manuscript we aim to extend the research on of hybrid nanomaterials into a new
direction, by proposing the concept of polarization-sensitive photochemistry utilizing generation
of hot electrons in chiral plasmonic nanocrystals (NCs). In our calculations, based on realistic
models, we predict that chiral plasmonic NCs are able to create a strong CD effect for the rates
of generation of HE. Since HEs are able to induce chemical reactions in a matrix 50,51 and on the
surface of a NC, 32,52,53 we predict that the strong CD effect for the HE generation in our system
can lead to related CD effects in a variety of photochemical reactions. Such HE-induced
photochemical reactions have been observed in several plasmonic systems. 32,50 -- 53
As a model system for the chiral HE generation effect, we take an assembly of two plasmonic
NRs, which was realized experimentally in Refs. 20,21,54. The fabrication of such bio-plasmonic
assemblies is feasible thanks to DNA origami nanofabrication technology,20,55 -- 57 which is
presently very well developed. Another suitable model for the HE generation effect in a chiral
plasmonic system is a single monolithic NC with a chiral shape; such NCs were proposed by us
in Ref. 58. Among the different examples of research published on such monolithic chiral NCs,
59 -- 63 the NCs reported in the very recent paper 63 are really suitable for the processes described
herein, since they have very large chiral asymmetry factors. The other part of the discussed
concept is the generation of HEs and injecting them into small TiO2 nanoparticles (or clusters)
deposited on the NC surface. The electrons generated by the NC plasmon and then transferred to
TiO2 can in a later step participate in chemical reaction in a liquid, taking advantage of the energy
5
alignment of TiO2 conduction band and certain molecular excited states. This scheme with HEs
and small TiO2 nanoparticles was realized by us in Refs. 50,51. In the papers referenced above,
long-lived plasmonic electrons injected into the small titania nanoparticles were transferred to
the solution creating reactive radicals that led to the photodegradation of dye molecules. On the
other hand, the NCs used in Refs. 50,51 were not chiral and, therefore, the observed strong
photochemistry was not be sensitive to the circular polarization of exciting light.
In this paper, we demonstrate very large CD signals in the HE generation, which can in turn
induce chemical reactions. 32,50 -- 53 In fact, we predict and propose polarization-sensitive
photochemistry with very large asymmetry g-factors, in the order of 0.15-0.6 (15-60%). Such
giant g-factors are only possible for chiral plasmonic systems with very strong absorption
resonances and strong plasmonic near-field interactions. To compare this category of systems
with chiral molecular systems, we now look at the typical g-factors of representative chiral
molecules. For example, the g-factor of proteins with the α-helix secondary structure, which is
strongly chiral, is ~ 10-3 64; some other molecules may have g-factors like ~ 0.05 (polyaromatic
compounds from Ref. 65) or ~0.01 (alleno -- acetylenic macrocycles from Ref. 66).
It is interesting to compare our photophysical and photochemical mechanisms with those of
traditional chiral photochemistry, which deals directly with chiral molecules. 67 The principle
difference between our proposed effects and traditional chiral photochemistry is that the latter
studies asymmetric photoreactions of chiral molecular species under CPL. But the mechanisms
under discussion here are asymmetric in their sensitivity to CPL, not in the selectivity of
molecular enantiomers. Therefore, the reacting molecules are either nonchiral or preserve their
enantiomeric ratios, but the plasmonic catalyst (chiral NCs) is chiral and, moreover, exhibits
unusually-large asymmetric responses to CPL. Such strong optical asymmetry in the absorption
6
of light between the two enantiomers of chiral species is fundamentally impossible (chiral
molecules can be too small in size and are just excitonic, with polarizabilities much weaker than
those for a plasmonic material). In chiral photochemistry, the central characteristic is the
preferential promotion of reactions with the handedness of incident light, and one can see cases
in which: (1) initial molecular states are chiral; (2) products are chiral; (3) both initial and final
molecular states are chiral. Products of chiral photochemistry can be nonchiral as a result of
photo-destruction. An excellent review on chiral photochemistry67 defines and discusses three
groups of photochemical reactions: Photoderacemization, asymmetric photodestruction, and
asymmetric synthesis. These mechanisms, which names should be self-explanatory, are of great
interest for their clear applications in pharmaceutical, and overall biological applications. The
origin of chiral photo reactions lies, of course, in the optical CD of chiral molecules, that is
asymmetric absorption. But the molecular CD is very small, and the asymmetry of photoreaction
products obtained from a racemic mixture, so-called enantiomeric excess (ee) would not be
observable directly after the absorption process. To see nonzero values of ee after such a process,
one should apply autocatalytic methods or other type of amplification, which are non-linear
kinetic methods in chemistry. In other words, chiral photochemistry can be only based on the
fundamental process of optical absorption asymmetry (i.e. CD), but should also involve non-
trivial chemical processes together with the CPL illumination to produce measurable results. To
give an example, the ee produced by some chiral molecular reaction can be as high as 4% (the 4-
cyclo-octenone system from refs. 67,68), which is higher than the CD for the same molecular
species. Since this paper is focused on the optoelectronic and photochemical properties of chiral
plasmonic NCs, providing a detailed description of chiral molecular photochemistry lies beyond
its scope, and would suggest the interested reader to look into the excellent review in Ref. 67.
7
We now resume the discussion on the subject of this paper, which is the asymmetric generation
of HEs in chiral NCs under CPL excitation. One can see the following applications for the
photochemical response of chiral NCs: Case 1: Polarization-sensitive photochemistry with
nonchiral molecules. In this case, the catalyst (i.e. the NC) is chiral and exhibits very large g-
factors for both CD and the HE rates. We can start with a solution containing one enantiomer of
the chiral NC, and this NC solution should have strong CD. Under CPL illumination, the HEs
induce nonchiral chemical reaction at the surface or in solution; one well-documented effect of
this HE injection is the photodegradation of dye molecules in solution. 50,51 Chirality of the dye
molecules and their products is not relevant in this case, as the photocatalytic process will not
carry its chiral asymmetry to the molecular reaction. The solution itself can be racemic, but the
NC system should have only one enantiomer. Then, the photoproduction of molecules (products
of degradation of dye) will strongly depend on the polarization of CPL, although the molecular
ee (such ee can also be zero) will not depend on it. Calculated g-factors of such non-chiral
reactions can be giant, of up to ~ 60%, as shown below. Such polarization-sensitive
photochemistry can be used as an alternative, indirect method to observe chirality of nanoscale
objects when usual CD spectroscopy cannot be used; for example, our method can be used when
a solution is fully opaque and the optical transmission is practically zero. Case 2: Chiral growth
of nanocrystals, or "chiral photochemistry" for the chiral plasmonic nanocrystals. It has been
established that NC growth can be induced by HEs. 32,52,53,69 The rate of growth should depend
on the total number of HEs generated in a complex. Then, a NC with a given chirality will grow
differently under LCP and RCP light. Or, in a racemic mixture of NCs, NCs with opposite
chirality can grow differently and the system will not remain racemic.70 In fact, this effect permits
chiral photochemistry at the level of a crystal structure of a colloidal complex-shape nanocrystal,
8
and we think that this regime has not obvious analogs in chiral photochemistry with molecules.
In this paper, we are showing that CD for HE-induced photochemistry can be remarkably strong
for chiral plasmonic objects. For nonchiral NCs, photochemical mechanisms need a special
consideration.71 Case 3: Asymmetric photodestruction of chiral plasmonic nanocrystals and
complexes. This application shares some similarity with asymmetric photodestruction for chiral
molecules, but, in our case, this mechanism is applied to a colloidal NC system. It entails the
selective asymmetric photo-destruction of one of the enantiomers of the plasmonic chiral
complex. Under intensive excitation, HEs can dissociate a bio-assembled NR-NR complex into
single NRs; in the case of single monolithic chiral NCs, the selective asymmetric destruction can
occur through intense surface chemistry. The rates of HE generation in the linear plasmonic
regime are: Rate
a I
, where can be RCP or LCP, ais the chiral-electronic coefficient,
and Iis the intensity of incident CPL radiation; RCP and LCP stand for the right- and left-
circular polarizations, respectively. In a chiral plasmonic NC or complex,
a
a
RCP
LCD
and its
0
CD is very strong, with the calculated g-factors up to 60%. In the experiments with single
monolithic NCs, the observed g-factors can be up to 20%.63 The probability for dissociation of a
chiral NR-NR complex bound via the DNA linkers certainly increases with the HE rates.
Therefore, if we start with a racemic mixture of NR-NR complexes (L-pair and R-pair), a strong
CPL excitation should after some time create unequal concentrations of the chiral states, mixed
with non-chiral products (single NRs). This scheme resembles the typical photocatalytic
experiments with chiral molecules.67
This paper is organized in the following way: the first section describes the model system and
defines the CD values and related g-factors for optical and HE responses; the following section
presents results for the chiroptical and photo-electronic properties of the NR-NR complexes; in
9
the next section, we look at a design for a broadband chiral complex; finally, we conclude this
study by summarizing the data and proposed effects.
Figure 1. (a) Mechanisms of generation and injection of hot carriers at the interfaces of
metal/semiconductor and metal/molecules, including the processes of photoexcitation and
relaxation of carriers. Energetic (hot) electrons are generated near the nanocrystal surface and
injected into a semiconductor material or into adsorbed chiral molecules. (b,c) Models of chiral
plasmonic NRs dimers with 45° and 90° rotation angles under incident LCP and RCP illuminations.
The NR sizes are: 12nm × 40nm. The DNA templates are depicted as the arrays of tubes (light
grey); the vertical gap between two NRs in this model equals to the thickness of the DNA template.
10
In (b,c), hot electrons are first injected into small TiO2 nanocrystals on the gold surface (as realized
in the experimental study in Ref.50), that later become transferred to the solution. During this
process, the charge carriers trigger the creation of reactive species (radicals) that are able to destroy
dye molecules (or catalyze some other reactions) which should be present in the solution. These
dye molecules can be used for an optical readout, through a process proposed and realized in Ref.
50.
Theoretical framework. Here we present an outline of the theory that supports our results and
discussion. While we separate its description in two parts, naturally following the formal
distinction between the fundamental elements in our study -chirality and hot electrons-, we also
articulate their connection in our model. On one hand, we describe the main concepts regarding
the optical response of chiral plasmonic assemblies, which is in turn built upon classical
electrodynamics. On the other hand, we detail the fundamentally quantum mechanical
phenomenon of hot electron generation in plasmonic particles.
Chiroptical response of a plasmonic system. The extinction cross section, ext, of a particle or
assembly quantifies the strength of its coupling with incident light, and it can be divided into two
components that account for two different types of light-matter interaction: absorption cross
section, abs, and scattering cross section,scat. Nanostructures made by noble metals (e.g. Au,
Ag) are notable examples of plasmonic systems, which exhibit interaction cross sections much
larger than their actual geometrical equivalents when illuminated at their resonant frequencies.
Notably, when the characteristic size of a nanostructure is much smaller than the operating
11
wavelength, i.e. in a regime well described by the quasi-static approximation,
scat can be
negligible when compared with
abs , yielding ext
abs
scat
abs
. Here,
abs is defined by the
ratio of the power absorbed by the system, Qabs, and the incident energy flux, I0:
abs
Q
abs
I
0
, (1)
Q
abs
Im(
metal
)
dV
8
E E ,
ω
*
ω
I
0
1/2
c
0
med
8
E
0
2
,
where metal and med are the dielectric constant of metal and environment (medium), respectively,
ω is the angular frequency of the incident light and 0c is the speed of light in vacuum. The vector
ωE is the complex electric field inside the metal and E0 denotes the field amplitude of the incident
light. The numerical results presented herein have been obtained using a relatively small value for
the incident flux, I0=3.6×103 W/cm2.
Two magnitudes stand at the center of the study of chiral optical activity, circular dichroism
(CD) and g-factor ( CDg ), which are defined in terms of the optical cross sections of a system:17,72
CD
ext,L
ext
ext,R
(2)
g
CD
(
ext,L
ext,L
ext,R
ext,R
) / 2
12
where the subscripts L and R stand in place of the abbreviations LCP and RCP, respectively. One
should note that, when considering an ensemble of colloidal particles, the extinctions used above
should be found via the following averaging procedure:
ext,
ext
k
,
ext,
k
,
x
k
,
ext,
3
ext,
k
,
z
y
,
,
L R
(3)
Here kx, ky and kz explicitly denote the use of incident wavevectors aligned with the x, y and
z directions, respectively. In other words, the values for the extinction spectra are obtained from
averaging over the three main incident directions.58,73 This treatment accounts for the random
orientations of individual plasmonic dimers within the ensemble in solution. This averaging
procedure is exact for nanocrystal complexes with relatively small dimensions.58,74 It is worth
noting that in the field of molecular spectroscopy the molar extinction coefficient molar is more
commonly used magnitude than extinction cross sections. Thus, let us briefly mention that the
conversion between
ext units of m-2 and molar (units of M-1cm-1) is the following:
molar
N
A ext
/ 0.2303
, here NA is Avogadro's number.
Generation of hot electrons in chiral systems: The photoexcitation process entails the absorption
of a photon and the connected transition of an electron to an excited state (Figure 1a). During this
process, two different types of excited electrons emerge in the system75: one is labeled
"thermalized", and their population is described by a Fermi-Dirac distribution with an effective
temperature
eT which is higher than that of the lattice temperature. The other type of excited
electrons is termed "non-thermalized", and their population cannot be described by the Fermi-
Dirac distribution, as it extends broadly to energies well above the Fermi energy (EF), and up to
13
FE . Such property of nonthermal, or hot, electrons is the key to use them in photocatalytic
processes, as they can be transferred to suitably aligned electronic states of semiconductors and
molecules that would not be energetically accessible by an internal promotion of energy
, as
Fig. 1a depicts schematically. In such a case, however, the energy threshold is set so that the
excited electron can overcome the energy barrier
bE between the materials. In plasmonic
nanostructures, the transition of electrons from the Fermi sea is a collective effect, resulting in a
spatial stepwise distribution of nonthermal HEs with high kinetic energy (the pink areas in Figure
1a highlight the regions, near to the NC surfaces, that will preferentially excite electrons well above
the Fermi level, with energies such that
E
F
k T
B e
E
F
). Likewise, hot holes (HHs)
undergo the same process of excitation and a fraction of them will have energies such that
E
F
E
F
k T
B e
(also, preferentially in the red areas in Figure 1a), naturally respecting the
energy and charge conservation. If we couple a plasmonic nanostructure to a molecule or a
semiconductor (e.g. TiO2), the generated HEs can be injected into the adjacent material if the
electron energy exceeds the barrier height, i.e. if
E
F
, where they can be used for
E
b
subsequent chemical reactions in the surrounding medium. It is relevant to note that the generation
of nonthermal hot carriers is an ultrafast and nonequilibrium process, leading to a fast relaxation
process via electron-electron (e-e) scattering, with the typical lifetime of hot carriers being shorter
than 100 fs.47,76 Therefore, most hot carriers generated near the surface of a plasmonic
nanostructure will be injected into the surrounding media. At the same time, under constant
illumination the system can generate hot carriers in a steady-state regime, offering a constant
stream of carriers to drive chemical reactions occurring at timescales much larger than the typical
lifetime of hot carriers inside the metal.
14
Now, to study the dependence of HE generation in the bio-assembled plasmonic complexes we
need to quantify the rates of HE generation. We can derive them by solving the quantum master
equation in a perturbative approach,75,77 -- 79 but let us just remind here the final expression for these
generation rates:
Rate
HE
2
2
2
e E
2
F
1
3
S
2
E
normal
ds
, (4)
where
E
normal
E n is the electric field internal to the nanostructure and normal to its surface,
as shown in Figure 1b. In Eq. 4, S is the surface of the metal structure. Importantly, Eq. 4 highlights
that the excitation of HE is a surface quantum effect, and it originates from the non-conservation
of the electron's linear momentum due to its scattering by the surface. In contrast, classical
electromagnetic absorption is a volume effect associated with the Ohmic heat (Eqs. 1).
When incident photons carry an energy
bE
, the corresponding excited fraction of the
distribution of HEs can effectively overcome the barrier between the metal and the adsorbed
molecules or semiconductor (see Figures 1b,c). The generation rates of such high-energy HEs
(Rate high energy) can be calculated as: 75,77 -- 79
Rate
high energy
2
2
2
e E
2
F
)
(
E
b
4
S
2
E
normal
ds
(5)
Obviously, Rate high energy includes only a fraction of Rate HE, for electrons with energies beyond a
given threshold, i.e. Rate high energy < Rate HE. Note that Rate high energy is usually not equal to the
injection rates of HEs (Rate injection) but it is reasonable to assume that the latter will be proportional
15
to the former, so that an increase in high energy HE generation will lead to a proportional increase
in HE injection. In practice, Rate injection is influenced by various factors, such as reflection at the
interface and a tunneling effect.36
Since these rates depend on the intensity of the electric field at the surface of the particle, they
will of course depend on the direction of incidence of light. Thus, as we deal with a system in
solution, we again need to average this physical quantity:
Rate
high energy
k
,
Rate
high energy,
3
Rate
high energy,
k
,
x
Rate
high energy,
,
k
y
Rate
high energy,
,
k
z
,
,
L R
(6)
Finally, to quantify the chiral response in terms of HE generation, itself a proxy for photocatalytic
reaction rates, we should define a novel extension to the aforementioned optical chiral parameters:
hot-electron CD and its related g-factor,
CD
high energy
Rate
high energy,L
Rate
high energy,R
(7)
g
CD,high energy
(
Rate
Rate
high energy,L
high energy,L
Rate
Rate
high energy,R
) / 2
high energy,R
,
These new variables are responsible for characterizing the polarization-sensitive generation of HEs
and the related photochemical CD.
Regarding specific implementations of chiral photocatalysts, let us briefly comment on the
viability of creating such systems and on their use in conjunction with TiO2 nanoparticles/clusters
16
to realize HE photo-chemical experiments, in a manner similar to experiments conducted with
achiral photocatalysts. 50,51 This computational paper studies as a model system a chiral NR-NR
assembly which has already been experimentally demonstrated. 20,21,54 Other systems suitable for
use as chiral photocatalysts have been realized experimentally, such as monolithic chiral NCs with
a cubic frame 63 which does not require the bio-assembly of separate achiral NCs. Simultaneously,
strong photo-chemistry based on plasmonic hot electrons has been demonstrated for achiral Au
NCs covered with TiO2 nanoparticles. 50,51 To activate our chiral NCs photo-chemically, one can
grow small TiO2 nanoparticles on the Au surface replacing some of the DNA ligands. One can
also deposit the chiral NR-NR dimers on silica spheres prior to the TiO2 growth, to make them
more stable. This strategy has already been successfully demonstrated in photo-chemical
experiments, where Refs. 50,51 report the use of such TiO2-Au-silica hybrids. However, as a final
remark on this consideration of the co-catalyst for the plasmonic system we note that, for the
asymmetric photo-destruction of chiral NR-NR pairs, it wouldn't be necessary to use any co-
catalytic TiO2 clusters.
Results and discussions. The model system under consideration is illustrated in Figures 1b,c,
which show two different chiral plasmonic dimers with relative rotation angles between the NRs
of 45° and 90°, respectively. Hereafter, we will refer to the systems by using these angles as labels.
Such well-defined systems can be straightforwardly fabricated using the DNA origami
nanofabrication technique. 20,21 In our model system, the dimers are made Ag and Au, materials
representative of noble metals with strong plasmonic responses. The sizes of the individual NR
are taken to be 12 nm × 40 nm, much smaller than the operating wavelength of the system, ranging
from 400 to 900 nm. The gap between the two NRs is varied within the interval 5-90 nm,
accounting for the capability of experimentally controlling this distance by changing the number
17
of DNA layers of origami template. For the 45° dimer, the centers of the two NRs overlap in the
projected x-y plane (Figures 1b,2). For the 90° case, their centers are shifted, and they instead
overlap closer to their corners, as shown in Figure 1c,2, in order to produce a chiral geometry.
These geometrical parameters correspond to those of existing experimental reports.20,56 The
complex permittivity of Ag and Au are taken from the literature,80 and the solvent is water
(
med
1.8
). In order to simplify the computational approach and to provide a clearer, more
general discussion of the phenomenon, we consider the generation of HEs in metallic structures
immersed in a homogeneous dielectric medium, neglecting the screening effect of very small
intermediate TiO2 nanocrystals. Furthermore, although our description does not include explicit
modelling of the carrier transfer through the TiO2 towards the molecular compounds, this
simplified description remains informative of the general picture because the above screening
effect can only cause a small red shift of the plasmonic peaks. 50
As an initial step towards understanding the differential chiral performance of HE generation
in these systems, we begin by discussing the chiral optical profiles of the chiral plasmonic dimers.
Using Eqs. 1,2 with data obtained from classical electrodynamic simulations (full details can be
found in the SI), we calculated the optical chiral performance of the two types of plasmonic dimers
with different gap sizes (5 and 13 nm). Results for Au can be found in Fig. 2, while those for Ag
are in the SI, and they depict the full optical profile of the systems under the two polarizations of
CPL, as well as the spectra of the differential chiral variables, CD and g-factor. As a first
observation, we can point out to Fig. 2d (and Fig. S2d), which showcase values of g-factor
exceeding 0.4 for the 45° Au dimer (above 0.6 for the 45° Ag dimer). These are orders of
magnitude larger than those for molecular systems, with g~10-3(-4), 64,81 and arise from the strongly
chiral symmetry of the engineered plasmonic assemblies. Now, if we look closely at the data in
18
Fig. 2, we will see that the illumination of our system by LCP and RCP light (Figs. 2a,b,c,f) excites
strong longitudinal plasmonic resonances in the interval 700-800 nm. This is accordance to what
one would expect from Au NRs of such aspect ratio immersed in water. Importantly, we observe
in our system a large Rabi-splitting of the individual NR plasmonic modes for small gaps, which
is induced by the near-field NR-NR interaction. We also see an absorption band in the interval
400-550 nm, which comes from the transverse plasmons in the NRs and from the interband
transitions in gold.
At this point, it is interesting to discuss in some more detail the physical picture of the
plasmonic modes and the Rabi-splitting effect, before delving into the discussion of the HE
generation in these systems. We can better see the underlying physics of the plasmonic modes by
considering the maps of free surface charge density for the 45° system at the top of Figure 2. In
them, the different characters of the chiral modes are apparent: the symmetrical charge distribution
at0 = 710 nm has a strongly dipolar character, while the asymmetrical charge distribution at
0=790 nm reveal a more complicated multipolar mode. Correspondingly, the higher energy mode
couples more strongly with the far-field than the lower energy, multipolar one (see small gap
curves in Figs. 2a,b, 3a,b, S2a,b, S3a,b). Such distributions correspond to the two normal modes
in the small-gap system and arise from the near-field interaction of the strong dipolar plasmons of
the two constituting NRs. As such, the Rabi splitting effect appears only for sufficiently small gaps
between the NRs, so that it is not appreciable for our dimers with a 13 nm gap but becomes clear
for the system with 5nm of interparticle distance. Furthermore, we make an interesting observation
in comparing the data in Figures 2,3,S2,S3 between the Rabi-split spectra for the 45° and 90°
dimers: the Rabi peaks are strongly asymmetric for the 45° NR pair but closely symmetric for the
90° complex. This can be understood in terms of the symmetry of the normal modes for each of
19
the dimers, and it can be easily appreciated if we look at the dipolar component of these modes
(Figure S4). The relevant comparison of surface charge maps and dipole diagram can be found in
the Supporting Information (Figure S4), together with more details about the Rabi-splitting effect
in the 45° and 90° NR dimers. To summarize, as one can see, we find very large chiral g-factors
for the NR-NR geometry. In addition, we have previously demonstrated computationally 20,21 that
such strong CD signals are robust against fluctuation of the geometrical angle in the NR-NR pair.
20
21
Figure 2. (a-b) Simulation results for the averaged extinction spectra of the NR dimers with 45°
angle under CPL light with different polarizations. (c-d) Chiral optical response of the 45° dimers.
These four panels show two curves each, for two different gaps between the rods. The insets on
the left show the 45° dimers from two orthogonal perspectives. (e-h) The equivalent information
than panels a-d, but for the 90° NR dimer. The top section of the figure shows four maps of the
surface charges for the Au 45° dimer with a gap of 5 nm, at the normal modes' wavelengths and
for both polarizations of light, revealing the nature of the two bright plasmonic modes in the dimers.
These modes make the positive and negative chiral bands for the Au NR dimer (panels c-d).
After discussing the main features of the optical response of the dimers, we now evaluate their
dynamics of HE generation by using Eqs. 6,7. The averaged generation rates of energetic HEs
(Rate high energy) and the corresponding chiral asymmetry metrics for the different Au dimers are
shown in Figure 3, whereas the equivalent data for Ag systems is in Fig. S3. It is relevant to
remember that Rate high energy denotes the generation rate of the over-barrier nonthermal HEs with
energies
bE , which is only a fraction of the total generation rate of HEs, Rate HE. This barrier
represents the energy threshold for the plasmonic photocatalytic system to contribute to the
chemical reactions, so that we need >Eb in order to have some HE that can propagate into
the TiO2 or molecular adsorbates. At the same time, this situation is more energetically favorable
than in absence of the plasmonic photocatalysts, as the energy barrier is lower than other energy
thresholds in the system, such as the bandgap of TiO2 or many relevant molecular transitions. For
our calculations, we took Eb = 1 eV as it is a representative value of a Schottky energy barrier. 45
It is interesting to compare the optical extinction and HE spectra in Figures 2,3. In these figures,
the extinction and HE spectra look similar for the spectral interval of the longitudinal plasmon
22
resonance (650-850nm) where the intraband electronic excitations dominate and the dielectric
constant is well described by the Drude model. The reason is that both effects (optical CD and HE
CDs) originate from the same plasmon resonances. However, for the interband excitation interval
in gold (400-600nm), the differences in the extinction-CD and HE-CD spectra should be apparent,
and we indeed see that in our calculations (Figures 2,3,S5). Moreover, we note an interesting
qualitative difference between the optical and photochemical CD effects. This difference comes
from the fundamental underlying phenomena causing these effects, as well as from their different
mathematical expressions. As mentioned above, the extinction signal for relatively small
complexes, such as the ones studied here, comes mostly from the absorption inside the plasmonic
components, which is a bulk effect. Correspondingly, the mathematical expressions for the optical
extinction and HE generations are different. The energy dissipation in our systems is a classical
effect (Drude absorption via the "frictional" Joule-heating mechanism) and occurs in a whole
volume of the NRs. The case of the HE generation is different in two respects. First, it is a quantum
effect (see Eqs. 4,5). Second, the HE generation is a surface effect appearing at the metal-
environment interfaces due to the breaking of the linear momentum conservation for the electrons.
In our NR complexes, this effect occurs preferentially at the end of the NRs, as shown in Figure 4.
To conclude the comparison between optical-CD and HE-CD, we should also mention that for the
Ag complexes the optical and optoelectronic spectra show different line-shapes at the main
plasmonic peaks as well (Figures S2,S3). This is because the Ag NCs exhibit much stronger
plasmonic enhancements and, therefore, the fine details of the physical processes start to play more
important roles. Another interesting case is that of the four interacting NRs (Figure 6), where the
interactions and plasmonic modes become much more complex and the physical and mathematical
differences between optical-CD and HE-CD become more evident in the resulting spectra.
23
At this point, and using the data in Figures 3 and S3 (see Supporting Information), we can
support the following conclusions: (1) Compared with the Au dimers, the Ag dimers exhibit a
larger Rate high energy and CD in the visible wavelength interval, due to the difference between the
dielectric functions of these two noble metals (see Supporting Information). (2) A remarkable Rabi
splitting can be observed in all studied systems, but especially for the ones with a smaller gap size
(5 nm) and for the more compact dimer (45° rotation angle); of course, this comes from the
increased coupling strength between the NRs. Importantly, the Rabi splitting boosts chiral
behaviors of Rate high energy, enhancing hot-electron g-factor of Au-dimers with 45° rotation angle
(Figures 3c,d). (3) The effects of Rate high energy become larger in the near-infrared (NIR) interval,
which is a signature of the frequency-dependent multiplier
(
)
4
in Eq. 5; this property looks
promising for NIR photochemistry. (4) The Rabi splitting strongly influences the spectral shape
of the CD signals, creating two distinct plasmon peaks (Figure 3); this effect is even stronger for
the Ag complex (Figure S2,S3). (5) Overall, we observe large values of the g-factors ( CDg
and
g
CD,high energy
), with a maximum of 0.65 for the Ag dimers with 45° rotation angle and 13 nm gap,
as seen in Figure S3d. Such plasmonic structures can serve as efficient chiral catalyzers in
photochemistry.
24
25
Figure 3. Simulation results for the averaged HE-generation spectra (a,b,e,f) and their related
chiral magnitudes (c,d,g,h) for the plasmonic dimers with 45° angle (a-d) and 90° angle (e-h).
Calculations were done for the Au-NRs in water for the gaps of 5 nm and 13 nm. The g-factors
obtained for both systems, but specially for the 45° dimer, indicates their suitability as
nanocomplexes to trigger polarization sensitive photocatalysis.
Now we take a closer look at the local maps of HE generation in Figure 4. The local generation
rate of energetic HEs per surface area is defined as:79
r
high energy
r
( )=
2
2
2
F
e E
2
)
(
bE
)
(
4
E
normal
r
2
(8)
where the position-dependent quantity high energy( )
r
r has the units of s-1m-2. In Figure 4, we show
the representative maps at the characteristic wavelengths of 710 nm and 790 nm (for the Au-dimer
with the 45° rotation angle and the 5 nm gap size), which correspond to the maximum (CDmax) and
the minimum (CDmin) of the CD spectra in Figures 2c,3c. In these maps, we only considered
incidence along the z-direction (kz) for incident circular polarized illumination, because the
direction kz dominates in the averaged CD spectra. Evidently, the local surface density of
generation of HEs is concentrated at the ends of NRs, resulting from the plasmonic hot spots and
their enhancement of Enormal at plasmonic modes which fundamentally excite a combination of
the dipole modes of the individual NRs (see the top of Fig. 2 and Fig. S4a). These hots spots are
the small surface areas where the maximum photochemical activity is expected and, furthermore,
they also have the maximum values for the local hot-electron CD (Figures 4c,f).
26
The above effects have relevant implications for some of the possible applications described
in the introduction. The total rate of NC growth can be assumed to be proportional to the total rate
of generation of HEs in a NC and, therefore, the enantiomers of a chiral NC should grow at
different rates. Moreover, the local crystal growth or surface photochemistry should be
proportional to the local rate of HE generation near the surface and, for a given CPL beam (RCP
or LCP), surface patterns of photochemistry will be different for each of the two chiral states of
the NR-NR complex. So, one particular enantiomer of the NR-NR complex will show different
growth patterns and rates under different CPL illuminations (LCP or RCP). In this way, the HE
CD signals can be translated to localized surface photochemistry and the related structural
differences between the two configurations of the plasmonic complex.
Figure 4. Maps of hot electron generation for the incident light propagating in the z-direction with
wavelengths of 710 nm (CDmax) and 790 nm (CDmin). These wavelengths correspond to the two
plasmonic resonances in the given dimer. Here, we show the case of the Au-dimer with 45° rotation
angle and the 5 nm gap size.
27
Finally, we summarize the optical and HE CD effects for both geometries and for the two
involved metals (Au and Ag). We observe the following features in Figure 5: (1) the HE CD mainly
follows the optical CD; (2) The CDs signals for both geometries show a maximum of CD at
moderate gap sizes (10-20nm). This gap size dependence can be understood physically in terms of
the NR-NR interactions: at long distances the NRs start behaving as uncoupled resonators, so that
the complex does not show chirality, while at very small gap sizes (<5 nm) the coupling is strong,
but the collective assembly's modes begin to become closer to those of an in-plane resonator with
the cross-like geometry, which is non-chiral.
Case 1: small gaps. For small gaps the CD signals decrease because the structure becomes
geometrically less chiral; if we were to fuse the NRs making their axes overlay in one common
plane, the structure would become non-chiral and the CD signals will vanish. We clearly see this
trend in Figure 5 for the small gaps. Mathematically, the CD signals at small separations depend
on the geometry in the following way:
C D
(
k g
)
(
f b
)
(9)
Here
k
2 /
is the wave vector of light, g is the gap size and b is the NR length. This
functional dependence follows from the analysis made in Refs.73,74. The factor k g
comes from
the retardation effects along the direction normal to both NRs (z-direction in our geometry in
Figure 1), and if it becomes too small the two interacting dipoles start collapsing into a single in-
plane structure.
Case 2: Large NR-NR separations. For large gaps, the CD decays since the dipole-dipole and
electromagnetic interactions between the NRs decay with the distance. The reason is that all chiral
28
responses in the chiral NR pairs originate from the NR-NR interactions between two single NRs
that are not chiral. Asymptotically, we see from the numerical calculations that, for large g ,
CD
g
1/
2
. We can understand this behavior in the following way: the dipole-dipole interaction
for long distances gives the typical behavior
( )
f b
b
1/
3
, but with the retardation coefficient k g
from Eq. 9, we indeed obtain
CD
g
1/
2
for g .
29
Figure 5. Peak values of the optical and HE chiral signals as functions of the bio-assembly gap
for the 45° Au dimer (a) and the 90 ° Au dimer (b). We observe that for an increasing gap size, the
CD signal decays faster for the case of the 90° dimer.
Practical applications often require, within the context of solar energy conversion, not only the
efficient generation of high energy HEs at one single frequency, but also a broadband utilization
of the entire solar spectrum. For this goal, we also designed a helical Au-tetramer superstructure
for broadband generation of energetic (over barrier) HEs (inset of Figure 6), where the length of
the AuNRs varies from 25 nm to 40 nm with a 5 nm step and their rotation angle is changed from
0° to 135° with a 45° step. The gap size is set at a constant 5 nm between each AuNR. As shown
in Figure 6 a,d, the extinction and Rate high energy now extend from 500 nm to 800 nm, due to the
different resonant frequencies of the plasmonic modes in the AuNRs. Likewise, the CD spectra
and the g-factor display broader spectra for the optical and HE responses, and each plasmonic
resonant mode can be recognized both in the optical and HE generation. Please note that we did
not perform a geometry optimization to maximize the system's bandwidth, but took instead
realistic parameters of NRs. Therefore, the performance can be further improved by optimizing
geometrical parameters such as the gap size, rotation angle and length of AuNRs. Interestingly,
the chiral performance is different between the optical extinction and Rate high energy. For example,
the optical CD spectrum and the extinction g-factor show a broadband bisignate line shape (Figures
6b,c). In contrast, the HE spectra (the HE CD and the g-factor for Rate high energy) primarily show
negative values (Figures 6e,f). The different chiral performances for the optical and HE chiral
effects appear because these effects depend on the different components of the electric fields: the
optical CD comes from the amplitude of ωE in the full volume of the particles, whereas the HE
CD depends on the internal normal field near the surface, Enormal. Furthermore, because of the
30
complexity of the tetramer, this structure has more complex NR-NR interactions as compared to
the NR dimer. Therefore, the CD spectra for the optical extinction and the HE effects differentiate
more clearly from each other in the more complex case of the NR tetramer.
Figure 6. The inset shows the broadband plasmonic Au-tetramer with the step of 45° rotation
angle, i.e. 0°, 45°, 90°, 135° (top-down), and 5 nm steps of the NR length, i.e. 40 nm, 35 nm,
30 nm, 25 nm (top-down), respectively. The gap between NRs is taken uniformly as 5 nm. (a-c)
Extinction spectra for the circularly-polarized light, extinction CD, and g-factor. (d-f) HE spectra
for the circularly-polarized light, HE CD, and the related g-factor.
In conclusion, we have demonstrated that chiral plasmonic nanostructures exhibit very large chiral
optical asymmetries that become transferred into the HE generation processes. We think that this
effect can be used as a new mechanism for polarization-sensitive plasmon-induced hot-electron
photochemistry. The proposed chiral photochemical effect has fundamentally different properties
31
as compared to chiral photochemistry operating asymmetrically over different molecular
enantiomers. However, we can also see some similarities between our proposed polarization-
sensitive photochemistry and the traditional case of chiral molecular photochemistry. Our models
and predictions are based on the realistic structures and conditions for the plasmonic nanoscale
systems. Unlike the case of molecular systems, we show that optical CD, HE generation rates and
rates for related photochemical processes should exhibit giant g-factors, or sensitivity to changes
in the polarization of CPL. In the Au-dimers, the g-factor for hot electron generation can be as
high as 15%, whereas the Ag-dimers exhibit even higher values for the chiral factor, up to ~ 60%.
Such remarkable values should be certainly observable in photo-chemical experiments involving
generation of HEs. For the broadband HE phenomena in the multi-NR structures, our results show
a spectrally broad generation of hot electrons, with strong plasmonic CD for the wider spectral
interval of 500 nm -- 800 nm. Overall, our results offer interesting possibilities for new
applications, such as polarization-sensitive photochemistry, asymmetric growth of chiral
plasmonic nanostructures, or asymmetric photo-destruction of plasmonic complexes.
ASSOCIATED CONTENT
Supporting Information
This information includes the model used in COMSOL, the chiro-optical properties of Ag dimers,
the structure of normal modes in the dimers and the maps of surface charge distribution.
AUTHOR INFORMATION
32
Corresponding Author
*E-mail: [email protected]
*E-mail: [email protected]
ORCID
Alexander O. Govorov: 0000-0003-1316-6758
Tim Liedl: 0000-0002-0040-0173
Lucas V. Besteiro: 0000-0001-7356-7719
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENT
T. Liu and L.V.B was supported by the Institute of Fundamental and Frontier Sciences, University
of Electronic Science and Technology of China. Z.W. was supported by National Basic Research
Program of China (Project 2013CB933301) and National Natural Science Foundation of China
(Project 51272038). This last author (A.O.G.) was funded via the 1000-talent Award of Sichuan,
China. The collaborative Germany-US component was generously funded by the Volkswagen
Foundation (T. Liedl and A.O.G.). Finally, A.O.G. holds Chang Jiang (Yangtze River) Chair
Professorship in China.
REFERENCES
(1) Maier, S. A. Plasmonics: Fundamentals and Applications; Springer US, 2007.
(2) Pendry, J. B.; Schurig, D.; Smith, D. R. Controlling Electromagnetic Fields. Science 2006,
312 (5781), 1780 -- 1782.
33
(3) Pendry, J. B. Negative Refraction Makes a Perfect Lens. Phys. Rev. Lett. 2000, 85 (18),
3966 -- 3969.
(4)
Jacob, Z.; Alekseyev, L. V.; Narimanov, E. Optical Hyperlens: Far-Field Imaging beyond
the Diffraction Limit. Opt. Express 2006, 14 (18), 8247.
(5) Ni, X.; Kildishev, A. V.; Shalaev, V. M. Metasurface Holograms for Visible Light. Nat.
Commun. 2013, 4 (1), 2807.
(6) Ni, X.; Emani, N. K.; Kildishev, A. V; Boltasseva, A.; Shalaev, V. M. Broadband Light
Bending with Plasmonic Nanoantennas. Science 2012, 335 (6067), 427.
(7) Kawata, S.; Ichimura, T.; Taguchi, A.; Kumamoto, Y. Nano-Raman Scattering Microscopy:
Resolution and Enhancement. Chem. Rev. 2017, 117 (7), 4983 -- 5001.
(8) Verma, P. Tip-Enhanced Raman Spectroscopy: Technique and Recent Advances. Chem.
Rev. 2017, 117 (9), 6447 -- 6466.
(9) Atwater, H. A.; Polman, A. Plasmonics for Improved Photovoltaic Devices. Nat. Mater.
2010, 9 (3), 205 -- 213.
(10) Clavero, C. Plasmon-Induced Hot-Electron Generation at Nanoparticle/Metal-Oxide
Interfaces for Photovoltaic and Photocatalytic Devices. Nat. Photonics 2014, 8 (2), 95 -- 103.
(11) Hou, W.; Cronin, S. B. A Review of Surface Plasmon Resonance-Enhanced Photocatalysis.
Adv. Funct. Mater. 2013, 23 (13), 1612 -- 1619.
(12) Govorov, A. O.; Fan, Z.; Hernandez, P.; Slocik, J. M.; Naik, R. R. Theory of Circular
Dichroism of Nanomaterials Comprising Chiral Molecules and Nanocrystals: Plasmon
34
Enhancement, Dipole Interactions, and Dielectric Effects. Nano Lett. 2010, 10 (4), 1374 --
1382.
(13) Collins, J. T.; Kuppe, C.; Hooper, D. C.; Sibilia, C.; Centini, M.; Valev, V. K. Chirality and
Chiroptical Effects in Metal Nanostructures: Fundamentals and Current Trends. Adv. Opt.
Mater. 2017, 5 (16), 1700182.
(14) Li, W.; Coppens, Z. J.; Besteiro, L. V.; Wang, W.; Govorov, A. O.; Valentine, J. Circularly
Polarized Light Detection with Hot Electrons in Chiral Plasmonic Metamaterials. Nat.
Commun. 2015, 6 (1), 8379.
(15) Fang, Y.; Verre, R.; Shao, L.; Nordlander, P.; Käll, M. Hot Electron Generation and
Cathodoluminescence Nanoscopy of Chiral Split Ring Resonators. Nano Lett. 2016, 16 (8),
5183 -- 5190.
(16) Hao, C.; Xu, L.; Ma, W.; Wu, X.; Wang, L.; Kuang, H.; Xu, C. Unusual Circularly Polarized
Photocatalytic Activity in Nanogapped Gold-Silver Chiroplasmonic Nanostructures. Adv.
Funct. Mater. 2015, 25 (36), 5816 -- 5822.
(17) Kong, X. T.; Khosravi Khorashad, L.; Wang, Z.; Govorov, A. O. Photothermal Circular
Dichroism Induced by Plasmon Resonances in Chiral Metamaterial Absorbers and
Bolometers. Nano Lett. 2018, 18 (3), 2001 -- 2008.
(18) Lan, X.; Wang, Q. Self-Assembly of Chiral Plasmonic Nanostructures. Adv. Mater. 2016,
28 (47), 10499 -- 10507.
(19) Govorov, A. O.; Gun'ko, Y. K.; Slocik, J. M.; Gérard, V. A.; Fan, Z.; Naik, R. R. Chiral
35
Nanoparticle Assemblies: Circular Dichroism, Plasmonic Interactions, and Exciton Effects.
J. Mater. Chem. 2011, 21 (42), 16806.
(20) Kuzyk, A.; Schreiber, R.; Zhang, H.; Govorov, A. O.; Liedl, T.; Liu, N. Reconfigurable 3D
Plasmonic Metamolecules. Nat. Mater. 2014, 13 (9), 862 -- 866.
(21) Shen, X.; Zhan, P.; Kuzyk, A.; Liu, Q.; Asenjo-Garcia, A.; Zhang, H.; García de Abajo, F.
J.; Govorov, A.; Ding, B.; Liu, N. 3D Plasmonic Chiral Colloids. Nanoscale 2014, 6 (4),
2077.
(22) Kong, X.-T.; Besteiro, L. V.; Wang, Z.; Govorov, A. O. Plasmonic Chirality and Circular
Dichroism in Bioassembled and Nonbiological Systems: Theoretical Background and
Recent Progress. Adv. Mater. 2018, 1801790.
(23) Ben-Moshe, A.; Maoz, B. M.; Govorov, A. O.; Markovich, G. Chirality and Chiroptical
Effects in Inorganic Nanocrystal Systems with Plasmon and Exciton Resonances. Chem.
Soc. Rev. 2013, 42 (16), 7028 -- 7041.
(24) Guerrero-Martínez, A.; Auguié, B.; Alonso-Gómez, J. L.; Džolić, Z.; Gómez-Graña, S.;
Žinić, M.; Cid, M. M.; Liz-Marzán, L. M. Intense Optical Activity from Three-Dimensional
Chiral Ordering of Plasmonic Nanoantennas. Angew. Chemie Int. Ed. 2011, 50 (24), 5499 --
5503.
(25) Smith, K. W.; Zhao, H.; Zhang, H.; Sánchez-Iglesias, A.; Grzelczak, M.; Wang, Y.; Chang,
W.-S.; Nordlander, P.; Liz-Marzán, L. M.; Link, S. Chiral and Achiral Nanodumbbell
Dimers: The Effect of Geometry on Plasmonic Properties. ACS Nano 2016, 10 (6), 6180 --
6188.
36
(26) Cushing, S. K.; Wu, N. Progress and Perspectives of Plasmon-Enhanced Solar Energy
Conversion. J. Phys. Chem. Lett. 2016, 7 (4), 666 -- 675.
(27) Brennan, L. J.; Purcell-Milton, F.; Salmeron, A. S.; Zhang, H.; Govorov, A. O.; Fedorov,
A. V; Gun'ko, Y. K. Hot Plasmonic Electrons for Generation of Enhanced Photocurrent in
Gold-TiO2 Nanocomposites. Nanoscale Res. Lett. 2015, 10 (1), 38.
(28) Mukherjee, S.; Libisch, F.; Large, N.; Neumann, O.; Brown, L. V.; Cheng, J.; Lassiter, J.
B.; Carter, E. A.; Nordlander, P.; Halas, N. J. Hot Electrons Do the Impossible: Plasmon-
Induced Dissociation of H 2 on Au. Nano Lett. 2013, 13 (1), 240 -- 247.
(29) Brongersma, M. L.; Halas, N. J.; Nordlander, P. Plasmon-Induced Hot Carrier Science and
Technology. Nat. Nanotechnol. 2015, 10 (1), 25 -- 34.
(30) Wu, K.; Chen, J.; McBride, J. R.; Lian, T. Efficient Hot-Electron Transfer by a Plasmon-
Induced Interfacial Charge-Transfer Transition. Science (80-. ). 2015, 349 (6248), 632 -- 635.
(31) Zhou, L.; Swearer, D. F.; Zhang, C.; Robatjazi, H.; Zhao, H.; Henderson, L.; Dong, L.;
Christopher, P.; Carter, E. A.; Nordlander, P.; et al. Quantifying Hot Carrier and Thermal
Contributions in Plasmonic Photocatalysis. Science (80-. ). 2018, 362 (6410), 69 -- 72.
(32) Cortés, E.; Xie, W.; Cambiasso, J.; Jermyn, A. S.; Sundararaman, R.; Narang, P.; Schlücker,
S.; Maier, S. A. Plasmonic Hot Electron Transport Drives Nano-Localized Chemistry. Nat.
Commun. 2017, 8, 14880.
(33) Zhang, Y.; He, S.; Guo, W.; Hu, Y.; Huang, J.; Mulcahy, J. R.; Wei, W. D. Surface-
Plasmon-Driven Hot Electron Photochemistry. Chem. Rev. 2018, 118 (6), 2927 -- 2954.
37
(34) Tian, Y.; García de Arquer, F. P.; Dinh, C. T.; Favraud, G.; Bonifazi, M.; Li, J.; Liu, M.;
Zhang, X.; Zheng, X.; Kibria, M. G.; et al. Enhanced Solar-to-Hydrogen Generation with
Broadband Epsilon-Near-Zero Nanostructured Photocatalysts. Adv. Mater. 2017, 29 (27).
(35) Sprague-Klein, E. A.; Negru, B.; Madison, L. R.; Coste, S. C.; Rugg, B. K.; Felts, A. M.;
McAnally, M. O.; Banik, M.; Apkarian, V. A.; Wasielewski, M. R.; et al. Photoinduced
Plasmon-Driven Chemistry in Trans -1,2-Bis(4-Pyridyl)Ethylene Gold Nanosphere
Oligomers. J. Am. Chem. Soc. 2018, 140 (33), 10583 -- 10592.
(36) Zhang, Y.; Nelson, T.; Tretiak, S.; Guo, H.; Schatz, G. C. Plasmonic Hot-Carrier-Mediated
Tunable Photochemical Reactions. ACS Nano 2018, 12 (8), 8415 -- 8422.
(37) DeSario, P. A.; Pietron, J. J.; Dunkelberger, A.; Brintlinger, T. H.; Baturina, O.; Stroud, R.
M.; Owrutsky, J. C.; Rolison, D. R. Plasmonic Aerogels as a Three-Dimensional Nanoscale
Platform for Solar Fuel Photocatalysis. Langmuir 2017, 33 (37), 9444 -- 9454.
(38) Kale, M. J.; Avanesian, T.; Xin, H.; Yan, J.; Christopher, P. Controlling Catalytic
Selectivity on Metal Nanoparticles by Direct Photoexcitation of Adsorbate -- Metal Bonds.
Nano Lett. 2014, 14 (9), 5405 -- 5412.
(39) Richardson, H. H.; Carlson, M. T.; Tandler, P. J.; Hernandez, P.; Govorov, A. O.
Experimental and Theoretical Studies of Light-to-Heat Conversion and Collective Heating
Effects in Metal Nanoparticle Solutions. Nano Lett. 2009, 9 (3), 1139 -- 1146.
(40) Jack, C.; Karimullah, A. S.; Tullius, R.; Khorashad, L. K.; Rodier, M.; Fitzpatrick, B.;
Barron, L. D.; Gadegaard, N.; Lapthorn, A. J.; Rotello, V. M.; et al. Spatial Control of
Chemical Processes on Nanostructures through Nano-Localized Water Heating. Nat.
38
Commun. 2016, 7 (1), 10946.
(41) Khosravi Khorashad, L.; Besteiro, L. V.; Wang, Z.; Valentine, J.; Govorov, A. O.
Localization of Excess Temperature Using Plasmonic Hot Spots in Metal Nanostructures:
Combining Nano-Optical Antennas with the Fano Effect. J. Phys. Chem. C 2016, 120 (24),
13215 -- 13226.
(42) Hühn, D.; Govorov, A.; Gil, P. R.; Parak, W. J. Photostimulated Au Nanoheaters in Polymer
and Biological Media: Characterization of Mechanical Destruction and Boiling. Adv. Funct.
Mater. 2012, 22 (2), 294 -- 303.
(43) Shi, X.; Ueno, K.; Oshikiri, T.; Sun, Q.; Sasaki, K.; Misawa, H. Enhanced Water Splitting
under Modal Strong Coupling Conditions. Nat. Nanotechnol. 2018, 13 (10), 953 -- 958.
(44) Zhao, H.; Zheng, X.; Feng, X.; Li, Y. CO2Reduction by Plasmonic Au Nanoparticle-
Decorated TiO2Photocatalyst with an Ultrathin Al2O3Interlayer. J. Phys. Chem. C 2018,
122 (33), 18949 -- 18956.
(45) Govorov, A. O.; Zhang, H.; Demir, H. V.; Gun'Ko, Y. K. Photogeneration of Hot Plasmonic
Electrons with Metal Nanocrystals: Quantum Description and Potential Applications. Nano
Today 2014, 9 (1), 85 -- 101.
(46) Zhang, H.; Govorov, A. O. Optical Generation of Hot Plasmonic Carriers in Metal
Nanocrystals: The Effects of Shape and Field Enhancement. J. Phys. Chem. C 2014, 118
(14), 7606 -- 7614.
(47) Hartland, G. V.; Besteiro, L. V.; Johns, P.; Govorov, A. O. What's so Hot about Electrons
39
in Metal Nanoparticles? ACS Energy Lett. 2017, 2 (7), 1641 -- 1653.
(48) Harutyunyan, H.; Martinson, A. B. F.; Rosenmann, D.; Khorashad, L. K.; Besteiro, L. V.;
Govorov, A. O.; Wiederrecht, G. P. Anomalous Ultrafast Dynamics of Hot Plasmonic
Electrons in Nanostructures with Hot Spots. Nat. Nanotechnol. 2015, 10 (9), 770 -- 774.
(49) Sykes, M. E.; Stewart, J. W.; Akselrod, G. M.; Kong, X. T.; Wang, Z.; Gosztola, D. J.;
Martinson, A. B. F.; Rosenmann, D.; Mikkelsen, M. H.; Govorov, A. O.; et al. Enhanced
Generation and Anisotropic Coulomb Scattering of Hot Electrons in an Ultra-Broadband
Plasmonic Nanopatch Metasurface. Nat. Commun. 2017, 8 (1), 986.
(50) Sousa-Castillo, A.; Comesaña-Hermo, M.; Rodríguez-González, B.; Pérez-Lorenzo, M.;
Wang, Z.; Kong, X.-T.; Govorov, A. O.; Correa-Duarte, M. A. Boosting Hot Electron-
Driven Photocatalysis through Anisotropic Plasmonic Nanoparticles with Hot Spots in Au --
TiO 2 Nanoarchitectures. J. Phys. Chem. C 2016, 120 (21), 11690 -- 11699.
(51) Negrín-Montecelo, Y.; Comesaña-Hermo, M.; Kong, X.-T.; Rodríguez-González, B.;
Wang, Z.; Pérez-Lorenzo, M.; Govorov, A. O.; Correa-Duarte, M. A. Traveling Hot Spots
in Plasmonic Photocatalysis: Manipulating Interparticle Spacing for Real-Time Control of
Electron Injection. ChemCatChem 2018, 10 (7), 1561 -- 1565.
(52) Zhai, Y.; DuChene, J. S.; Wang, Y.-C.; Qiu, J.; Johnston-Peck, A. C.; You, B.; Guo, W.;
DiCiaccio, B.; Qian, K.; Zhao, E. W.; et al. Polyvinylpyrrolidone-Induced Anisotropic
Growth of Gold Nanoprisms in Plasmon-Driven Synthesis. Nat. Mater. 2016, 15 (8), 889 --
895.
(53) Wang, P.; Krasavin, A. V.; Nasir, M. E.; Dickson, W.; Zayats, A. V. Reactive Tunnel
40
Junctions in Electrically Driven Plasmonic Nanorod Metamaterials. Nat. Nanotechnol.
2018, 13 (2), 159 -- 164.
(54) Kuzyk, A.; Yang, Y.; Duan, X.; Stoll, S.; Govorov, A. O.; Sugiyama, H.; Endo, M.; Liu, N.
A Light-Driven Three-Dimensional Plasmonic Nanosystem That Translates Molecular
Motion into Reversible Chiroptical Function. Nat. Commun. 2016, 7 (1), 10591.
(55) Douglas, S. M.; Dietz, H.; Liedl, T.; Högberg, B.; Graf, F.; Shih, W. M. Self-Assembly of
DNA into Nanoscale Three-Dimensional Shapes. Nature 2009, 459 (7245), 414 -- 418.
(56) Lan, X.; Liu, T.; Wang, Z.; Govorov, A. O.; Yan, H.; Liu, Y. DNA-Guided Plasmonic Helix
with Switchable Chirality. J. Am. Chem. Soc. 2018, 140 (37), 11763 -- 11770.
(57) Cecconello, A.; Besteiro, L. V.; Govorov, A. O.; Willner, I. Chiroplasmonic DNA-Based
Nanostructures. Nat. Rev. Mater. 2017, 2 (9), 17039.
(58) Fan, Z.; Govorov, A. O. Chiral Nanocrystals: Plasmonic Spectra and Circular Dichroism.
Nano Lett. 2012, 12 (6), 3283 -- 3289.
(59) Cathcart, N.; Kitaev, V. Monodisperse Hexagonal Silver Nanoprisms: Synthesis via
Thiolate-Protected Cluster Precursors and Chiral, Ligand-Imprinted Self-Assembly. ACS
Nano 2011, 5 (9), 7411 -- 7425.
(60) Mark, A. G.; Gibbs, J. G.; Lee, T.-C.; Fischer, P. Hybrid Nanocolloids with Programmed
Three-Dimensional Shape and Material Composition. Nat. Mater. 2013, 12 (9), 802 -- 807.
(61) Ben-Moshe, A.; Wolf, S. G.; Sadan, M. B.; Houben, L.; Fan, Z.; Govorov, A. O.;
Markovich, G. Enantioselective Control of Lattice and Shape Chirality in Inorganic
41
Nanostructures Using Chiral Biomolecules. Nat. Commun. 2014, 5 (1), 4302.
(62) Wang, P.; Yu, S.-J.; Govorov, A. O.; Ouyang, M. Cooperative Expression of Atomic
Chirality in Inorganic Nanostructures. Nat. Commun. 2017, 8, 14312.
(63) Lee, H.-E.; Ahn, H.-Y.; Mun, J.; Lee, Y. Y.; Kim, M.; Cho, N. H.; Chang, K.; Kim, W. S.;
Rho, J.; Nam, K. T. Amino-Acid- and Peptide-Directed Synthesis of Chiral Plasmonic Gold
Nanoparticles. Nature 2018, 556 (7701), 360 -- 365.
(64) Fasman, G. D. Circular Dichroism and the Conformational Analysis of Biomolecules;
Springer US: Boston, MA, 1996.
(65) Walters, R. S.; Kraml, C. M.; Byrne, N.; Ho, D. M.; Qin, Q.; Coughlin, F. J.; Bernhard, S.;
Pascal, R. A. Configurationally Stable Longitudinally Twisted Polycyclic Aromatic
Compounds. J. Am. Chem. Soc. 2008, 130 (48), 16435 -- 16441.
(66) Alonso-Gómez, J. L.; Rivera-Fuentes, P.; Harada, N.; Berova, N.; Diederich, F. An
Enantiomerically Pure Alleno-Acetylenic Macrocycle: Synthesis and Rationalization of Its
Outstanding Chiroptical Response. Angew. Chemie Int. Ed. 2009, 48 (30), 5545 -- 5548.
(67) Rau, H. Chapter 1: Direct Asymmetric Photochemistry with Circularly Polarized Light. In
Chiral photochemistry; CRC Press, 2004.
(68) Fukui K, Naito Y, Taniguchi S, I. Y. Book of Abstracts. International Journal of
Pharmaceutics. 1st. Osaka: Intl Symp Asymm Photochem 2001, p L 106.
(69) Watanabe, M.; Araki, S.; Hayashi, K. Directed Growth of Metal Nanoparticles on
Substrates by Polarized Light Irradiation. In 2015 IEEE SENSORS; IEEE, 2015; pp 1 -- 4.
42
(70) If we start the process of CPL illumination (LCP or RCP) over a racemic mixture of two
chiral states of a NC (enantiomers: A-form and B-form), with time the NC mixture will
depart from being racemic. We will instead have a system with A'- and B'-forms, where A'
and B' are not enantiomers anymore, but new chiral states. The crystal volumes of the A'
and B' forms will be different and their shapes will not be mirror-symmetric, since the total
and local rates of HE generations exhibit a very strong CD. This case has not an obvious
analogue in chiral molecular photochemistry, since we deal here with a fundamentally-
different system -- chiral plasmonic NCs. This fundamental difference stems, of course,
from the fact that our system has a very large number of atoms, as compared to molecular
systems.
(71) Yeom, J.; Yeom, B.; Chan, H.; Smith, K. W.; Dominguez-Medina, S.; Bahng, J. H.; Zhao,
G.; Chang, W.-S.; Chang, S.-J.; Chuvilin, A.; et al. Chiral Templating of Self-Assembling
Nanostructures by Circularly Polarized Light. Nat. Mater. 2015, 14 (1), 66 -- 72.
(72) Kong, X.-T.; Zhao, R.; Wang, Z.; Govorov, A. O. Mid-Infrared Plasmonic Circular
Dichroism Generated by Graphene Nanodisk Assemblies. Nano Lett. 2017, 17 (8), 5099 --
5105.
(73) Fan, Z.; Zhang, H.; Govorov, A. O. Optical Properties of Chiral Plasmonic Tetramers:
Circular Dichroism and Multipole Effects. J. Phys. Chem. C 2013, 117 (28), 14770 -- 14777.
(74) Fan, Z.; Govorov, A. O. Plasmonic Circular Dichroism of Chiral Metal Nanoparticle
Assemblies. Nano Lett. 2010, 10 (7), 2580 -- 2587.
(75) Besteiro, L. V.; Kong, X.-T.; Wang, Z.; Hartland, G.; Govorov, A. O. Understanding Hot-
43
Electron Generation and Plasmon Relaxation in Metal Nanocrystals: Quantum and Classical
Mechanisms. ACS Photonics 2017, 4 (11), 2759 -- 2781.
(76) Hartland, G. V. Optical Studies of Dynamics in Noble Metal Nanostructures. Chem. Rev.
2011, 111 (6), 3858 -- 3887.
(77) Zhang, H.; Govorov, A. O. Optical Generation of Hot Plasmonic Carriers in Metal
Nanocrystals: The Effects of Shape and Field Enhancement. J. Phys. Chem. C 2014, 118
(14), 7606 -- 7614.
(78) Govorov, A. O.; Zhang, H.; Gun'ko, Y. K. Theory of Photoinjection of Hot Plasmonic
Carriers from Metal Nanostructures into Semiconductors and Surface Molecules. J. Phys.
Chem. C 2013, 117 (32), 16616 -- 16631.
(79) Kong, X.-T.; Wang, Z.; Govorov, A. O. Plasmonic Nanostars with Hot Spots for Efficient
Generation of Hot Electrons under Solar Illumination. Adv. Opt. Mater. 2017, 5 (15),
1600594.
(80) Johnson, P. B.; Christy, R. W. Optical Constants of the Noble Metals. Phys. Rev. B 1972, 6
(12), 4370 -- 4379.
(81) Barron, L. D. Molecular Light Scattering and Optical Activity; Cambridge University Press:
Cambridge, 2004.
44
Supplementary Information
Figure S1. The COMSOL model used in the calculations. A plasmonic dimer (colored here as
gold) was designed and its center of mass placed at the origin of coordinates, immersed in a 500
nm-radius water sphere (blue) and surrounded by a 200 nm-thick perfectly matched layer (PML,
gray) that suppresses reflections at the boundaries and thus recreate the conditions of an isolated
dimer. In the calculation, for the dimer, identically tetrahedral meshes were utilized for both
plasmonic nanorods. To obtain the averaged extinction and generation rates of hot electrons,
incident circularly polarized light was sent in orthogonal directions corresponding to the cartesian
axes (x, y, and z). Then, the results were averaged for the three directions of incidence.
45
Spectra for the Ag NR dimers. It is interesting to compare dimers composed of different
materials, and here we add data for Ag plasmonic NRs to complement the data for gold NCs
presented in the main text. Experimentally, Au-NRs are widely available, whereas Ag-NRs are
less common and comparatively harder to fabricate through wet-chemistry methods. As we pointed
out in the main text, the Ag dimers showcase a larger Rabi splitting, as one does expect from the
generally stronger and sharper plasmonic resonances present in good-quality silver.
46
Figure S2. Chiro-optical data for the Ag-NRs. Calculated chiral extinction spectra for plasmonic
dimers with 45° rotation angle (a-d), and 90° rotation angle (e-h). The optical data include the
extinction spectra for LCP and RCP light, the CD spectra and the g-factor. These calculations were
performed for the Ag-dimers with 5 nm and 13 nm gap sizes. The legend includes information
about material, rotation angle of NRs and gap size.
47
Figure S3. Optoelectronic (HE generation) data for the Ag-NRs under CPL. Calculated chiral
spectra for plasmonic dimers with 45° rotation angle (a-d), and 90° rotation angle (e-h). The data
include the HE generation rate spectra for LCP and RCP light, as well as the related CD spectra
and g-factor. These calculations were performed for the Ag-dimers with 5 nm and 13 nm gap sizes.
The legend includes information about material, rotation angle of NRs and gap size.
48
Mode intensity and spectral asymmetry in the Rabi-splitting regime. In the extinction spectra
in Figure 2 (see main text), we clearly observe that the 45° NR structure produces strongly
asymmetric spectra, whereas the 90° NR dimer exhibit a symmetric Rabi-split spectrum. We can
easily understand this interesting difference by looking at the total plasmonic dipolar moment in
our structures at the resonances. Figure S4(a) below shows that the 45° dimer has one weak and
one strong dipolar mode, while the 90° dimer has two equally strong bright modes. Figure S4(b)
shows the Rabi-splitting magnitudes as a function of the gap size between the two NRs for different
materials (Au and Ag) and rotation angles.
Figure S4: (a) Optical dipoles derived from the COMSOL-computed charge-density maps for both
chiral geometries at the two plasmon peaks. (b) Rabi-splitting, expressed as the spectral distance
49
between the peaks in units on energy, for different NR-NR complexes as a function of the NR-NR
gap.
Comparison of the extinction-CD and HE-CD spectra for the short-wavelength interval of
400-600nm. When we deal with the strong transversal resonance, the CD and HE-CD spectra
behave similarly, as expected, but at the region of interband transitions in Au the difference
between extinction-CD and HE-CD should be strong. Indeed, we observe this behavior in Figure
S5. The CD g-factor is mostly positive for in the interval 400-500nm, whereas the HE-CD g-factor
is negative in the same interval of 400-500nm. The physical reason for such difference between
the optical and HE CD and g-factor spectra is that the extinction CD in the interval of 400-500nm
in gold has a strong contribution from the interband absorption. Mathematically, absorption by a
volume V inside a NC is
. The case of HE-CD is different. The HE-CD
~ Im
Q
abs
V
Au
E
2
depends on the electric field directly and the corresponding dissipation,
does not include the wavelength-dependent factor Im Au .
Q
~HE
E
,
normal
2
,
S
Figure S5: Chiroptical spectra of the 45°-dimer for the short-wavelength interval. We see a strong
difference between the g-factors for extinction-CD and HE-CD.
50
Therefore, the wavelength dependencies are expected to be behave differently, as seen in Figure
S5.
Pulsed excitation regime. Our formalism (Eqs.4,5 in the main text) can also be used for the pulsed
regime of excitation. The average number of high-energy (over the barrier) HEs during a short
pulse can be evaluated as: S1
N
high energy,
Rate
high energy,
ee
(1
1
/
ee
e
t
/
t
ee
),
,
L R
(S1)
where t is the pulse duration. To obtain Eq. S1, we used a rectangular pulse. In Eq. S1, ee is the
average electron-electron scattering lifetime of high energy HEs, which can be derived from the
Fermi liquid theory:
ee
4
E
2
0 F
)
(
2
(S2)
0
128
3
2
1
p
128
3
2
m
*
0
ne
2
(
q
0)
where 0 is the intrinsic electron-electron scattering lifetime in bulk metal, p is the plasma
frequency, which is associated with the effective electron mass of bulk metal m*, the permittivity
of free space is represented as 0 , n is the electron density, and e is the elementary charge of
electron. Note that we considerp in the approximation of long wavelength (q≈0, with q being
the wavevector).
The estimate (Eqs. S2) for the average electron-electron (e-e) scattering lifetime ( ee ) is
strongly energy-dependent. In the visible and NIR wavelength ranges, the typical value of ee is
~10 fs, leading to
and
ee
t
N
high energy
ee
Rate
high energy
, obtained from Eq. S1. Qualitatively,
longer wavelengths are more favorable to obtain larger values of
N
high energy,
because of the
involved frequency dependences:
Rate
high energy ~ (
)
4
and
)
2
ee
(
.S1,S2
Then, the corresponding CD value for the average number of generated carriers is defined as
N
high energy
N
high energy,L
N
high energy,R
Although the calculations in the main text were done for steady-state generation and injection
of HEs, experiments in the pulsed regime are also very popular and relevant to understand the
51
dynamics of excited photocarriers.S1,S2 Here we evaluate the average number of high energetic
) during an ultrashort pulse with a duration of 80
HEs (N high energy) and the related CD (
N
high energy
fs. As a model system, we use again the Au-dimer with 45° rotation angle and a 5 nm gap. Our
calculations are based on Eqs. S1-S2 and shown below in Figure S6. More details for such time-
dependent formalism can be found in Ref. S1. Figure S6 shows the results. As it is also the case
for the CW illumination regime, we observe a CD effect in the optoelectronic responses. The CD
signal for the average number of HEs during the pulse time interval is shown in Figure S6b.
Figure S6. (a) Average number of generated hot electrons during the 80-fs pulse for LCP light
(red) and RCP light (blue), respectively. (b) CD of the average number of generated hot electrons
during the pulse. We show the case of Au-dimers with 45° rotation angle and 5 nm gap size. The
flux during the pulse in this calculation was 2.5×108 W/cm2, like in the experiment of Ref. S2.
52
Reference
[S1] M. E. Sykes, J. W. Stewart, G. M. Akselrod, X. T. Kong, Z. Wang, D. J. Gosztola, A. B. F.
Martinson, D. Rosenmann, M. H. Mikkelsen, A. O. Govorov and G. P. Wiederrecht, Nat.
Commun., 2017, 8, 986.
[S2] H. Harutyunyan, A. B. F. Martinson, D. Rosenmann, L. K. Khorashad, L. V. Besteiro, A.
O. Govorov and G. P. Wiederrecht, Nat. Nanotechnol., 2015, 10, 770 -- 774.
53
|
1911.02619 | 1 | 1911 | 2019-11-06T20:25:54 | Ionically gated perovskite solar cell with tunable carbon nanotube interface at thick fullerene electron transporting layer: comparison to gated OPV | [
"physics.app-ph"
] | We demonstrate an ionically gated planar PS-PV solar cell with ultra-thick fullerene ETL with a porous CNT electron collector on top of it. Perovskite photovoltaic devices usually have undoped electron transport layers, usually thin like C60 due to its high resistance. Metallic low work function cathodes are extremely unstable in PS-PV due to reaction with halogens I-/Br-, and it would be desirable to have stable carbon cathodes on top of thick low resistance ETL for enhancing the stability of PS-PVs. We show that gating such top CNT cathode in ionic liquid, as part of a supercapacitor charged by Vg tunes the Fermi level of CNT by EDL charging, and causes lowering of a barrier at of C60/C70 ETL. Moreover, at higher gating voltage ions further propagates into fullerene by electrochemical n-doping, which increases dramatically PV performance by raising mostly two parameters: Isc and FF, resulting in PCE efficiency raised from 3 % to 11 %. N-doping of ETL strongly enhances charge collection by ETL and CNT raising Isc and lowering series resistance and thus increasing strongly PCE. Surprisingly Voc is not sensitive in PS-PV to external Vg gating, on the contrary, to strongly enhanced Voc in ionically gated organic PV, where it is the main gating effect. This insensitivity of Voc to lowering of the work function of Vg gated CNT electrode is a clear indication that Voc in PS-PV is determined by inner p-i-n junction formation in PS itself, via accumulation of its intrinsic mobile ionic species halogens and cations and their vacancies. | physics.app-ph | physics | Ionically gated perovskite solar cell with tunable carbon nanotube interface at thick fullerene electron
transporting layer: comparison to gated OPV
D.S. Saranin1,2, D.S. Muratov3, R. Haroldson4, A. G. Nasibulin5, A.R. Ishteev1,3, D.V. Kuznetsov1,3, M.N. Orlova2, S.I. Didenko2 and
A.A. Zakhidov1,4
1Energy Efficiency Center, National University of Science and Technology MISiS, Moscow, 119049, Russia
2Department of Semiconductor Electronics and Device Physics, National University of Science and Technology MISiS, Moscow, 119049, Russia
3 Department of Functional Nanosystems and High Temperature Materials, National University of Science and Technology MISiS, Moscow,
119049, Russia
4Physics Department and The NanoTech Institute, The University of Texas at Dallas, Richardson, 75080, USA
5Center for Photonics and Quantum Materials, Skolkovo University of Science and Technology, Moscow,143026, Russia
Abstract
We demonstrate an ionically gated planar PS-PV solar cell with ultra-thick fullerene ETL with porous CNT electron collector on top
of it. Perovskite photovoltaic devices usually have undoped electron transport layers, usually thin like C60 due to its high
resistance. Metallic low work function cathodes are extremely unstable in PS-PV due to reaction with halogens I-/Br-, and it would
be desirable to have stable carbon cathodes on top of thick low resistance ETL for enhancing stability of PS-PVs. We show that
gating such top CNT cathode in ionic liquid, as part of a supercapacitor charged by Vg tunes the Fermi level of CNT by EDL charging,
and causes lowering of barrier at of C60/C70 ETL. Moreover, at higher gating voltage ions further propagates into fullerene by
electrochemical n-doping, that increases dramatically PV performance by raising mostly two parameters: Isc and FF, resulting in
PCE efficiency raised from 3 % to 11 %. N-doping of ETL strongly enhances charge collection by ETL and CNT raising Isc and lowering
series resistance and thus increasing strongly PCE. Surprisingly Voc is not sensitive in PS-PV to external Vg gating, on the contrary
to strongly enhanced Voc in ionically gated organic PV, where it is the main gating effect. This insensitivity of Voc to lowering of
work function of Vg gated CNT electrode, is a clear indication that Voc in PS-PV is determined by inner p-i-n junction formation in
PS itself, via accumulation of its intrinsic mobile ionic species halogens and cations and their vacancies.
Introduction
Development of perovskite photovoltaics has started from dye-sensitized solar cells (DSSC) conception in pioneer works1
by using a metal organic semiconductor as a promising absorber instead of usual dye sensitization. Later, as perovskite solar cells
(PSC) began presenting a separate scientific direction of photovoltaics,2 DSSC structures gradually gave way to mesoscopic and
planar configurations3,4 due to advantages of perovskite semiconductor properties in thin film device application: ambipolar
transport,5 -- 7 suppressed recombination,8 -- 10 big diffusion length,11 direct12,13 and easily tunable band gap.14,15
Record PCE of more than 25.2%16 was demonstrated for perovskite solar cells fabricated on n-i-p mesoscopic
architectures, which require mesoporous and compact TiO2 film. Despite the fact that such a device structural concept was realized
in large-scale printing techniques of DSSC and perovskite n-i-p modules,17 -- 19 this configuration limits the flexible lightweight device
application due to the high temperature sintering process needed for TiO2 (up to 450°C). Hole transport layers for most high-
performing n-i-p devices were fabricated with small molecule Spiro-Ometad (2,2',7,7'-Tetrakis[N,N-di(4-methoxyphenyl)amino]-
9,9'-spirobifluorene),20 which should be p-doped to increase the life time21 and enhance the conductivity of the hole.22 PSC can
display high PCE with various Spiro-Ometad HTL thicknesses ranging from the common 200 nm to an anomalistic 600 nm,
depending on the doping level to avoid surface recombination, form high open-circuit voltage and use rough perovskite absorber
films for higher level of absorption23. On the other hand, doping impurities (widely used Li-TFSI, tBP) are the main reason for
interface degradation and device operation -- instability,24 such as accelerated photo oxidation.
Another interface problem is from the perovskite side, which has high defect density at interfaces due to accumulation
of mobile iodine vacancies, excessive ions,25 and ferroelectric poling with slow polarization.26 One promising effort for perovskite
interface stabilization is the use of fullerene based passivation layers27 -- C60, C70, PCBM, and others. It was shown27 that fullerene
passivation can effectively increase the surface recombination lifetime with total decreasing of surface trap density by capping
perovskite grain boundaries and C60 penetrating into the bulk between grains. For p-i-n (inverted) PCS, the fullerene thickness
range for device operation feasibility lays is between extremely thin 2.5 nm to the commonly used 50 nm28. Devices with C60 ETL
thicknesses below 25 nm usually show low shunt resistance with a reduced short circuit current and filling factor. Fullerene layer
with thicknesses of 30 -- 50 nm lift the overall PCE to high performance level, but the concept of a thick, robust, passivation electron
transport layer with suitable conductivity as analogue to doped Spiro-Ometad HTLs is highly desirable but not yet presented,
However, it can be potentially realized with co-evaporation techniques using n- type dopants like PhIm, acrydine orange, CoCp2,
etc,29 -- 31 similar to OLED technology but these methods require expensive materials and complicated technological processes.
Electrode interface also suffers from instability factors: metal diffusion into transport layers and perovskite absorber,32
appearance of bubbles in metal films during evaporation,33 and halogenic oxidation.34 Most prospects for metal replacement focus
around carbon because this material can be printed or laminated without vacuum processes, stabilize the interface, and
demonstrates hydrophobic properties. Currently, most stable PSC are fabricated with carbon electrodes;35 moreover, carbon can
be effectively used for semitransparent solar cells in forming nanotubes. Several papers show competitive results of cells with
single-walled CNT anodes,36,37 but this material is still being pursued for metal conductivity and Wf engineering.
A promising effort for improving the electrode-transport layer junction is interface engineering and accumulation of
charge carriers. Electrode and transport layer energy levels can be aligned via interfacial polarization and dipoles caused by a
buffer layer based on ionic liquids. As shown in works with organic solar cells by Yu,38 Zhang,39 and Kang,40 using ionic liquids at
the cathode interface improved electron collection with metal oxide and polymer transport layers by reducing electrode Wf and
forming an ohmic contact. Ionic liquids (IL) are room temperature melting salts usually presented by long chain organic cations
and inorganic anions such as BF4, Cl, PF6, and TCA with a wide electrochemical window of stability. Using ILs does not require
vacuum deposition of insulators and can be processed with solution techniques. The simplification of fabrication processes has a
wide range of approaches in energy related, thin film devices with liquid-solid interfaces.
The gate field effect is another way to improve charge collection or accumulation in thin film solar cells. Cook41 and co-
workers have shown tunable OPV with ionically gated CNT cathode electrode, in which Fermi level of CNT raised significantly,
decreasing w.f. by 0.5 eV by negative gate voltage Vg ~ 1.5 V. Causing dramatic increase of all parameters of OPV: Voc from 0.1 V
to 0.6 V, FF raising to 0.7 and resulting in significant PCE increase. Zhou and co-authors42 designed and fabricated OPV cells with
a cathode interface gate with significant improvement in charge extraction (+24 % to Jsc) under gate bias. A more advanced concept
was presented in a paper by Karak41, where electron extraction was achieved in the AC field with an ionic liquid cathode interfacial
layer, Additionally, charge accumulation is an effective solution for electro physical property tuning of transport semiconductors.
This area of study is the main problem in developing thin film FET (field effect transistors) for modern electronics based on organic
semiconductors, metal oxides and carbon derivatives (fullerenes, nanotubes, graphene). Ionic liquids can form a double electric
layer (EDL) with applied bias and act as a gate due their specific capacity (can reach more than µF/cm2 versus nF/cm2 order in
SiO2), which influences the quantity of induced carriers at low gate voltages <5 V 43. Ionic gate is presented in various types of
super capacitors, fuel cells, dye sensitized solar cells, and field effect related devices. And recently ionic gating was applied even
in OLED and OLET devices, making them brighter and more efficient by enhanced charge injection at gated interface44.
In this work, we found motivation through a combined approach of interface engineering with ionic liquid and n- type
accumulation of charge carriers by ionic gating under Vg applied to gate. Carbon nanotubes as a high-specific surface network is
an advantageous concept for ionic gate. High surface of materials gives possibility to induce higher concentration of carriers in
electrostatic regime. In this case, carbon network can act as effective charge plate for a supercapacitor based on ionic liquid. In its
turn, such gate concept can be effectively used in three-interface interaction structure, when supercapacitor is placed on
semiconductor surface and ionic liquid penetrates through CNT porous web to semiconductor surface and then to its bulk by
diffusion/drift.
Herein, we present a novel type of inverted perovskite planar cell with ultra-thickness (up to 300 nm), a robust fullerene
electron transport layer (ETL), and a tunable cathode interface based on highly porous carbon nanotubes sheets. We have
developed a horizontal architecture, then fabricated and characterized the p-i-n structure with horizontal CNT ionic gate placed
at back electrode. In this device, high specific resistance of the undoped ETL layer and the series resistance on the electrode
interface are reduced by charge carrier accumulation regimes. The use of the ionic gate is demonstrated as an effective, low
voltage tool for improving solar cell output parameters in reversible electrostatic regimes. We compared device structures with
two fullerene ETLs -- C60 and C70 ranging from 200 -- 300 nm thicknesses at different gate voltages finding correlations to output PV
parameters.
(a) (b)
(c) (d)
Figure 1. (a) Device schematics of inverted planar MAPbI3 cell with ionic gate between SWCNT cathode and MWCNT
electrode, (b) Device band diagram,(c) Device top view,(d) device cross side view
Experimental part
Ink preparation
Perovskite ink was fabricated in 1.5 M concentration from methylamine iodide (MAI, Dyesol) and lead iodide (Alfa Aesar). Firstly, MAI was
dissolved in mixture of DMF: γ-BL (1:1 volume ratio, from Sigma Aldrich, anhydrous), and then PbI2 was dissolved in MAI solution with a 10%
volume addition of DMSO. Ink was heated overnight at 60 °C and cooled to room temperature in 5 minutes before device fabrication. PEDOT:
PSS water dispersion from Hereaus Clevious (1,4 % Al 4083) was filtrated through 0.45 um PTFE filter prior HTL spin coating.
Device fabrication
Pixelated ITO substrates (Lumtec, 15 ohm/□) were cleaned in an ultrasonic bath with acetone, toluene, and IPA (ME grade). Substrates were
preliminary activated by 20 minutes UV-ozone treatment for PEDOT: PSS HTL deposition. A 30 nm layer was spin-coated at 3000 RPM during 60s
and then annealed at 150 °C. A CH3NH3PbI3 450 nm photoactive layer was formed after a 2-step spin coating process: 20 s at 1000 RPM and 25
s at 4000 RPM and toluene dripping procedure (10s before 2nd step ending) with final annealing at 100°C during 10 minutes (all processes were
provided in glovebox with inert atmosphere < 1 ppm O2, <1 ppm H20). C60 and C70 electron transport layers (200,250,300 nm thicknesses) were
thermally evaporated at 2*10-6 Torr with 0.5 A/s rate.
SWCNTs (synthesized accordingly, route described in previous work) were laminated as 3 mm stripe on the top of device structure and densified
with HFE.
MWCNTs (synthesized accordingly, route described in previous work) were laminated as 3 mm stripe on the top of device structure and densified
with HFE.
Ionic gate deposition
MWCNT as counter electrode was laminated (5 layers) in parallel to the SWCNT cathode in 3 mm distance and densified with HFE.
Finally, a drop (2 µl) of DEME -BF4 ionic liquid (Kanto Chemicals) was squeezed between the SWCNT and MWCNT electrodes by glass coverslip
on top of the warm wet planar layer.
Preliminary, 50 and 100 nm ETL devices were tested, but ionic liquid interpenetrated the perovskite layer under pressure of cover slip and
partially dissolved photoactive film. For the 50 nm, the ETL layer device degraded in first minute of testing; for 100 nm, the ETL device started
to degrade after 5 minutes of measurement and showed poor performance (presented in supplement material). Therefore, thick fullerene film
also acts as a protection layer for perovskite in devices with ionic gate.
Characterization
JV characterization were provided at standard 1.5 AM G 100 mW/cm2 conditions with Newport ABB solar simulator (calibrated with Si certified
cell) and two Keithley 2400 SMU (JV sweep, gate applying) in inert atmosphere.
Ionic gate JV Sweeps
Quantum efficiency measurements were provided on XP-6 system (PV MEASUREMENTS) calibrated with KG-9 Si reference cell.
Raman characterization was done on Thermo DXR Raman microscope with 532 nm laser.
Results
The output performance of
fabricated cells was measured
in
two
regimes: without
ionic
liquid
in
ITO/HTL/Perovskite/ETL/SWCNT device structure and with SWCNT/DEME-BF4/MWCNT ionic gate at different biases. Gate voltage
was applied with positive connection to the counter electrode and negative to the SWCNT device cathode for respective inducing
of DEME+ ions at SWCNT-ETL interface and BF4
- ions at MWCNT counter.
The presence of cation and anion induce polar charge accumulation (electrons and holes respectively) at semiconductor
interface provided a doping effect and Fermi level shift in electrostatic regime. As described in our previous works45,46 in OPV
devices, high concentration of cations at carbon nanotube-acceptor interfaces induced by gate voltage raises the Fermi level, as
shown in the band diagram of Figure 2(a). At initial conditions, JV curves of cells without ionic liquid gate show a strong S-shaped
character caused by high contact resistance between SWCNT cathode and thick, intrinsic C60 and C70 ETLs, as shown in Figures 2
and 5(g -- h). For both fullerene materials, such ETL thicknesses between 200 -- 300 nm affect device output performance with series
resistance values from 2000 Ohm*cm2
(C60, C70 200 nm thick) to >12000 Ohm*cm2
(C60, 300 nm) and >7500 Ohm*cm2 (C70,
300 nm) due to the intrinsic low conductivity >1013 *cm47. Accordingly, typical thicknesses of fullerene based ETLs in perovskite
solar cells range 15 -- 50 nm. Moreover, sheet resistance of SWCNTs48 used for electrodes did not reach the level of typical metal
electrodes and conductive oxides49. As a result, the devices' FF values are below 0.25, less than that for a resistor-like curve. JV
behavior changes dramatically with appearance of SWCNT cathode-ionic liquid interface and n-type accumulation accordingly to
gate bias value. Firstly, initial contact between DEME-BF4 with SWCNT cathode without applied bias has dropped series resistance
for both fullerene ETL to more than order for 200,250 nm thicknesses and >5 times for 300 nm thick C60 and C70 ETLs cells.
Sequential increasing of gate bias from 0.00 V to 1.00 V reduced series resistance to acceptable values of <80 ohm*cm2
for C60
and <40 ohm*cm2
for C70 ETL devices and transferred s-shaped curve to diode-like JV with 0.40 -- 0.50 FF values. Next slight Vgate
increasing with 0.25 V step (from 1.00 V to 2.50 Vgate) showed same trend in Rs reducing to level below 20 Ohms*cm2. The highest
FF values were obtained at gate biases >1.50 V in range 0.6-0.7 for 200 -- 300 nm C60 samples and 0.5 -- 0.6 for 200 -- 300 nm C70
samples at Vgate >1.75 V. A big difference between C60 and C70 ETL devices was obtained in current generation behavior with
dependence to thickness of layers. Output performance of cells differs in dynamics of photocurrent increments with increase of
Vgate bias. This distinction is clearly observed in Figures 5(a, b), where Jsc values are shown like function of Vgate. Spread of Jsc gain
for C60 ETL devices have more expressed dependence to ETL thickness with increasing of Vgate than for C70 cells. We have found
that Jsc gate response is different at range of bias 1.00 V; 1.00 > 2.00 V; and >2.00 V for different thicknesses. The 200 nm ETL
devices have not shown significant gain of Jsc with increasing gate biases from 0.00 V t0 1.00, for both fullerene types. The C60 ETL
cell had Jsc growth only +0.56 mA/cm2 from 13.11 mA/cm2, while the C70 ETL cell had gain of +0.60 mA/cm2 from 18.35 mA/cm2
(<4 % in both cases). On the other hand, relative contribution of ionic gating for Jsc growth with thicker ETL is much higher, +34.8%
and +41.3% for 250 nm and 300 nm C60 films; +31.9 % and +25.3 % for C70 films respectively. At higher gate bias range 1.00 >
2.00 V, C70 ETL devices have approximately the same dynamics of Jsc increment, as response to the increase of Vgate with all used
thicknesses, and much lower spread of values in comparison to C60 ETL cells in dependence to thicknesses. At gate voltages >2.00
V, no meaningful Jsc was observed. Total improvement in output performance has impressive indicators. The use of ionic gate, as
a tool for tuning cathode-ETL interface, increased the PCE of devices more than five times, on average. As shown in Figures 5(e,
f), C60 samples improvement is 2.46% to 11.28 % (200 nm); 2.05% to 8.67 % (250 nm); 0.72% to 6.55 % (300 nm), and C70 is 3.35%
to 9.74 % (200 nm); 1.47% to 10.74 % (250 nm); 0.72% to 10.31 % (300 nm).
(a)
Output performance improvement of 200 nm thick C60 cell
from S -- shaped JV curve to diode-like characteristics at
0…1.00 V gate with sequential power increase up to 1.75
Vgate
(b) Output performance of 200 nm thick C60 cell at Vgate>2.00 V
with slightly decreasing of Jsc
(c) Output performance improvement of 200 nm thick C70 cell
from S -- shaped JV curve to diode-like characteristics at
0…1.00 V gate with sequential power increase up to 2.00
Vgate
(d) Output performance 200 nm thick of C70 cell at Vgate>2.00 V
with slightly decreasing of Voc
Figure 2. JV curves of MAPBI3 cell with 200 nm thick C60 (a)(b) and C70 (c)(d) ETL 0 to…2.50 V range of gate bias
(a) Short current density V gate dependence of 200-300 nm thick C60
cell
(b) Short current density V gate dependence of 200-300 nm thick C70
cell
(c) Filling factor V gate dependence of 200-300 nm thick C60 cell
(d) Filling factor V gate dependence of 200-300 nm thick C70 cell
(e) Power conversation efficiency V gate dependence of 200-300 nm
thick C60 cell
(f) Short current density V gate dependence of 200-300 nm thick C70
cell
(g) Series resistance V gate dependence of 200-300 nm thick C60 cell
(h) Series resistance V gate dependence of 200-300 nm thick C70 cell
Figure 3. JV Output parameters V gate dependence versus ETL thickness
During quantum efficiency measurements in ionic gating regimes, we obtained the same trend of device performance
improvement. Quantum efficiency spectrums have stable growth with increasing Vgate for both types of fullerene films and
thicknesses (external quantum efficiency spectra of 200 nm C60 and C70 ETL devices presented in Fig. 4, spectra for 250 and 300
nm presented in supplementary). EQE spectra are typical for MAPbI3 perovskite solar cells with PEDOT: PSS and C60-C70 transport
layers. Considerable changes are clearly observed in increment of photon quantity converted to electricity. The level of photon
conversion grows with same dynamics as short current density during gating in JV testing. As is presented in Figure 4(a), EQE
spectra shift upward with bigger gain at 1.00 -- 2.00 V gate bias for the C60 sample and have less response to gate voltage for the
C70 cell (Figure 4(b)). In the maximum point of spectra for 200 nm C70 cell, EQE was achieved at 80%, and that is 40% higher than
cells at initial conditions without ionic liquid gate, respectively for C60 ETL devices, maximum was achieved at 65%. The level of
photon conversion decreases more than 5 -- 10% at Vgate>2.00 V [Figure 4(b), 5(b)] with the appearance of QE shoulder in near UV
region; it corresponds to perovskite absorber degradation and will be discussed in the next chapter. In general, growth of the EQE
spectra follows the trend of Jsc changes of solar cells during gating, and confirms improvement of charge collection.
(a) Growth of light conversation EQE level from 350 to 850 nm
wavelength at 2.00 V (200 nm thick C60 cell)
(b) Changes of light conversation EQE level from 350 to 850 nm
wavelength at ≥ 2.00 V (200 nm thick C60 cell)
(c) Growth of light conversation EQE level from 350 to 850 nm
(d) Changes of light conversation EQE level from 350 to 850 nm
wavelength at 2.00 V (200 nm thick C70 cell)
wavelength at ≥ 2.00 V (200 nm thick C70 cell)
Figure 4. External quantum efficiency spectrums of MAPBI3 cell with 200 nm thick C60 (a)(b) and C70 (c)(d) ETL at
0…2.50 V gate bias
Analysis of fullerene-SWCNT-ionic liquid interface operation was done in separate structures for JV measurement of
semiconductor junction. ITO/C60 (C70)/SWCNT-DEME-BF4-MWCNT samples were fabricated in the same route as layers in solar
cells, respectively. As presented in Figures 6(a, b) for 200 nm C60 and C70 films, initial resistance of SWCNT-fullerene junction is
two orders more than resistance at 2.5 V of gate voltage at ionic liquid gate. This means that n-type accumulation regime
significantly increased the ohmic of electrode-semiconductor contacts and provided energy levels matching between SWCNT work
function and fullerene LUMO. In addition, JV curve behavior changed two times during the ionic gating process. Initial conditions
without an ionic liquid JV curve had a resistor-like type of an approximate straight line (accordingly to Ohm law). Then, asymmetry
of forward (V>0) and reverse JV curves (V<0) appeared during increasing gate voltage. The slope of the forward curve grew much
higher in comparison to the reverse curve, which, in opposition, moved to axis line. For both ETL materials, the maximum relative
between high current values of forward curve and minimum values of reverse curve were achieved at 1.5 gate bias. Valve ratios
were calculated as 10 for 200 nm C60 film and 4 for C70 respectively. Moreover, until 1.5 V at gate, reverse current decreased
dynamics or tendency to leakage removing. We suggest that such JV behavior changes result from the i-n junction appearing in
accumulation regime under electrostatic gating. At gate biases >1.50 V, reverse curves swiftly gained the slope and formed a
resistor-like JV graph with much smaller differential resistivity in comparison to initial conditions. We believe this effect is
consequently due to inducing the n-type doping for the whole depth of ETL. Decreasing of Voc in devices with occurs due solvation
of fullerene ETLs in IL in analogue to effect described in work of Maciel50 and have long-term time dependence.
(a)
JV behavior changes from resistor (till 1.0 V gate) to diode like
curve (at 1.5 Vgate) of ITO/C60 200 nm/SWCNT-IL gate structure with
sequential reverse to resistor characteristic with lower differential
resistance
(b)
JV behavior changes from resistor (till 1.0 V gate) to diode like
curve (at 1.5 Vgate) of ITO/C70 200 nm/SWCNT-IL gate structure with
sequential reverse to resistor characteristic with
lower differential
resistance
Figure 5. C60 (a) and C70 (b) diode characteristics with gate bias dependence
Discussions
Large progress in PSC performance improvement was achieved with development of doping techniques of thick polymer
hole transport layers like Spiro-Ometad, P3HT, PTAA, etc., but not many works tended to analogue approaches with organic
electron transport materials. In this work, we show that thick fullerene layers >200 nm can be used as electron transport layers
for perovskite solar cells without complicated doping methods like co-evaporation. The use of ionic gate in advanced planar device
architecture compensates initial low conductivity of thick ETL and non-metal electrodes in the accumulation charging regime.
Electrostatic inducing of n-type carriers at ETL-cathode interface improved mismatching energy levels and provided a significant
gain of charge collection in initially intrinsic semiconductors (C60/C70, SWCNT).
The use of horizontal ionic gate was addressed for managing three critical points in devices with thick ETL and undoped
SWCNT electrodes: first, high sheet resistance of cathode; second, high resistance of thick ETL; and third, not optimal fullerene
LUMO-SWCNT Wf matching. For optimal electron collection and provision of ohmic contact, zero barrier between the transport
layer and electrode is required. The measurement of the SWCNT's work function is imprecise due containment of metallic and
semiconductor parts and approximate acceptance equal to 4.7 eV, according to data from the literature51. In this case, initially the
ETL-cathode junction has 0.4 eV of energy loss caused by mismatching of ETL LUMO and electrode Wf. On the other side, high
ohmic fullerene films cannot provide high electric current flow with 60 Ohm/ SWCNT, and this significantly reduces the slope of
the IV curve. Such affection was clearly observed in JV behavior of tested solar cells and diodes in structures without ionic gate.
Therefore, formulation of problem for junction was defined as necessity of SWCNT Ef raise and increasing of local conductivity in
ETL via higher concentration of charge carriers (n-type doping in both cases). According to this statement, the device's operation
mode should provide accumulation of positive ions (cations) to induce carriers with opposite charge at back electrode interface.
During negative charge accumulation at SWCNT-DEME+
cation, the interface Fermi level of SWCNT starts to shift up, providing better
work function and matching with LUMO of fullerene. In turn, the ETL surface is also working under gate accumulation regime via
soaking by ionic liquid through CNT network, and near-surface layers of fullerenes become n- doped in electrostatic mode. The
level of accumulation was controlled by gate voltage (0.00 -- 2.50 V), and in turn, high concentrations of induced carriers are
provided via the high capacity of ionic liquid43 and the high specific area of the SWCNT acting as mediators for positive ions. The
charging process of ion induction and distribution on the SWCNT cathode and MWCNT counter was determined as a time
dependent process between condenser plates. Cathode-counter current flow has an initial level of pA without gate voltage and
falls with hyperbolic dependence from 100 nA to saturation at 1 nA after applying the gate bias. Therefore, such additional
power loses (nW) for ionic charging has no critical contribution to the device operation. Typical charging I(t) plots presented in
supplement material.
There is a very significant qualitative difference in the effect of Vg on Voc of perovskite PV as compared to our earlier results on
ionic gating in Polymeric PV52 and small molecule PV 53: in both later cases the strongest effect was the increase of Voc: from 0.1
V to 0.5 V in Poly-PV, and similar in small molecule PV. As shown at Fig.6 of comparison below: It reflects Voc related to increase
of Fermi level in SWCNT with EDLC charging of CNT in all OPVs. On the contrary in Pero-PV the Voc do not change at all, but only
shape of IV curve is changing at Voc, reflecting the strong change of series resistance, Rs (which is the slope of dV/dI at Voc point).
This is a clear indication that the w.f. of electrode is not important for PS-PV, since the internal p-i-n junction quickly forms inside
PS, as sketched in Fig X. this raises Fermi levels in n-doped part of PS layer adjusting it to Fermi level of electrode CNT,
independently where this level is initially. So as in all PS PV devices, the Voc does not depend much on w.f. of electrode, once the
internal p-i-n is formed by dynamic photopoling26,54 or poling55. Therefore, the most important effect of Vg is the increase of
photocurrent Isc due to better charge collection by n-doped ETL and lower series resistance Rs of the device becoming smaller
with Vg increase causing increase of FF.
Figure 6. Schematic operation of PS-PV tuned by gating: a) at small Vg and upon light the PV operation is effective mainly by
generating larger Isc with better FF, upon Vg raise, while Voc is not changing at all, since Vg is determined by inner p-i-n junction
in PS, shown as intrinsic ions accumulated in PS at interfaces. Larger Vg> 2 V causes strong doping of ETL into n+ by plus ions of
Ionic liquid DEME+ accumulation in C60 and lowering both barriers at C60/CNT and PS/C60 interfaces. However this processes do
not effect much neither Isc, nor FF, since charge collection cannot be further much improved. B) the equivalent circuit of gated
PS-PV showing a supercapacitor SC on top of ionically tunable photodiode I-PV. Charging of SC can be enhanced by self-charging
by Voc, or even solely done by Voc, which is increased by PIN formation in PS layer.
For comparison, we present the results of IV curves change with Vg for a similar Vg ranges in OPV (as originally presented in work
of Voroshilov and co -- authors53): S-shape changes to good FF even with Vg=0, due to polarization by EDL formation at interface
(as in our case). Then further doping at higher Vg increases Voc due to Fermi level raise of SWCNT electrode, till the CB of ETL,
when the conductivity of SWCNT.
So, the great and unusual observation is that Voc is not increased at all, although the Fermi level is known to raise in
SWCNTs. This means that w.f. of SWCNT is not important for Voc, it already was high due to PS physics. And this physics is now
showing its nature in this gating difference.
Fig. 7. IV curves of two typical OPV upon Ionic gating, demonstrating strong increase of Voc and Isc
In OPV the main effect is the dramatic change of Voc as shown in Fig. Y for two types of OPV: PHT/PCBM bulk heterojunction (from
the work of Cook56 and CuPc/C60 small molecule OPV (from work with ionic gating for small molecule OPV)53.
This insensitivity of Voc on ionic gate Vg in PS-PV is an indirect prove that in PS the inner pin junction is formed, via the
accumulation of its intrinsic ionic species, as shown at Fig. 7 above. In case of PS-PV the strong increase of Isc is due to improved
collection of carriers by the N-doped ETL, and partially by lowering the series Rs.
The amazing second difference of PS-PV from OPV is the insensitivity of operation on Vg at higher Vg > 2.0 -- 25 V. this is again can
be understood as due to n+ heavy doping of C60.C70 at higher Vg, which only slightly increase charge collection, and do not
degrade the PV operation, like in OPV, as observed by Saranin56 in gated OPV tandems at high Vg. The interplay of inner p-i-n
junction in PS layer with ions coming from ETL at higher Vg is an interesting challenge, that needs further investigation.
Mainly C70 and C60 ETL based devices have different levels of current generation and output performance response to
gate voltage in dependence to thickness. At initial conditions, devices with 200 nm C70 based ETL have 1.5 times bigger short
current density 16 mA/cm2 in comparison to 12 mA/cm2 of 200 nm C60 based ETL cells (as it shown on Figure 2(a)(c)). The
difference of Jsc values is determined to higher absorption of C70 films in 450 -- 600 nm visible range of light, and gives its
contribution in photon harvesting and larger photocurrent due the larger quantity of generated excitons, as shown in works with
OPV devices and planar PSC.57,58 With an increase in the thickness of the electron transport layer from 200 to 300 nm, a difference
between Jsc performance of C60 and C70 devices becomes more obvious. Different gate bias response of devices is explained due
energy structure, surface state, and interfaces between fullerene ETL, SWCNT cathode, and induced positive ions-DEME+. First, a
significant drop of series resistance was observed after deposition of ionic liquid on the cathode surface without applied gate bias.
In turn, decreasing the contact resistance corresponds to doping effects, which occur at room temperature between the SWCNT-
fullerene-IL interfaces. Carbon nanotubes-IL interfaces can have strong interaction via Van-Der-Waals bonding, - bonding, and
sidewall adsorption, as it presented in several reviews with modelling and experimental results59,60.
At the same time, fullerene ETL-IL interface has another interaction in physics. Theoretical calculations, provided by
Chaban and co-workers,61 show that a very significant polarization effect appears between the room temperature ionic liquid and
the C60 molecule. With use of the hybrid density functional theory (HDFT) and powered Born-Oppenheimer molecular dynamics
(BOMD) simulations, authors showed that ion adsorption at C60 surface acquires systematically positive electrostatic charges 0.1 --
0.2e with imidazolium IL. Moreover, conduction and valence band orbitals are shared between fullerene and ions of ionic liquid
as result of polarizing action. Semiconductor energy levels can be shifted up or down in the presence of different anions of ionic
liquid (authors showed LUMO raise with Cl-, NO3
- ions and drop with PF6
- anion), with band gap tuning. Therefore, we suggest,
that initial contact between C60/C70 ETL with BF4
- anion shifted down LUMO with improving of energy level matching with SWCNT
Wf. As result, such polarizing action can give contribution to initial drop of contact resistance between ETL and cathode during IL
cathode soaking. Applying this model to our case with three-side interface, direct charge transfer can be done in accumulation
regime at Vgate2.00 V, when ETL LUMO and SWCNT Wf have equal energy level. As It shown in JV gate dependence of ETL film
diode structure, asymmetry of forward reverse curve corresponds to n- type accumulation in depth of ETL with sequential
decreasing of differential resistance and rectification of JV curve at Vgate>2.00 V. Thereby, induced i-n junction and expansion of
n-doped area in ETL with increasing Vgate have improved charge collection and current generation due lower LUMO position and
higher concentration of carriers. Advantage of junction in transport layers for PSC was presented already in a work by Jung at al.,62
where authors developed n-i-p Spiro-Ometad HTL with co-evaporation doping methods. N-type doped spiro-OMeTAD was adjusted
to perovskite HOMO for efficient hole extraction while a p-type layer formed optimal level matching with the gold electrode. Response
of output parameters from gate bias for C60 and C70 ETL devices caused by different energy positions of LUMO levels and molecule
arrangement with different carbon bonds curvature. It is well known that C70 have lower a position in comparison to C60,
approximately 4.2 eV to 3.9 eV57,63, hence, C70 ETL initially has better alignment with SWCNT Wf and lower potential losses.
Consequently, C60 ETL requires a higher-level pf n- type accumulation for level matching with cathode Wf and electron extraction.
C60 molecule is buckminsterfullerene or buckyball, and its packing changes gradually from hexagonal to a cubic form, while C70
elongated fullerenes have a rugby ball-like shape, and prefer the hexagonal packing64. Different surface curvature due to chemical
bonding and molecule packing allows for specific ion adsorption and distribution under gate bias, which influences the depth of
charges accumulation.
Conclusions
In summary, we demonstrated a combined approach of interfacial engineering and n-type accumulation in advanced
planar p-i-n perovskite solar cell with ionic gate and ultra-thick ETL based on fullerene. Operation capability of such thick 200 -- 300
nm fullerene-based ETL is presented for the first time for perovskite solar cells. Initial high series resistance of SWCNT with thick
200 nm ETL interface was dropped from extremely high kOhms*cm2 values to 20 Ohm*cm2 due to dipolar polarization by
DEME+ and BF4
- ions and Vg gate bias induced n-type doping of CNT/ETL via injected carriers. Sequential increase of gated bias
successfully transformed SWCNT-fullerene Shottky contact to ohmic junction while simultaneous initial intrinsic and high ohmic
nature thick 200 -- 300 nm C60/C70 fullerene films were compensated by charge accumulation in ETL depths. Appearance of i-n
junction (induced by gate bias) was confirmed by diode structure, JV measurements with presence of rectification ratio at Vgate ≥
1.5 V.
The different response of JV performance to gate bias and ETL thickness was observed for C60 and C70 devices. This
effect occurs due to the higher curvature of C70 molecules packing that allows better IL cations-anions adsorption and distribution
under the applied field, and initially lower LUMO position, which provides matching SWCNT Wf at lower gate bias. Therefore, C70
ETL have lower spread of Jsc
in dependence to thickness with increased Vgate, while C60 ETL devices have brighter expression of
dependence to layer thickness.
Finally, devices showed dramatic improvement of output parameters with changes starting from the S-shape JV curve to
11+ % PCE performance. Best efficiencies were demonstrated on 200 nm C60 cells and 250 nm C70 devices at Vgate = 2.25 V and
2.50 V, respectively. For C60, most performing cells Jsc gain was + 54.2%, FF increased in 2.7 times, and efficiency grew from 2.46%
to 11.29%, respectively, and for C70, Jsc improved +44 %, FF increased in 3 times, and efficiency grew from 1.47% to 11.13%.
Using ionic gate for cathode junction opens new perspectives for improving electron transport materials of various
thicknesses and promising materials for electrode application with ambient, vacuum-free processing.
Acknowledgements
Support for this work was provided from the Ministry of Education and Science of the Russian Federation in the framework of
Increase Competitiveness Program of NUST "MISiS" (No. K2-2019 -- 014). Partial financial support by Welch grant AT-1617 is highly
appreciated.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
Kojima, A., Teshima, K., Shirai, Y. & Miyasaka, T. Organometal halide perovskites as visible-light sensitizers for photovoltaic cells. J.
Am. Chem. Soc. 131, 6050 -- 6051 (2009).
McGehee, M. D. Perovskite solar cells: Continuing to soar. Nat. Mater. 13, 845 -- 846 (2014).
You, J., Meng, L., Hong, Z., Li, G. & Yang, Y. Inverted planar structure of perovskite solar cells. in Organic-Inorganic Halide Perovskite
Photovoltaics: From Fundamentals to Device Architectures 307 -- 324 (2016). doi:10.1007/978-3-319-35114-8_12
Ying, Y., Jing, G., Jia-Rui, C. & Xue-Yi, G. Research Progress of Perovskite Solar Cells. J. Inorg. Mater. 30, 1131 -- 1138 (2015).
Gonzalez-Pedro, V. et al. General working principles of CH3NH3PbX3 perovskite solar cells. Nano Lett. 14, 888 -- 893 (2014).
Zhao, Y., Nardes, A. M. & Zhu, K. Solid-State Mesostructured Perovskite CH3NH3PbI3 Solar Cells: Charge Transport, Recombination,
and Diffusion Length. J. Phys. Chem. Lett. 140120101155009 (2014). doi:10.1021/jz500003v
Calió, L., Kazim, S., Grätzel, M. & Ahmad, S. Hole-Transport Materials for Perovskite Solar Cells. Angewandte Chemie - International
Edition 55, 14522 -- 14545 (2016).
Yamada, Y., Nakamura, T., Endo, M., Wakamiya, A. & Kanemitsu, Y. Photocarrier recombination dynamics in perovskite CH3NH3PbI3
for solar cell applications. J. Am. Chem. Soc. 136, 11610 -- 3 (2014).
Adhikari, N. et al. Interfacial Study to Suppress Charge Carrier Recombination for High Efficiency Perovskite Solar Cells. ACS Appl.
Mater. Interfaces 7, 26445 -- 26454 (2015).
Suarez, B. et al. Recombination study of combined halides (Cl, Br, I) perovskite solar cells. J. Phys. Chem. Lett. 5, 1628 -- 1635 (2014).
Stranks, S. D. et al. Electron-Hole Diffusion Lengths Exceeding 1 Micrometer in an Organometal Trihalide Perovskite Absorber. Science
(80-. ). 342, 341 -- 344 (2013).
Ke, X., Yan, J., Zhang, A., Zhang, B. & Chen, Y. Optical band gap transition from direct to indirect induced by organic content of
CH3NH3PbI3 perovskite films. Appl. Phys. Lett. 107, 091904 (2015).
Unger, E. L. et al. Roadmap and roadblocks for the band gap tunability of metal halide perovskites. J. Mater. Chem. A 5, 11401 -- 11409
(2017).
Qaid, S. M. H. et al. Band-gap tuning of lead halide perovskite using a single step spin-coating deposition process. Mater. Lett. 164,
498 -- 501 (2016).
Amat, A. et al. Cation-induced band-gap tuning in organohalide perovskites: Interplay of spin-orbit coupling and octahedra tilting.
Nano Lett. 14, 3608 -- 3616 (2014).
Saliba, M. et al. Cesium-containing triple cation perovskite solar cells: improved stability, reproducibility and high efficiency. Energy
Environ. Sci. 9, 1989 -- 1997 (2016).
Ku, Z., Rong, Y., Xu, M., Liu, T. & Han, H. Full printable processed mesoscopic CH3NH3PbI3/TiO2 heterojunction solar cells with carbon
counter electrode. Sci. Rep. 3, 3132 (2013).
Miyasaka, T. Toward printable sensitized mesoscopic solar cells: Light-harvesting management with thin TiO2 films. Journal of Physical
Chemistry Letters 2, 262 -- 269 (2011).
Hu, Y. et al. Stable Large-Area (10 × 10 cm 2 ) Printable Mesoscopic Perovskite Module Exceeding 10% Efficiency. Sol. RRL 1600019
(2017). doi:10.1002/solr.201600019
Jung, H. S. & Park, N. G. Perovskite solar cells: From materials to devices. Small 11, 10 -- 25 (2015).
Namatame, M. et al. Direct observation of dramatically enhanced hole formation in a perovskite-solar-cell material spiro-OMeTAD by
Li-TFSI doping. Appl. Phys. Lett. 110, (2017).
Nguyen, W. H., Bailie, C. D., Unger, E. L. & McGehee, M. D. Enhancing the hole-conductivity of spiro-OMeTAD without oxygen or
lithium salts by using spiro(TFSI)2 in perovskite and dye-sensitized solar cells. J. Am. Chem. Soc. 136, 10996 -- 11001 (2014).
Marinova, N. et al. Light Harvesting and Charge Recombination in CH3NH3PbI3 Perovskite Solar Cells Studied by Hole-transport-layer
Thickness Variation. ACS Nano 150313232238003 (2015). doi:10.1021/acsnano.5b00447
Sanchez, R. S. & Mas-Marza, E. Light-induced effects on Spiro-OMeTAD films and hybrid lead halide perovskite solar cells. Sol. Energy
Mater. Sol. Cells 158, 189 -- 194 (2016).
Zhang, Y. et al. Charge selective contacts, mobile ions and anomalous hysteresis in organic -- inorganic perovskite solar cells. Mater.
Horizons 1, 96 (2015).
Deng, Y., Xiao, Z. & Huang, J. Light-induced self-poling effect on organometal trihalide perovskite solar cells for increased device
efficiency and stability. Adv. Energy Mater. 5, (2015).
Shao, Y., Xiao, Z., Bi, C., Yuan, Y. & Huang, J. Origin and elimination of photocurrent hysteresis by fullerene passivation in CH3NH3PbI3
planar heterojunction solar cells. Nat. Commun. 5, 1 -- 7 (2014).
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
Lee, K. M. et al. Thickness effects of thermally evaporated C60 thin films on regular-type CH3NH3PbI3 based solar cells. Sol. Energy
Mater. Sol. Cells 164, 13 -- 18 (2017).
Momblona, C. et al. Efficient vacuum deposited p-i-n and n-i-p perovskite solar cells employing doped charge transport layers. Energy
Environ. Sci. (2016). doi:10.1039/C6EE02100J
Schubert, S. et al. Highly doped fullerene C60 thin films as transparent stand alone top electrode for organic solar cells. Sol. Energy
Mater. Sol. Cells 118, 165 -- 170 (2013).
L??ssem, B., Riede, M. & Leo, K. Doping of organic semiconductors. Physica Status Solidi (A) Applications and Materials Science 210, 9 --
43 (2013).
Domanski, K. et al. Not All That Glitters Is Gold: Metal-Migration-Induced Degradation in Perovskite Solar Cells. ACS Nano 10, 6306 --
6314 (2016).
Xiong, J. et al. Interface degradation of perovskite solar cells and its modification using an annealing-free TiO 2 NPs layer. Org.
Electron. 30, 30 -- 35 (2016).
Sanehira, E. M. et al. Influence of Electrode Interfaces on the Stability of Perovskite Solar Cells: Reduced Degradation Using MoO x /Al
for Hole Collection. ACS Energy Lett. 1, 38 -- 45 (2016).
Zimmermann, I. et al. One-Year stable perovskite solar cells by 2D/3D interface engineering. Nat. Commun. 1 -- 8 (2017).
doi:10.1038/ncomms15684
Aitola, K. et al. Carbon nanotube-based hybrid hole-transporting material and selective contact for high efficiency perovskite solar
cells. Energy Environ. Sci. 9, 461 -- 466 (2016).
Habisreutinger, S. N., Nicholas, R. J. & Snaith, H. J. Carbon Nanotubes in Perovskite Solar Cells. Adv. Energy Mater. 7, 1601839 (2017).
Yu, W. et al. Efficiency exceeding 10% for inverted polymer solar cells with a ZnO/ionic liquid combined cathode interfacial layer. J.
Mater. Chem. A 3, 10660 -- 10665 (2015).
Zhang, J. Efficient Inverted Polymer Solar Cells with self-assembled Ionic Liquid Layer as a Cathode Buffer Layer. in International
Photonics and OptoElectronics PT3D.4 (OSA, 2015). doi:10.1364/PFE.2015.PT3D.4
Kang, H., Hong, S., Lee, J. & Lee, K. Electrostatically self-assembled nonconjugated polyelectrolytes as an ideal interfacial layer for
inverted polymer solar cells. Adv. Mater. 24, 3005 -- 3009 (2012).
Karak, S. et al. A perovskite based plug and play AC photovoltaic device with ionic liquid induced transient opto-electronic conversion.
J. Mater. Chem. A 4, 9019 -- 9028 (2016).
Zhou, M. et al. Enhancement of power conversion efficiency of PTB7:PCBM-based solar cells by gate bias. Organic Electronics: physics,
materials, applications 32, 34 -- 40 (2016).
Fujimoto, T. et al. Electric-double-layer field-effect transistors with ionic liquids. Phys. Chem. Chem. Phys. 15, 8983 (2013).
Organic Electronics: Emerging Concepts and Technologies - Google Книги. Available at:
https://books.google.ru/books?id=uQYeAAAAQBAJ&pg=PT265&lpg=PT265&dq=OLET+ionic+gating&source=bl&ots=6XoPbPXC80&sig
=ACfU3U3qSFGgGbmlswuyuAI9Hq04fcBkGA&hl=ru&sa=X&ved=2ahUKEwjFzMWhstblAhWc7aYKHbn9D6cQ6AEwAHoECAgQAQ#v=one
page&q=OLET ionic gating&f=false. (Accessed: 6th November 2019)
Saranin, D. et al. Tunable organic PV parallel tandem with ionic gating. J. Renew. Sustain. Energy 9, 021204 (2017).
Cook, A. B., Yuen, J. D., Micheli, J. W., Nasibulin, A. G. & Zakhidov, A. Ambient Method for the Production of an Ionically Gated Carbon
Nanotube Common Cathode in Tandem Organic Solar Cells. J. Vis. Exp. e52380 (2014). doi:10.3791/52380
Mort, J. et al. Photoconductivity in solid films of C60/70. Chem. Phys. Lett. 186, 281 -- 283 (1991).
Zhou, Y. & Azumi, R. Carbon nanotube based transparent conductive films: progress, challenges, and perspectives. Sci. Technol. Adv.
Mater. 17, 493 -- 516 (2016).
Hosono, H. & Ueda, K. Transparent C 58. Transparent Conductive Oxides. (2017). doi:10.1007/978-3-319-48933-9_58
Maciel, C. & Fileti, E. E. Molecular interactions between fullerene C60 and ionic liquids. Chem. Phys. Lett. 568 -- 569, 75 -- 79 (2013).
Su, W. S., Leung, T. C. & Chan, C. T. Work function of single-walled and multiwalled carbon nanotubes: First-principles study. Phys.
Rev. B - Condens. Matter Mater. Phys. (2007). doi:10.1103/PhysRevB.76.235413
Saranin, D. et al. Tunable organic PV parallel tandem with ionic gating. J. Renew. Sustain. Energy 9, 021204 (2017).
Saranin, D. S., Voroshilov, P. M., Simovski, C. R. & Zakhidov, A. A. Ionically Gated Small Molecule OPV: Controlled n-doping of Thick
Fullerene acceptor layers. (2018).
Gets, D. et al. Light-emitting perovskite solar cell with segregation enhanced self doping. Appl. Surf. Sci. 476, 486 -- 492 (2019).
Leijtens, T. et al. Mapping electric field-induced switchable poling and structural degradation in hybrid lead halide perovskite thin
films. Adv. Energy Mater. (2015). doi:10.1002/aenm.201500962
Saranin, D. et al. Tunable organic PV parallel tandem with ionic gating. J. Renew. Sustain. Energy (2017). doi:10.1063/1.4979900
Zhang, K. et al. Fullerenes and derivatives as electron transport materials in perovskite solar cells. Sci China Chem 60, 144 -- 150 (2017).
Zhang, F. et al. Influence of PC 60 BM or PC 70 BM as electron acceptor on the performance of polymer solar cells. Sol. Energy Mater.
Sol. Cells 97, 71 -- 77 (2011).
Fatemi, S. M. & Foroutan, M. Recent findings about ionic liquids mixtures obtained by molecular dynamics simulation. J.
Nanostructure Chem. doi:10.1007/s40097-015-0155-0
Polo-Luque, M. L., Simonet, B. M. & Valcárcel, M. Functionalization and dispersion of carbon nanotubes in ionic liquids. TrAC Trends
Anal. Chem. 47, 99 -- 110 (2013).
Chaban, V. V & Fileti, E. E. Strong electronic polarization of the C60 fullerene by imidazolium-based ionic liquids: accurate insights
from Born-Oppenheimer molecular dynamic simulations. Phys. Chem. Chem. Phys. 17, 15739 -- 45 (2015).
Jung, M.-C., Raga, S. R., Ono, L. K. & Qi, Y. Substantial improvement of perovskite solar cells stability by pinhole-free hole transport
layer with doping engineering. Sci. Rep. 5, 9863 (2015).
Zhuang, T. et al. Fullerene derivatives as electron donor for organic photovoltaic cells. Appl. Phys. Lett. 103, 203301 (2013).
Dietz, P., Fostiropoulos, K., Krätschmer, W., Hansma, P. K. & Krmschmer, W. Size and packing of fullerenes on C 60 /C 70 crystal
surfaces studied by atomic force microscopy Size and packing of dullerenes on C6,/C& crystal surfaces studied by atomic force
microscopy. Cit. Appl. Phys. Lett. J. Vac. Sci. Technol. B Microelectron. Nanom. Struct. Process. Meas. Phenom. 60, (1992).
56.
57.
58.
59.
60.
61.
62.
63.
64.
|
1806.02202 | 2 | 1806 | 2018-06-11T12:46:23 | Graphene Reflectarray Metasurface for Terahertz Beam Steering and Phase Modulation | [
"physics.app-ph",
"cond-mat.mtrl-sci",
"physics.optics"
] | We report a THz reflectarray metasurface which uses graphene as active element to achieve beam steering, shaping and broadband phase modulation. This is based on the creation of a voltage controlled reconfigurable phase hologram, which can impart different reflection angles and phases to an incident beam, replacing bulky and fragile rotating mirrors used for terahertz imaging. This can also find applications in other regions of the electromagnetic spectrum, paving the way to versatile optical devices including light radars, adaptive optics, electro-optical modulators and screens. | physics.app-ph | physics | Graphene Reflectarray Metasurface for Terahertz Beam Steering and Phase
Modulation
M. Tamagnone 1,2,∗,†, S. Capdevila1,†, A. Lombardo3, J. Wu3, A.
1Laboratory of Electromagnetics and Antennas, ´Ecole Polytechnique F´ed´erale de Lausanne, Lausanne, Switzerland
Centeno4, A. Zurutuza4, A. M. Ionescu5, A. C. Ferrari3, J. R. Mosig1
2Harvard John A. Paulson School of Engineering and Applied Sciences,
Harvard University, Cambridge, Massachusetts 02138, USA
3Cambridge Graphene Centre, University of Cambridge,
9 J.J. Thompson Avenue, Cambridge CB3 OFA, UK
4Graphenea SA, 20018 Donostia-San Sebasti´an, Spain
5Nanoelectronic Devices Laboratory, ´Ecole Polytechnique F´ed´erale de Lausanne, Lausanne, Switzerland
∗Corresponding authors and
†These Authors contributed equally
We report a THz reflectarray metasurface which uses graphene as active element to achieve beam
steering, shaping and broadband phase modulation. This is based on the creation of a voltage
controlled reconfigurable phase hologram, which can impart different reflection angles and phases
to an incident beam, thus replacing bulky and fragile rotating mirrors used for terahertz imaging.
This can also find applications in other regions of the electromagnetic spectrum, paving the way
to versatile optical devices including light radars, adaptive optics, electro-optical modulators and
screens.
Metasurfaces are planar devices based on a periodic
or quasi-periodic bi-dimensional array of cells (typically
dielectric or metallic elements placed on a layered sub-
strate) capable of manipulating impinging light to ob-
tain various functionalities, such as focusing[1, 2], beam
steering and shaping[3–5], unilateral propagation[6], po-
larization control[7, 8], frequency filtering[9, 10], non-
linear phenomena[11], and dynamic modulation[6, 12–
18]. When optically tuneable materials are embedded
in the cells, metasurfaces can be designed to dynamically
steer a beam in different directions. This can be achieved
at microwave frequencies in reflectarray (RA) meta-
surfaces using, e.g., micro-electrical-mechanical-systems
(MEMS)[3, 4], or voltage controlled capacitors[3, 19]. Ex-
tending beam steering to THz, infrared and visible fre-
quencies is challenging due to the scarcity of compact
tuneable elements operating at shorter wavelengths.
This technological issue can be solved by using sin-
gle layer graphene (SLG), which is an ideal material for
photonics and optoelectronics due to its rich physics and
gate-tuneable optical properties[20]. Compared to bulk
materials, the possibility of inducing high carrier den-
sities in SLG is the key to achieve optical tuneability
both for optical intraband processes [14, 21] and for inter-
band processes[22] (which are relevant for optoelectronic
modulators and photodetectors[23]). Furthermore, high
mobility of SLG allows mid infrared plasmon-polaritons,
also tuneable by gating[24, 25].
SLG is ideally suited to modulate terahertz waves be-
cause of its high mobility and easy integration on Si
technology. The mobility is linked to the massless na-
ture of carriers in SLG and allows for a larger conduc-
tivity tuneability range than Si, for a given carrier den-
sity interval [14]. Unlike radio-frequency MEMS[3, 4],
SLG does not require packaging [26], its switching speed
is several order of magnitude faster[22] and it does not
suffer from reliability issues[3, 4]. Tuneable capacitors,
instead, are dominated by resistive losses above the mi-
crowave range[3]. Thus, SLG is an ideal choice for THz
modulation[14, 16, 17].
We report a THz reflectarray metasurface exploiting
SLG as active element to achieve beam steering, shaping
and broadband phase modulation. Our device achieves
dynamical beam steering thanks to an array of cells (in-
cluding metal and gated SLG) built on a reflective sub-
strate, see Fig.1a. This consists of a dielectric spacer
layer (20 µm float zone Si, with dielectric constant 11.7
at 1 THz[2]), on a metallic reflective film (140 nm Ag
on 60 nm Al)[9, 21]. An additional AlO2 layer is used
to gate the SLG, allowing for the dynamical tuning of
its conductivity and hence of its optical behavior at THz
frequencies. High resistivity Si (for our sample the resis-
tivity is 10 kΩ·cm) is transparent at THz frequencies[9],
but allows injected carriers to charge the gate capaci-
tance and tune SLG via field effect[14]. The unit cell in
Figs.1b,c is inspired by bow tie antennas[27], with two
trapezoidal Au elements that concentrate the impinging
electromagnetic energy in a 3 µm narrow gap where SLG
is placed.
Fig.5 in Methods M1 illustrates the RA substrate fab-
rication process flow. Our device requires a substrate
comprising a reflective conductive ground plane and a
dielectric spacer with thickness in the order of a quarter
wavelength (in the material itself). We achieve this by
using high resistivity Si as the dielectric spacer. An an-
odic bonding process is used to bond a metallic reflective
layer (Ag + Al) to a supporting glass substrate. Fig.6
in Methods M1 summarizes the fabrication of the RA
8
1
0
2
n
u
J
1
1
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
2
v
2
0
2
2
0
.
6
0
8
1
:
v
i
X
r
a
2
FIG. 1: a, Cross section of the device mounted on a printed circuit board (PCB) and wire-bonded. Thickness of the layers
from bottom to top: glass substrate 525 µm, evaporated Al 100 nm; Ag 100 nm, Si 20 µm, Al2O3 200 nm, Au 100 nm; 5 nm
Cr adhesion layer). b, Scanning electron microscope (SEM) picture of a representative device (scale bar 50 µm). c Unit cell.
width L = 100 µm, height H = 20 µm, antenna gap G = 3 µm, first trapezoid base W1 = 70 µm, second trapezoid base W2 = 7
µm, width of interconnecting line I = 2µm. The final size of the array area is 8×8 mm2, hence 80x400 cells. d, characterization
of the electric modulation of the complex reflection coefficient. Light is s-polarized and incident with an angle θi = 45◦. In this
case, all columns are driven with the same voltage (V1 = V2 = V3 = ... = VN ), therefore SLG has the same σ in all the cells.
Since the cells are smaller than half wavelength, light is reflected specularly (θr = 45◦) and the reflection coefficient depends
on the frequency and σ.e, simulations and f, measurements of the complex reflection coefficient of the array. The colored lines
in the polar plots represent the evolution of the reflective index with frequency (indicated in red), and each curve represents a
different SLG impedance or biasing voltage. The transverse grey lines join data points measured at the same frequency.
starting from the substrate chips.
The beam steering RA can work as intended only if
the cell can modulate its reflection coefficient between
two values (ΓON, ΓOFF) with a phase modulation π (see
Methods M3). The amplitude of the reflection coeffi-
cient should be maximized and remain constant in the
two states. This is equivalent to creating a metasurface
where each cell has a tuneable surface impedance, since
the reflection coefficient Γ and the surface impedance ZS
are related by[21, 27]:
Γ =
ZS − η
ZS + η
(1)
(cid:113)
µ0ε−1
where η =
0 (cid:39) 377 Ω is the free space
impedance, µ0 is the vacuum magnetic permeability and
ε0 is the vacuum dielectric permittivity. The cell can
then be designed to obtain suitable values of ZS start-
ing from the SLG's sheet impedance Zg = σ−1, where σ
is SLG's conductivity. This can be changed via electric
field gating between a maximum (σON) and minimum
(σOFF).
It is possible to control the reflection phase
in a binary way (two opposite values of the phase) if
the metasurface is designed to have complete absorption
(Γ = 0) for ZS = (cid:112)ZS,ON · ZS,OFF. Because Γ = 0
comes (cid:112)ZS,ON · ZS,OFF = η. The geometric average is
ZS = η when Zg = (cid:112)Zg,ON · Zg,OFF. For our samples
(cid:112)Zg,ON · Zg,OFF = 1789 Ω = 4.75η. This implies that
implies ZS = η, the approximate design condition be-
used here, to ensure that ΓON = −ΓOFF, providing bi-
nary phase modulation with the same amplitude in two
states. The metasurface design must therefore achieve
we measure Zg,ON = 800 Ω, Zg,OFF = 4000 Ω. Therefore
the cell must be designed to scale down SLG's impedance
to a factor 4.75 to be at the optimal working point.
To achieve this, we first chose a Salisbury screen
configuration[28]. This consists of a dielectric spacer on
a reflective metallic layer[21]. The spacer is a Si layer
having thickness t:
t =
λ0
4n
=
c
4nf0
(2)
where f0 = 1 THz is the design frequency, λ0 is the
corresponding free space wavelength, c is the speed of
abcdeGlass substrateSiPCB supportWireAl2O3Al+AgAu AntennasGrapheneBondsLGrapheneAuHW1W2GV1V2V3V4V5V6V7V8EE800150040000°45°90°135°180°-135°-90°-45°10.750.250.5 1 THz1.1 THz1.2THz0.9THzf20 V10 V0 V-10 V-20 V-30 V-40 V0°45°90°135°180°-135°-90°-45°10.750.250.51 THz1.1 THz0.9THzIi�r�xyz 3
FIG. 2: a, Unit cell dimensions after optimization. L = 100 µm, H = 20 µm, G = 3 µm, W1 = 70 µm, W2 = 7 µm, I = 2 µm.
b, Screenshot of the unit cell simulation setup (the used software is Ansys HFSS). c, Artistic view of the final array.
√
light, and n =
11.7 is the Si refractive index. The
purpose of this structure is to cancel the contribution of
the reflective layer to the free space impedance, obtaining
ZS (cid:39) 0 in absence of other structures on top of the spacer,
as discussed in Ref.21. From Eq.2 we get t = 21.9 µm.
For our experiments we use t = 20 µm due to limitations
in the available silicon on insulator (SOI) wafers.
The metallic structure in Fig.2 is chosen to concen-
trate the impinging field on a SLG rectangular load over
a broad-band, hence the choice of the bow-tie antenna
element. SLG is prolonged on one side, to contact an
additional Au bias line used to improve the connectiv-
ity of the column, so that the applied voltage is uniform
even in case of cracks in one or more of the SLG loads.
Voltage is applied to both ends of the column.
The cells (20 × 100 µm2) are smaller than half of the
wavelength (300 µm at 1 THz). Each reflects the incident
waves with a reflection coefficient that can be modulated
applying different voltages to SLG. Numerical simula-
tions and measurements of the reflection coefficient are
in Fig.1d-f. These measurements are performed by gating
all the cells with the same voltage and then measuring
the overall reflection coefficient Γ of the surface, Fig.1d.
Γ is a complex number describing both the amplitude
and the phase of the reflected wave, with the phase de-
lay normalized with respect to a reference mirror (Au
deposited on the same chip directly on Al2O3). A THz
fiber-coupled time domain system is used to measure Γ
(see Methods M2), focusing the incident beam on a small
area of the array to avoid probing areas outside it. Note
that, because of the subwavelength nature of the array,
only one reflected beam exists, without diffraction.
The unit cell geometry is optimized so that, at the
target design frequency of 1.05 THz, different σ cause
the reflection coefficient to vary from one value to its
opposite, passing close to the total absorption condition
(Γ = 0). In this way, by switching the cell between these
two states (ON and OFF) a local phase modulation of
π can be achieved. This is similar to the concept pro-
posed in Refs.15, 16 and demonstrated experimentally at
microwave frequencies in Refs.3, 19. The slight shift mea-
sured reflection coefficient with respect to the simulations
visible in the figures is due to fabrication tolerances.
Beam steering can then be achieved by switching the
cell state in the array so that a dynamical and reconfig-
urable phase hologram is created, obtaining a fully re-
configurable RA. We focus on beam steering and shap-
ing in one plane. This allows for a simplification of the
control network, whereby all cells belonging to the same
column are connected together, and each column can be
controlled by an individual voltage.
The far electric field radiation pattern obtained illumi-
nating the array with a plane wave having electric field
amplitude E0 and angle of incidence θi (in our case fixed
to 45◦) can be estimated based on the interference of
discrete radiators[27]:
abcLGrapheneAuHW1W2GIt 4
FIG. 3: a, Beam steering principle. Columns are alternatively set ON and OFF (VON = 26 V, VOFF = -44 V). b-c coarse
beam steering obtained using periodic patterns.b Beam profiles at 0.977 THz, where the deflected field is higher. Amplitudes
are normalized to the largest deflected field for any angle and frequency. c, control voltage patterns used in b. d angular
dispersion of the beam for different frequencies. The dashed white line is the expected theoretical dispersion from Eq.4. e-f,
fine beam steering from quasi-periodic patterns. e beam profile at 0.977 THz. f control voltage patterns used in e. g beam
broadening using chirped patterns. h dual beam operation. i time response of steered beam, compared to the pulse (not in
scale) of our THz time domain system (TDS). j time response of steered beam de-convoluted to remove the measurement pulse
shape. Both i and j show a standard deflected beam (P = 6) and one obtained with a chirped pattern. The time response of
the chirped pattern is also chirped: the red and green traces are aligned in the centre, but the chirped pattern is delayed with
respect to the non-chirped one both at the end and at the beginning of the pulse. k chirped and dual beam control patterns.
E(θ, r) = E0 g(r) fC(θ) fA(θ) = E0 g(r) fC(θ)
Γn e jnk0L(sin θ−sin θi)
N(cid:88)
(3)
n=1
where θ is the deflection angle, fC(θ) is the radiation
pattern of a single isolated column, fA(θ) is called array
factor [27], Γn is the reflection coefficient of the n-th cell,
N is the total number of cells in the array, k0 = 2π/λ
is the wavenumber, L is the cell width, r is the distance
from the RA and g(r) = r−1exp(−jkr). The fC(θ) fac-
tor is negligible here, as it does not show sharp variations
in θ due to the sub-wavelength size of the unit cell. The
summation (hence fA(θ)) is maximized when its elements
are in phase. If a linear phase profile is created setting
the Γn elements such that Γn = ejnφ then the maxi-
mum (hence the reflected beam direction) is obtained for
θ = arcsin
, which can be changed dynam-
(cid:16)
sin θi − φ
k0L
(cid:17)
aV1V2V3V4V5V6V7V8i�r�cdbfeColumn 1Column 80Column 40Pattern "Period 5":Pattern "Period 6":Pattern "Period 7":Pattern "Period 8":Pattern "Period 4":Pattern "All ON":Column 1Column 80Column 40Pattern "Period 5.75":Pattern "Period 6":Pattern "Period 6.25":Pattern "Period 6.5":Pattern "Period 5.5":Pattern "All ON":-5051015202530Deflection Angle(deg)00.20.40.60.8Field amplituder�Period 4Period 5Period 7Period 6Period 8All ON-5051015202530Deflection Angle(deg)00.20.40.60.8Field amplituder�Period 5.5Period 5.75Period 6.25Period 6Period 6.5All ON0510Deflection Angler(deg)00.10.20.30.4Field amplitudeAll ONPeriod 6Chirp 1Chirp 20102030Deflection Angler(deg)Period 6Dual BeamAll ON=VOFF=VON =VOFF=VON=VOFF=VONr�-0.0100.01Reference pulsePeriod 6Chirp 2859095100Time (ps)-0.0100.01Period 6Chirp 2Amplitude (A.U.)ghijkColumn 1Column 80Column 40Pattern "Chirp 1":Pattern "Chirp 2":Pattern "Dual Beam":Pattern "Period 6":Pattern "All ON":=VOFF=VON0.98 THz0.98 THzPeriod 60.85THz1.05 THzr�= 6.5°r�= 6.5° 5
FIG. 4: a Geometric discrete phase modulation principle. The solid black line indicates how the delay of the deflected beam
can be modified by shifting the control voltage pattern. b-d measured geometrical phase modulation by shifting a periodic
pattern with P=4, 6, 8. The resulting phase shift keying (PSK) constellations on the complex plane are shown on a polar plot
with lines that represent the symbol evolution from 0.94 to 1THz. Details on background removal are given in Method M2.
Close to each symbol is the supercell used to generate it. Each measurement is done at the angle of maximum intensity, see
Fig.3b.
ically tuning the phase profile. It is possible to show (see
Methods M3 for a full mathematical derivation) that this
principle still holds if the reflection phase is quantized to
just two values (0 and π) for all the elements, thus reduc-
ing the gradient to a periodic set of segments with phase
0 alternated with segments of phase π. The periodicity
P of the pattern expressed in terms of number of cells is
then given by P = 2π/φ and by the beam steering law:
(cid:18)
(cid:19)
θ = arcsin
sin θi − λ
P L
(4)
where λ is the wavelength.
If P is an even integer the pattern consists of a rep-
etitions of a supercell of P cells (with P/2 cells set to
phase 0 and P/2 to π). Fig.3a illustrates the case P = 4.
However, it is possible to generalize the pattern to odd
and even fractional values of P using a pattern genera-
tion technique described in Methods M4, thus achieving
continuous beam steering.
Beam-steering with integer P (from 4 to 8) is shown
in Fig.3b, for the voltage patterns in Fig.3c. The angu-
lar steering range reaches 25◦. The beam is well-formed
with the exception of small side lobes which appear for
odd P , due to the technique used to emulate odd and
fractional P values. Fig.3d plots the beam steering as a
function of frequency, compared with the prediction of
Eq.4, while Figs.3e,f demonstrate the continuous beam-
steering achieved with fractional P .
Our device can also reconfigure the beam shape. This
is achieved by smoothly changing the local P from one
extreme to the other of the array using a chirped pat-
tern, as shown in Fig.3k, thus having slightly different
deflection angles across the array, emulating a parabolic
profile. The device operates as a parabolic mirror with
tunable curvature, which we use here to generate a wider
beam, Fig.3g. The same principle can be used to achieve
tuneable focusing (limited here to one dimension).
Besides focusing and widening the beam, more com-
plex operations can be performed. E.g., Fig.3h plots
the generation of a double beam by filling two halves
of the array with patterns having different P (changing
abruptly in the middle of the array, as shown in the dual-
beam pattern in Fig.3k). Another important application
is the ability to manipulate an impinging THz pulse at
the time domain level. This is possible because the in-
abcdAll ONAll OFF4-PSKAll ONAll OFF6-PSKAll ONAll OFF8-PSK0°45°90°135°180°-135°-90°-45°10.750.250.50°45°90°135°180°-135°-90°-45°10.750.250.50°45°90°135°180°-135°-90°-45°10.750.250.50.94 - 1THz0.94 - 1THz0.94 - 1THz=-5°r�=10°r�=17.5°r� cident pulse reaches at different times each RA element.
Therefore, the voltage pattern selected on the array is
transferred to the time response of the system (within
some limits due to the spectral response of each cell and
to the total size of the illuminated portion of the array).
Figs.3i,j show that a periodic pattern with P = 6 gener-
ates a sinusoid (of finite time duration due to the finite
size of the array). The chirped pattern used for beam
broadening gives a chirped sinusoid in the time response.
Similar transformations can be achieved with more com-
plex patterns.
We now consider the effect of shifting a periodic pat-
tern (with P = 2, 4, 6) of a finite number of cells, and
we verify that the corresponding time domain sinusoid
is similarly de-phased. This is equivalent to the phase
shift associated to a lateral translation of an optical
grating[29], but the movement here is emulated by the
reconfigurable control patterns. The experiment is il-
lustrated in Fig.4a and the measurements, better repre-
sented in the frequency domain, are plotted in Figs.4b-d.
These are the complex reflection coefficients for each of
the aforementioned cases, and for each possible shift of
the pattern. E.g., the P = 4 patterns can be shifted in
4 ways, with shifts of 0, 1, 2, 3 cells, while shifting of
4 cells is identical to 0 and so on. Each cell shift corre-
sponds to a phase delay of 2π/P regardless of the beam
frequency. This scheme, here referred to as geometrical
PSK (phase shift keying[30]), provides a way to perform
a precise phase modulation on a wide band (60GHz at
1THz).
In summary, we reported a reconfigurable RA metasur-
face for terahertz waves using SLG. Beam steering, shap-
ing and modulation were achieved. Our results demon-
strate that graphene can be embedded in metasurfaces
providing an unprecedented control and modulation ca-
pabilities for THz beams, with applications for adaptive
optics, sensing and telecommunications. Our approach
can be extended to mid infrared, and to two dimensional
beam steering, by using individual cell control.
Acknowledgements
We dedicate this work to the memory of Prof. Julien
Perruisseau-Carrier. We thank Giancarlo Corradini,
Cyrille Hibert, Julien Dorsaz, Joffrey Pernollet, Zdenek
Benes, and the rest of EPFL CMi staff for the use-
ful discussions. We acknowledge funding from the EU
Graphene Flagship, the Swiss National Science Foun-
dation (SNSF) grants 133583 and 168545, the Hasler
Foundation (Project 11149), ERC Grant Hetero2D, EP-
SRC grant nos. EP/509 K01711X/1, EP/K017144/1,
EP/N010345/1, EP/M507799/5101 and EP/ L016087/1.
6
Methods
M1. Fabrication process flow
The starting point is a Si on insulator (SOI) wafer
(produced by Ultrasil Corp.) having a device layer with
the required characteristics for our dielectric spacer with
20 µm thickness and high resistivity ∼ 1 kΩ·cm, Fig.5a.
Ag is deposited (e-beam evaporation) to create the re-
flective layer, followed by an Al layer (vacuum is not
broken between the two depositions, Fig.5b). The Al
coated face of the SOI wafer is then bonded via anodic
bonding (Fig.5c) to a borosilicate glass wafer (Borofloat
33, very similar to Pyrex in composition[31]), acting as
a support for the device layer, which is too fragile to
be handled alone given its thickness. Bonding is per-
formed at atmospheric pressure with a Suss Microtec SB6
tool immediately after evaporation. Glass wafers are also
cleaned in a hot piranha bath immediately before bond-
ing to remove organic impurities. A second borosilicate
wafer is used below the one to be bonded, as a sacrificial
substrate to collect the excess Na ions, thus preventing
contamination. To prevent accidental bonding of the two
borosilicate wafers, the sacrificial substrate is thinned us-
ing wafer grinding, and the non-polished surface is placed
in contact with the borosilicate substrate to be bonded.
The aim of the following steps is to eliminate the SOI
handle and box layer, to expose the device layer. This is
done by first grinding the Si handle wafer down to 100
µm (Fig.5d). This is a mechanically aggressive process,
therefore further thinning could damage the substrate or
cause the failure of the bonding. The remaining Si is
dry-etched using a fluorine-based chemistry, with a pro-
cess having 200:1 selectivity with respect to SiO2, Fig.5e.
This ensures that the box layer survives the process, pre-
serving the device layer as well. The box is then dis-
solved in HF 49% (Fig.5f), selected over buffered oxide
etch (BHF) because it etches faster SiO2 [32] and, unlike
BHF, does not attack Al [32].
The gate oxide (200 nm Al2O3) is prepared using
atomic layer deposition (ALD) on all the wafer, as shown
in Fig.5g. Afterwards, dual layer photo-litography (LOR
+ AZ1512) is performed, followed by evaporation of 100
nm Au with an adhesion layer of 5 nm Cr and liftoff
(Fig.5h,i). During this step, the reference mirrors (one
for each chip), bonding pads, and dicing markers are
defined on the full wafer (Fig.8a). Oxygen plasma de-
scum is performed prior to the evaporation to ensure
maximum adhesion, important for the subsequent wire-
bonding step. The wafer is then diced (Fig.5j) using an
automatic dicing saw (Disco DAD-321). During the dic-
ing process, the wafer is protected by a photo-resist layer,
then stripped in remover on each chip, lifting the dicing
residues.
SLG is grown on Cu foil (99.8% purity) by chemical
7
FIG. 5: Fabrication process flow (part 1). a, Initial SOI wafer. b, E-beam evaporation of 140 nm Ag followed by 60 nm Al.
c, Anodic bonding between the deposited Al layer and a support Pyrex wafer. This is cleaned in a hot piranha bath prior
bonding. d, Grinding of the Si handle wafer down to 100 µm. e, Dry etching of the remaining Si. f, wet etching of SiO2 box
layer. g, ALD deposition of 200 nm of AlO2. h, Optical lithography for the bonding pads: dual layer photo-resist spin coat,
exposure and development. i, Oxygen plasma de-scum, e-beam evaporation of 100 nm Au pads with 5 nm Cr for adhesion and
liftoff. j, Dicing.
vapor deposition (CVD) on a tube furnace as for Ref.33.
The Cu foil is annealed in H2 (flow 20 sccm) at 1000 ◦C
for 30 min. After annealing, CH4 (flow 5 sccm) is in-
troduced for 30 min while keeping the temperature at
1000 ◦C, leading to the growth of SLG. This is then
transferred onto the Al2O3/Si/Ag/Pyrex by wet trans-
fer (Figure 6a,b)[34], where polymethyl methracrylate
(PMMA) is used as a sacrificial layer to support SLG
during Cu etching in ammonium persulfate[34]. After
transfer, PMMA is dissolved in acetone. Raman spec-
troscopy is used to monitor the sample quality through-
out the process by using a Renishaw inVia spectrome-
ter equipped with 100X objective and a 2400 groves/mm
grating at 514.5 nm. Representative Raman spectra of
SLG placed onto the Al2O3/Si/Ag/Pyrex substrate are
shown in Fig.7. The spectrum of graphene on Cu shows
not significant D peak, indicating a negligible defect den-
sity [35]. After transfer, the position of G peak is 1590
cm−1 and its full width at half maximum is 17 cm−1,
the position of 2D peak is 2692 cm−1, while the ratio of
the 2D to G peaks intensities, I(2D)/I(G), is ∼1.84 and
the ratio of their areas, A(2D)/A(G), is ∼4, indicating
a Fermi level ∼0.2-0.4eV and a charge carrier concen-
tration ∼1012 cm−2 [36]. The D peak is present in the
spectrum of the transferred SLG, suggesting that some
defects have been introduced during the process. From
I(D)/I(G) ∼0.13 and given the Fermi level, we can esti-
mate a defect density ∼7x1010 cm−2 [37, 38].
SLG is then e-beam patterned using PMMA resist fol-
lowed by oxygen plasma and stripping in acetone at 45
SOI Si device (20 �m)SOI SiO2 box (0.2 to 4 �m)SOI Si handle (300 to 600 �m)SOI Si device (20 �m)SOI SiO2 box (0.2 to 4 �m)SOI Si handle (300 to 600 �m)Ag (140 nm) + Al (60 nm) SOI Si device (20 �m)SOI SiO2 box (0.2 to 4 �m)SOI Si handle (300 to 600 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)SOI Si device (20 �m)SOI SiO2 box (0.2 to 4 �m)Ground Si handle (100 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)SOI Si device (20 �m)SOI SiO2 box (0.2 to 4 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)SOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)SOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)SOI Si device (20 �m)Ag (140 nm) + Al (60 nm)Pyrex support (525 �m)Al2O3 (200 nm)SOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)SOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)abcdefAu Padsghij 8
FIG. 6: a, Chip as prepared in Fig.5. b SLG transfer. c, Lithography for SLG patterning; PMMA spin coat e-beam lithography
and development. d SLG etching with oxygen plasma and PMMA stripping in hot acetone. e, Lithography for metallic antennas:
MMA/PMMA spin coating, e-beam lithography and liftoff. f, Oxygen plasma de-scum, e-beam evaporation of 100 nm Au pads
with 5 nm Cr for adhesion and liftoff. g, Mounting chip on PCB support and wire-bonding.
FIG. 7: Raman spectrum of the SLG before and after transfer.
SOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)aSOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)GrapheneSOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)SOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)SOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)SOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)bcdefgSOI Si device (20 �m)Ag (140 nm) + Al (60 nm) Pyrex support (525 �m)Al2O3 (200 nm)Printed Circuit Board 9
FIG. 8: a, Full wafer, optical lithography mask for Au pads, reference mirrors and traces. b, Chip with all fabrication layers (Au
optical lithography, SLG etch and Au e-beam lithography). c Magnification of the top part of the RA, showing RA columns.
◦C, Fig.6c,d). Apart from patterning SLG in the RA,
during this process all the bonding and traces are also
exposed to the oxygen plasma to ensure that no SLG re-
mains on them, to avoid short circuits. Subsequently, a
new e-beam lithography (MMA + PMMA) is performed
to define the metallic antennas via evaporation and liftoff
in acetone, Figs. 6e,f, 8a,b. Finally the chip is glued to
the PCB substrate and all the columns are connected via
wire-bonding to the PCB traces, Fig.6g). The ground
plane is contacted laterally with Ag paint.
M2. Measurement setup and post processing
The RA is characterized using a commercial fiber cou-
pled THz time domain setup (by Menlo Systems, model
TERA K15 mark II). The THz emitter is mounted at 45
degrees of incidence with respect to the RA, while the
detector is placed on a motorized rotation stage. The
system is first aligned in transmission to maximize the
THz signal intensity, and subsequently in reflection, us-
ing a reference mirror mounted on the sample holder.
This is mounted on a translation stage (motorized XY
linear stages plus manual Z stage), to automatically al-
ternate between sample and reference mirror. All mea-
surements are normalized with respect to the reference
mirror, created on the same chip of the RA during the
optical lithography process on the full wafer.
Two different measurement modes are used, Figure 9:
• Collimated The beam is collimated and impinges
on a large area of the sample (∼1 cm2). The re-
duced spread of the angular components of the
beam (when decomposed into a superposition of
plane waves) ensures precise measurement of an-
gles and radiation patterns, but part of the beam
interacts with the area around the RA.
• Focused: The beam is focused by an additional
pair of lenses so that it impinges completely inside
the array. However, this requires larger spread of
the angular components, hence this mode is not
accurate for angles and radiation pattern measure-
ment. Instead, it is used to measure the reflection
coefficient and the efficiency of the array.
The measurements in Fig.1 are performed in the fo-
cused mode, while those in Figs.3,4 use the collimated
abcGraphene etchGold (Ebeam litho)Gold (Optical litho) 10
FIG. 9: a-b, Collimated beam measurement, schematics and picture. TX, THz emitter; RX, detector. RA, reflectarray; REF,
reference mirror. c-d, Focused beam measurement, schematics and picture. The focusing lens of the detector rotates together
with the detector itself.
mode. The latter is to be preferred to characterize the
geometric phase shift keying (G-PSK, Fig.4), since pre-
cise phase modulation relies on the interaction between
the beam and all of the columns of the RA, which can
be illuminated completely only in the collimated mode.
The drawback of the collimated mode is the interac-
tion of the beam with areas outside the RA metasur-
face. This can be addressed with the following post-
processing method. For radiation pattern characteri-
zation, two measurements are performed, one with the
chosen control sequence, and the other with the oppo-
site (logical NOT) control sequence.
In this way, the
steered beam will have opposite phase in the two cases
(see Method M3) and the radiation pattern can be ex-
tracted by subtracting (frequency by frequency and an-
gle by angle) these two measurements. Any contribu-
tion from the area outside the array is canceled by the
subtraction. For the G-PSK case, the same is accom-
plished by subtracting from all the signals the average in
the complex plane (frequency by frequency). An overall
phase factor e−j(ωτ +φ) is removed from all the traces. A
unique value of the delay τ is used for all the symbols
in each G-PSK measurement. This is done to remove
the free-space phase delay of the measured beam, due to
slight differences in the paths when measuring the array
and the reference mirror. Similarly, the removed phase
factor φ is unique for all the traces, and it is used to
align the symbols to the real and imaginary axes of the
complex plane.
M3. Radiation pattern theory
The array geometry does not depend on the y direc-
tion since the array has a periodicity smaller than half
wavelength in that direction, independently of the control
pattern. We assume that the incident wave is propagat-
ing in the xz plane (i.e. ky = 0). In the low cell-to-cell
bdacTXTXRXFiber to THz setupRX Fiber to THz setup Fiber to THz setupFiber to THz setupRAREFPCBPCBRAREFE field(s-pol)E field(s-pol)LensFocusing coupling approximation[3], the electric field of an antenna
array in the x direction (assuming equidistant elements
separated by L) in far field conditions is given by[3, 27]:
N(cid:88)
n=1
E(θ, r) = g(r)
wnfC(θ) e jnk0L sin θ = g(r)fC(θ)
N(cid:88)
wn e jnk0L sin θ
(5)
n=1
where fC is the single cell radiation pattern, k0 is the
wavenumber, wn is the amplitude associated to the n-th
element. For a RA we can write wn as the product of the
incident field at the element position times a reflection
coefficient:
wn = Γn Ei(x = nL, z = 0)
(6)
where we assume for simplicity and without loss of
generality that the RA is in the z = 0 plane. The electric
field Ei of an incident wave (with incident angle θi with
respect to the normal) is given by:
Ei(x, z) = E0 e−jk0(x sin θi+z cos θi)
(7)
Combining Eqs.7,6,5 we get:
for λ = 2π/k0:
sin θ − sin θi = l
λ
L
l ∈ Z
11
(12)
Because λ/L = 3 in our case, only the specular reflec-
tion l = 0 satisfies this condition, for any θi.
Super-cells can be implemented by creating periodic
patterns of reflection coefficients, fulfilling the condition
Γn = Γn+P , where P is a positive integer number of cells
in the super-cell.
If a super-cell with periodicity P is
implemented, then the array factor can be rewritten, by
splitting the summation in two levels: an external sum
over all the super-cells, and an internal one over the cells
in a super-cell:
N/P(cid:88)
P(cid:88)
fA(θ) =
P(cid:88)
n=0
m=1
Γm e jmk0L(sin θ−sin θi)
N/P(cid:88)
m=1
n=0
Γm e j(nP +m)k0L(sin θ−sin θi) =
e jnP k0L(sin θ−sin θi)
(13)
We can then define the supercell factor fSC:
fSC(θ) =
Γm e jmk0L(sin θ−sin θi)
(14)
P(cid:88)
m=1
N/P(cid:88)
N(cid:88)
N(cid:88)
n=1
N(cid:88)
E(θ) = E0 g(r)fC(θ)
Γn e jnk0L(sin θ−sin θi) =
and the superarray factor fSA:
n=1
E0 g(r)fC(θ)fA(θ)
(8)
fSA(θ) =
e jnP k0L(sin θ−sin θi)
(15)
where we define the array factor fA(θ) as:
fA(θ) =
Γn e jnk0L(sin θ−sin θi)
(9)
We notice that,
if all the reflection coefficients are
phase-modulated of π (thus reversing their sign), the to-
tal phase of the scattered field will also be out of phase
of π, which is used to suppress the background in our
measurements. If all the cells have the same Γ, then:
E(θ) = E0 g(r) fC(θ) Γ
e jnk0L(sin θ−sin θi)
(10)
n=1
A maximum in the reflection is obtained for θ = θi
(which is the direction of the specular reflection) where
all the contributions of the summation add in phase.
More generally, a maximum is obtained if:
k0L(sin θ − sin θi) = 2πl
l ∈ Z
(11)
n=0
and use them to decompose the array factor as fA(θ) =
fSC(θ)fSA(θ). The final expression for the electric far
field is then:
E(θ) = E0 g(r)f (θ) = E0 g(r)fC(θ)fA(θ) =
E0 g(r)fC(θ)fSC(θ)fSA(θ)
(16)
The angular part of the radiation pattern f (θ) is de-
composed in three factors (ordered here from the most
directive to the least):
• fSA, associated to the super-array, identifies a set
of possible directions where light can be scattered,
and behaves as a diffraction grating.
• fSC, associated to the super-cell, gives different
weights to the possible diffraction orders accord-
ingly to the phase of the cells in the supercell.
• fC, associated to the cell, slowly varying with θ in
the subwavelength case and does not affect signifi-
cantly the final pattern.
The directions of diffracted beams launched by the ar-
ray is then given by fSA(θ), and for each beam the fol-
lowing must be satisfied:
sin θ − sin θi = l
λ
P L
l ∈ Z
(17)
Evaluating fSC for each of the diffracted beams:
P(cid:88)
P(cid:88)
m=1
fSC(l) =
Γm e jmk0L(sin θ−sin θi) =
Γm e j2πml/P
l ∈ Z
(18)
m=1
In our case θi = 45◦ and we are operating at 1 THz.
Also, let us consider the simplest case, with N even inte-
ger and with a supercell formed by N/2 cells with reflec-
tion phase 0◦ followed by N/2 cells with reflection phase
l = −1 represents the steered beam of interest.
180◦.
The specular reflection (l = 0) is suppressed because fSC
vanishes for θ = θi. This is due to the fact that in our
super-cell half of the cells have a reflection coefficient Γ
and the remaining ones −Γ, so the total sum is zero.
Then evaluating the summation for the considered su-
percell:
P/2(cid:88)
m=1
Γ
fSC(l) =
e j2πml/P −
P(cid:88)
P(cid:88)
m=1
m=P/2+1
Γm e j2πml/P =
e j2πml/P
(19)
The same cancelation holds for any even value of l.
Any beam for odd values of l different from 1 and -1 is
also strongly attenuated. For all our choices of period (P
between 4 and 8) the l = 1 beam does not exist, since
no real θ solves Eq.17 for the chosen beam wavelength
and incident angle. Imperfections in the array may still
cause smaller side lobes, i.e. unwanted beams for l (cid:54)= −1.
This is especially true for odd and fractional values of P ,
where the analysis of Eq.19 does not apply rigorously
(though it still describes qualitatively the situation).
The geometric PSK operation can be understood in-
specting fSC for l = −1, and noting that a translation
of any supercell pattern described as Γm ← Γ(m+1)modP
creates a phase shift 2π/P (for an infinite array approx-
imation):
fSC(l = −1) =
P(cid:88)
m=1
Γm e −j2πm/P
(20)
12
P(cid:88)
m=1
Γ(m+1)modP e −j2πm/P =
P−1(cid:88)
r=0
Γr e −j2π(r−1)/P =
P(cid:88)
ej2π/P
Γr e −j2πr/P
r=1
(21)
The software Ansys HFSS is used to compute and
optimize the reflection coefficient of the cells.
In par-
ticular, the effect of the metallic bias lines on the cell
impedance is reduced thanks to the optimization. Simu-
lations are performed using the Floquet boundary condi-
tions (equivalent to Bloch periodic boundary conditions,
see for example[6]), with an incident angle 45◦. The op-
timized bias line width is 2 µm, so that its inductance
per unit length is sufficient to reduce its effect on the
structure.
M4. Control unit and list of control sequences
The RA is glued on a support PCB substrate and each
column is wirebonded to a PCB trace. Both ends of each
column are connected to a high voltage transistor stage,
and all the stages are controlled by two Arduino units.
Fig.10a is an Arduino unit connected to a PCB driver
with 40 transistor stages. A similar unit controls the re-
maining 40 columns. Fig.10b shows a circuit schematic
of the high voltage stage, while Fig.10c illustrates the
support PCB substrate. Fig.10d has two PCB drivers
connected to the support PCB. The high voltage stage
can switch on and off the voltage of each column, reach-
ing values close to the supply voltage in one case and
close to ground in the other. The needed positive and
negative gate voltages are achieved by applying an inter-
mediate voltage on the retroreflector of the RA, so that
the voltage difference is negative in one state and posi-
tive in the other one. The total voltage is set via a high
voltage DC generator, and the retroreflector voltage is
controlled by a potentiometer.
Tables I, II, III, IV show the control sequences used in
each of our experiments. These are programmed into the
Arduino module by the control computer. The latter can
control the Arduino modules, the rotary stage, the XY
stage, the high voltage generator and the terahertz setup,
so that the measurements are completely automatized.
ON and OFF states (or equivalently 1 and 0) corre-
spond to the following voltages between SLG and retrore-
flector: VON = 26V , VOFF = −44V . The sequences are
generated by discretizing sinusoids and chirped sinusoids
with different periods, which leads to quasi-periodic sig-
nals for odd and fractional P values.
13
Sequence name Sequence
TABLE I: List of control sequences for coarse beam steering
All OFF
All ON
00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111
Period 4
¬ Period 4
Period 5
¬ Period 5
Period 6
¬ Period 6
Period 7
¬ Period 7
Period 8
¬ Period 8
10011001 10011001 10011001 10011001 10011001 10011001 10011001 10011001 10011001 10011001
01100110 01100110 01100110 01100110 01100110 01100110 01100110 01100110 01100110 01100110
11001100 01110011 00011100 11000111 00110001 11001100 01110011 00011100 11000111 00110001
00110011 10001100 11100011 00111000 11001110 00110011 10001100 11100011 00111000 11001110
00011100 01110001 11000111 00011100 01110001 11000111 00011100 01110001 11000111 00011100
11100011 10001110 00111000 11100011 10001110 00111000 11100011 10001110 00111000 11100011
10001110 00011110 00111000 01111000 11100001 11100011 10000111 10001110 00011110 00111000
01110001 11100001 11000111 10000111 00011110 00011100 01111000 01110001 11100001 11000111
11100001 11100001 11100001 11100001 11100001 11100001 11100001 11100001 11100001 11100001
00011110 00011110 00011110 00011110 00011110 00011110 00011110 00011110 00011110 00011110
Sequence name Sequence
TABLE II: List of control sequences for fine beam steering
All OFF
All ON
00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111
Period 5.5
¬ Period 5.5
Period 5.75
¬ Period 5.75
Period 6
¬ Period 6
Period 6.25
¬ Period 6.25
Period 6.5
¬ Period 6.5
01110001 11001110 00111001 11000111 00111000 11100011 00011100 01100011 10001100 01110001
10001110 00110001 11000110 00111000 11000111 00011100 11100011 10011100 01110011 10001110
11100111 00011100 01110001 11001110 00111000 11100011 10001100 01110001 11000111 00011000
00011000 11100011 10001110 00110001 11000111 00011100 01110011 10001110 00111000 11100111
10001110 00111000 11100011 10001110 00111000 11100011 10001110 00111000 11100011 10001110
01110001 11000111 00011100 01110001 11000111 00011100 01110001 11000111 00011100 01110001
00111100 01110001 11000111 00011110 00111000 11100011 10000111 00011100 01110001 11000011
11000011 10001110 00111000 11100001 11000111 00011100 01111000 11100011 10001110 00111100
01110001 11100011 10001111 00011100 01111000 11100001 11000111 00001110 00111000 01110001
10001110 00011100 01110000 11100011 10000111 00011110 00111000 11110001 11000111 10001110
Sequence name Sequence
TABLE III: List of control sequences for beam shaping
All OFF
All ON
00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111
Period 6
10001110 00111000 11100011 10001110 00111000 11100011 10001110 00111000 11100011 10001110
¬ Period 6
01110001 11000111 00011100 01110001 11000111 00011100 01110001 11000111 00011100 01110001
Chirp 1
01110011 10001100 01110001 10001110 00111000 11100011 10001110 00011100 01111000 11100001
¬ Chirp 1
10001100 01110011 10001110 01110001 11000111 00011100 01110001 11100011 10000111 00011110
Chirp 2
00111001 11001110 01110001 11001110 00111000 11100011 10001111 00011100 00111000 01111000
¬ Chirp 2
11000110 00110001 10001110 00110001 11000111 00011100 01110000 11100011 11000111 10000111
Dual beam 11001100 11001100 11001100 11001100 11001100 11110000 11110000 11110000 11110000 11110000
¬ Dual beam 00110011 00110011 00110011 00110011 00110011 00001111 00001111 00001111 00001111 00001111
Sequence name
Sequence
TABLE IV: List of control sequences for geometric PSK
All OFF
All ON
00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111
4-GPSK Symbol 0 10011001 10011001 10011001 10011001 10011001 10011001 10011001 10011001 10011001 10011001
4-GPSK Symbol 1 00110011 00110011 00110011 00110011 00110011 00110011 00110011 00110011 00110011 00110011
4-GPSK Symbol 2 01100110 01100110 01100110 01100110 01100110 01100110 01100110 01100110 01100110 01100110
4-GPSK Symbol 3 11001100 11001100 11001100 11001100 11001100 11001100 11001100 11001100 11001100 11001100
6-GPSK Symbol 0 00011100 01110001 11000111 00011100 01110001 11000111 00011100 01110001 11000111 00011100
6-GPSK Symbol 1 00111000 11100011 10001110 00111000 11100011 10001110 00111000 11100011 10001110 00111000
6-GPSK Symbol 2 01110001 11000111 00011100 01110001 11000111 00011100 01110001 11000111 00011100 01110001
6-GPSK Symbol 3 11100011 10001110 00111000 11100011 10001110 00111000 11100011 10001110 00111000 11100011
6-GPSK Symbol 4 11000111 00011100 01110001 11000111 00011100 01110001 11000111 00011100 01110001 11000111
6-GPSK Symbol 5 10001110 00111000 11100011 10001110 00111000 11100011 10001110 00111000 11100011 10001110
8-GPSK Symbol 0 11100001 11100001 11100001 11100001 11100001 11100001 11100001 11100001 11100001 11100001
8-GPSK Symbol 1 11000011 11000011 11000011 11000011 11000011 11000011 11000011 11000011 11000011 11000011
8-GPSK Symbol 2 10000111 10000111 10000111 10000111 10000111 10000111 10000111 10000111 10000111 10000111
8-GPSK Symbol 3 00001111 00001111 00001111 00001111 00001111 00001111 00001111 00001111 00001111 00001111
8-GPSK Symbol 4 00011110 00011110 00011110 00011110 00011110 00011110 00011110 00011110 00011110 00011110
8-GPSK Symbol 5 00111100 00111100 00111100 00111100 00111100 00111100 00111100 00111100 00111100 00111100
8-GPSK Symbol 6 01111000 01111000 01111000 01111000 01111000 01111000 01111000 01111000 01111000 01111000
8-GPSK Symbol 7 11110000 11110000 11110000 11110000 11110000 11110000 11110000 11110000 11110000 11110000
14
FIG. 10:
a, Arduino unit connected to a PCB driver. b Transistor stage used to drive the columns (there are 40 of these
in each PCB driver). c PCB support, on which the RA is glued and wirebonded. d Connections between PCB support and
drivers.
M5. Device efficiency
with the Drude model, as [6]:
Our RA is based on the modulation of the reflected sig-
nal from each cell due to the variation of σ with voltage.
Ref.6 demonstrated that specific upper bounds to the ef-
ficiency of such two-state modulators exist, and depend
uniquely on the conductivity σON and σOFF of graphene
in the two states. If the corresponding cell reflection co-
efficients in the two states are ΓON and ΓOFF then this
bound is given by the following inequality[6]:
γmod (cid:44)
ΓON − ΓOFF2
(1 − ΓON2)(1 − ΓOFF2)
≤ γR (cid:44)
σON − σOFF2
4 Re(σON) Re(σOFF)
(22)
σ at THz frequencies can be estimated analytically
σON,OFF = (RON,OFF (1 + jωτ ))−1
(23)
In our samples we have RON = 800 Ω, ROFF = 4000 Ω,
τ = 45 fs. From Eq. 22 we get:
γR =
(RON − ROFF)2(1 + ω2τ 2)
4 RON ROFF
(cid:39) 0.856
(24)
For the RA, ideally, the reflection coefficients must
have opposite phases and same absolute values.
In
practice this happens only approximately, and the ac-
tual deflected signal
is proportional to the difference
D (cid:44) ΓON − ΓOFF, while the sum S (cid:44) ΓON + ΓOFF is
responsible for an unwanted specular reflection compo-
nent. Expressing the bound in S and D we get from Eq.
22:
To columnControl2540-2540-2510-2540-2540-X1X2X3X4X6X52510-chipabcd 16D2
(4 − S2 − D2)2 − 4(Re(SD∗))2
16D2
≥
(4 − D2)2
γmod =
and:
16D2
(4 − D2)2 ≤ γR
(25)
(26)
Eq. 26 can now be inverted to find the maximum
achievable D with our SLG parameters. We get:
(cid:115)
D ≤ Dmax (cid:44)
√
4(γR + 2 − 2
γR
γR + 1)
(cid:39) 0.7832
(27)
From the measurements in Fig.1, D (cid:39) 0.5 at peak
efficiency, below the computed Dmax. This is likely due
to losses in the metal, therefore there is room for im-
provement in the cell design. The final power efficiency
of the deflected beam is given by the average differential
reflection coefficient (D/2 (cid:39) 0.25, i.e. ∼ −12 dB, while
the efficiency of the optimal device would be ∼ −8 dB.
Dmax is a strict bound that cannot be exceeded[6], how-
ever its value can be increased using SLG having higher
mobility or with gate oxide with better breakdown volt-
age and hence larger SLG tunability. The efficiency can
be further increased by reducing the temperature to in-
crease mobility and reduce the effect of thermal carriers.
[1] Khorasaninejad, M., Chen, W. T., Devlin, R. C., Oh,
J., Zhu, A. Y., Capasso, F. Metalenses at visible wave-
lengths: Diffraction-limited focusing and subwavelength
resolution imaging, Science 352, 6290, 1190-1194 (2016).
[2] Headland, D., Carrasco, E., Nirantar, S., Withayachum-
nankulm, W., Gutruf, P., Schwarz, J., Abbott, D.,
Bhaskaran, M., Sriram, M., Perruisseau-Carrier, J. et
al. Dielectric resonator RA as high-efficiency nonuniform
terahertz metasurface, ACS Photonics 3, 6, 1019-1026
(2016).
[3] Hum, S. V., Perruisseau-Carrier, J. Reconfigurable reflec-
tarrays and array lenses for dynamic antenna beam con-
trol: A review, Antennas and Propagation, IEEE Trans-
actions on 62, 1, 183-198 (2014).
[4] Debogovic, T., Bartolic, J., and Perruisseau-Carrier,
J., Dual-polarized partially reflective surface antenna
with mems-based beamwidth reconfiguration, Antennas
and Propagation, IEEE Transactions on 62, 1, 228-236
(2014).
[5] Carrasco, E, Tamagnone, M., Mosig, J. R., Low, T.,
Perruisseau-Carrier, J. Gate-controlled mid-infrared light
bending with aperiodic graphene nanoribbons array,
Nanotechnology, 26, 13, 134002 (2015).
15
[6] Tamagnone, M., Fallahi, A., Mosig, J. R., Perruisseau-
Carrier, J. Fundamental limits and near-optimal design
of graphene modulators and non-reciprocal devices, Na-
ture Phohonics 8, 7, 556-563 (2014).
[7] Mueller, J. B.,Rubin N. A., Devlin, R. C., Groever, B.,
Capasso, F. Metasurface polarization optics:
Indepen-
dent phase control of arbitrary orthogonal states of polar-
ization, Physical Review Letters 118, 11, 113901 (2017).
[8] Niu, T., Withayachumnankul, W., Upadhyay, A.,
Gutruf, P., Abbott, D., Bhaskaran, M., Sriram, S.,
Fumeaux, C. Terahertz reflectarray as a polarizing beam
splitter, Optics Express 22, 13, 16148-16160 (2014).
[9] Hasani, H. et al. Tri-band, polarization-independent re-
flectarray at terahertz frequencies: Design, fabrication,
and measurement, Terahertz Science and Technology,
IEEE Transactions on 6, 2, 268-277 (2016).
[10] Mittra, R., Chan, C. H., Cwik, T. Techniques for analyz-
ing frequency selective surfaces – a review," Proceedings
of the IEEE 76, 12, 1593-1615 (1988).
[11] Lee, J. et al. Giant nonlinear response from plasmonic
metasurfaces coupled to intersubband transitions," Na-
ture 511, 7507, 65-69 (2014).
[12] Sun Z., Martinez, A., Wang, F. Optical modulators with
2D layered materials, Nature Photonics 10, 4, 227-238
(2016).
[13] Li, Z., Yu, N. Modulation of mid-infrared light us-
ing graphene-metal plasmonic antennas, Applied Physics
Letters 102, 13, 131108 (2013).
[14] Sensale-Rodriguez, B. et al. Extraordinary control of ter-
ahertz beam reflectance in graphene electro-absorption
modulators, Nano Letters 12,9, 4518-4522 (2012).
[15] Kim, M., Jeong, J., Poon, J. K., Eleftheriades G. V.
Vanadium-dioxide-assisted digital optical metasurfaces
for dynamic wavefront engineering JOSA B 33, 5, 980-
988 (2016).
[16] Sherrott, M. C. et al. Experimental demonstration of
> 230◦ phase modulation in gate-tunable graphene-gold
reconfigurable mid-infrared metasurfaces Nano Letters
17, 5, 3027-3034 (2017).
[17] Miao, Z., Wu, Q., Li, X., He, Q., Ding, K., An, Z., Zhang,
Y., Zhou, L., Widely tunable terahertz phase modulation
with gate-controlled graphene metasurfaces, Physical Re-
view X, 5, 4, 041027 (2015).
[18] Yao, Y, Shankar, R, Kats, M. A., Song, Y., Kong, J.,
Loncar M., Capasso, F. Electrically tunable metasurface
perfect absorbers for ultrathin mid-infrared optical mod-
ulators, Nano Letters, 14, 11, 6526-6532 (2014).
[19] Rodrigo, D., Jofre, L., Perruisseau-Carrier, J. Unit cell
for frequency-tunable beamscanning reflectarrays, An-
tennas and Propagation, IEEE Transactions on 61, 12,
5992-5999 (2013).
[20] Bonaccorso, F., Sun, Z., Hasan, T., Ferrari, A. C.
Graphene photonics and optoelectronics. Nat Photon 4,
611622 (2010).
[21] Tamagnone, M., Moldovan, C., Poumirol, J.-M., Kuz-
menko, A.B.,Ionescu, A. M., Mosig, J. R., Perruisseau-
Carrier, J. Near optimal graphene terahertz non-
reciprocal
isolator, Nature Communications 7, 11216
(2016).
[22] Liu, M. et al. A graphene-based broadband optical mod-
ulator. Nature 474, 6467 (2011).
[23] Koppens, F. H. L. et al. Photodetectors based on
graphene, other two-dimensional materials and hybrid
systems. Nat Nano 9, 780793 (2014).
[24] Rodrigo, D. et al. Mid-infrared plasmonic biosensing with
graphene. Science 349, 165168 (2015).
[25] Yan, H. et al. Damping pathways of mid-infrared plas-
mons in graphene nanostructures. Nat Photon 7, 394399
(2013).
[26] Sagade, A. A. et al. Highly air stable passivation of
graphene based field effect devices, Nanoscale, 7, 8,
35583564 (2015).
[27] Balanis, C. A. Antenna Theory: Analysis and Design.
Antenna Theory: Analysis and Design (John Wiley &
Sons, 2005).
[28] Munk, B. A. Frequency selective surfaces theory and de-
sign (John Wiley & Sons, 2000).
[29] Lee, J. Y., Jiang, G. A. Displacement measurement using
a wavelength-phase-shifting grating interferometer, Op-
tics express 21, 25553-25564 (2013).
[30] Blahut, R. E. Principles and practice of information
theory. (Addison-Wesley Longman Publishing Co., Inc.,
1987).
[31] See
manufacturer
http://www.us.schott.com
data-sheet
at
[32] Williams, K. R. & Muller, R. S. Etch rates for micro-
16
machining processing. Journal of Microelectromechanical
systems 5, 256269 (1996).
[33] Li, X. S. et al. Large-Area Synthesis of High-Quality and
Uniform Graphene Films on Copper Foils. Science 324,
1312-1314 (2009).
[34] Bae, S. et al. Roll-to-roll production of 30-inch graphene
films for transparent electrodes. Nature Nanotechnology
5, 574-578 (2010).
[35] Ferrari, A. C., Basko, D. M. Raman spectroscopy as a
versatile tool for studying the properties of graphene. Na-
ture Nanotechnology 8, 235-246 (2013).
[36] Das, A. et al. Monitoring dopants by Raman scattering in
an electrochemically top-gated graphene transistor. Na-
ture Nanotechnology 3, 210-215 (2008).
[37] Bruna, M., Ott, A. K. , Ijas, M. , Yoon, D. ,Sassi, U., and
Ferrari, A. C., Doping Dependence of the Raman Spec-
trum of Defected Graphene, ACS Nano, 8, 7, 74327441
(2014).
[38] Cancado, L. G. et al. Quantifying Defects in Graphene
via Raman Spectroscopy at Different Excitation Ener-
gies. Nano Letters 11, 3190-3196 (2011).
|
1904.04887 | 1 | 1904 | 2019-04-09T19:59:25 | Electron transport through self-assembled monolayers of tripeptides | [
"physics.app-ph",
"cond-mat.mes-hall"
] | We report how the electron transport through a solid-state metal/Gly-Gly-His tripeptide (GGH) monolayer/metal junction and the metal/GGH work function are modified by the GGH complexation with Cu2+ ions. Conducting AFM is used to measure the current-voltage histograms. The work function is characterized by combining macroscopic Kelvin probe and Kelvin probe force microscopy at the nanoscale. We observe that the Cu2+ ions complexation with the GGH monolayer is highly dependent on the molecular surface density and results in opposite trends. In the case of a high density monolayer the conformational changes are hindered by the proximity of the neighboring peptides, hence forming an insulating layer in response to copper-complexation. Whereas the slightly lower density monolayers allow for the conformational change to a looped peptide wrapping the Cu-ion, which results in a more conductive monolayer. Copper-ion complexation to the high- and low-density monolayers systematically induces an increase of the work functions. Copper-ion complexation to the low-density monolayer induces an increase of electron transport efficiency, while the copper-ion complexation to the high-density monolayer results in a slight decrease of electron transport. Both of the observed trends are in agreement with first-principle calculations. Complexed copper to low density GGH-monolayer induces a new gap state slightly above the Au Fermi energy that is absent in the high density monolayer. | physics.app-ph | physics | Electron transport through tripeptides
self-assembled monolayers
Evgeniy Mervinetsky,1,2 Israel Alshanski,1,2 Stephane Lenfant,3 David Guerin,3
Leonardo Medrano Sandonas,4,6 Arezoo Dianat,4 Rafael Gutierrez,4
Gianaurelio Cuniberti,4,5,6 Mattan Hurevich,1,2
Shlomo Yitzchaik,1,2 and Dominique Vuillaume3
1) Institute of Chemistry, The Hebrew University of Jerusalem, Safra campus,
Givat Ram, Jerusalem, 91904, Israel.
2) Center for Nanoscience and Nanotechnology, The Hebrew University of
Jerusalem, Safra campus, Jerusalem, 91904, Israel.
3) Institute for Electronics Microelectronics and Nanotechnology, CNRS, Univ.
Lille, 59652, Villeneuve d'Ascq, France.
4) Institute for Materials Science and Max Bergmann Center of Biomaterials, TU
Dresden, 01069 Dresden, Germany.
5) Dresden Center for Computational Materials Science, TU Dresden, 01062
Dresden, Germany.
6) Center for Advancing Electronics Dresden, TU Dresden, 01062 Dresden,
Germany
Emails : [email protected]; [email protected];
[email protected]
KEYWORDS. peptide monolayer, electron transport, work function, Cu
complexation, bio-molecular electronics.
ABSTRACT. We report how the electron transport through a solid-state metal/
Gly-Gly-His tripeptide (GGH) monolayer/metal junction and the metal/GGH
work function are modified by the GGH complexation with Cu2+ ions. Conducting
AFM is used to measure the current-voltage histograms. The work function is
characterized by combining macroscopic Kelvin probe and Kelvin probe force
microscopy at the nanoscale. We observe that the Cu2+ ions complexation with the
GGH monolayer is highly dependent on the molecular surface density and results
in opposite trends. In the case of a high density monolayer the conformational
changes are hindered by the proximity of the neighboring peptides, hence forming
an insulating layer in response to copper-complexation. Whereas the slightly
lower density monolayers allow for the conformational change to a looped peptide
wrapping the Cu-ion, which results in a more conductive monolayer. Copper-ion
complexation to the high- and low-density monolayers systematically induces an
!2
increase of the work functions. Copper-ion complexation to the low-density
monolayer induces an increase of electron transport efficiency, while the copper-
ion complexation to the high-density monolayer results in a slight decrease of
electron transport. Both of the observed trends are in agreement with first-
principle calculations. Complexed copper to low density GGH-monolayer induces
a new gap state slightly above the Au Fermi energy that is absent in the high
density monolayer.
!3
INTRODUCTION
Molecular devices made of peptides molecules and proteins are gaining interest
as nanoscale devices in bioelectronics.1-2 Understanding the electron transport
mechanisms through these biomolecules is a key issue in biology. Similarly,
biomolecules can be used to build and study various bioelectronic nanomaterials
and devices.3-6 For example, the "doping" of a polypeptide chain (7-alanine) by
one single tryptophan substitution enhances the electron transport, a mechanism
ascribed to the introduction of an additional energy level close to the Fermi energy
of the Au electrodes.4 Single molecule measurements (scanning tunneling
microscope break junction) show that the conductance of a metal/polypeptide/
metal junction can be controlled by the pH of the surrounding media due to a
conformational change from a compact helical to a more extended structure,7 and
pH titration was also demonstrated from single peptide conductance
measurements.8 The electron transport through a helical peptide is also dependent
whether the electrons are injected parallel or anti-parallel to the peptide dipole.9
Combining peptides and redox species (e.g. ferrocene),10-12 or π-conjugated
moieties,3 is also a powerful way to modify the electron transport properties of the
peptides and tailor electronic functionality in biologically relevant
macromolecules. Spin-dependent transport through chiral peptides has also been
demonstrated as an example of the CISS (chirality induced spin selectivity) effect.
13
!4
Some of us have reported that a self-assembled monolayer of Gly-Gly-His
tripeptide (GGH) is a very sensitive electrochemical sensor of copper ions.14 The
GGH molecules in the monolayer chelate with the Cu2+ ions by a conformational
change, forming a dense barrier, which prevent the access of redox species to the
electrode during the electrochemical impedance spectroscopy measurements,
leading to an increase of the electrochemical impedance.
Here, we study the electronic properties of a solid-state metal/GGH
monolayer/metal junction and we examine how they are modified by the GGH
reaction with Cu2+ ions. The electron transport through the metal/GGH
monolayer/metal junction is studied by measuring the current versus voltage
curves histograms at the nanoscale with a conducting AFM. The work function of
metal/GGH is characterized by combining macroscopic Kelvin probe and Kelvin
probe force microscopy at the nanoscale. The self-assembled monolayers (SAM)
of GGH were also characterized by ellipsometry, XPS and topographic AFM to
assess the formation of the SAM on gold ultra-flat electrodes and the reaction with
Cu2+ ions. We study two samples prepared to have a slightly different molecular
packing of the molecules in the SAMs. In both cases, we observe that the Cu2+
ions reaction with the GGH monolayer systematically induces an increase of the
work function, in agreement with Density-Functional Theory (DFT) calculations.
The effect of Cu2+ exposure on the electron transport through the metal/GGH/C-
AFM junction depends on the GGH molecular packing. For the denser SAM, the
current passing though the metal/GGH/C-AFM junctions is slightly reduced when
!5
exposed to Cu2+ ions, while it is increased (about a decade) for the slightly less
dense SAM. DFT calculations help to rationalize these results. In the less dense
case, the GGH undergoes a large conformational change (GGH folding around the
Cu2+) to fully chelate the Cu2+ ion (Cu2+ surrounded by the 4 N atoms of the
GGH) with square-planar configuration, while in the denser SAM, due to steric
hindrance, the Cu2+ ions are partly chelated with less than 4 N atoms of the GGH
(figure 1). In this former case, DFT calculations show that the increase of electron
transport is consistent with the modification of the molecular orbitals at the metal/
GGH system and the appearance of metal (Cu) induced gap state slightly above
the Au Fermi energy.
METHODS
GGH self-assembled monolayers - The SAMs were formed on template-
stripped Au (TSAu) electrodes. The very flat TSAu surfaces were prepared
according to the method developed by the Whitesides group.15 In brief, a 300−500
nm thick Au film was evaporated on a very flat silicon wafer covered by its native
SiO2 (rms roughness of ~0.4 nm), which was previously carefully cleaned by
piranha solution (30 min in 7:3 H2SO4/H2O2 (v/v); Caution: Piranha solution is a
strong oxidizer and reacts exothermically with organics), rinsed with deionized
(DI) water, and dried under a stream of nitrogen. A clean glass piece
(ultrasonicated in acetone for 5 min, ultrasonicated in 2-propanol for 5 min, and
UV irradiated in ozone for 10 min) was glued (UV polymerizable glue) on the
!6
evaporated Au film and mechanically stripped with the Au film attached on the
glass piece (Au film is cut with a razor blade around the glass piece). This very
flat (rms roughness of ~0.4 nm, the same as the SiO2 surface used as template)
and clean template-stripped TSAu surface was immediately used for the formation
of the SAM.
Ellipsometry - We recorded spectroscopic ellipsometry data in the visible range
using an UVISEL (Jobin Yvon Horiba) spectroscopic ellipsometer equipped with
DeltaPsi 2 data analysis software. The system acquired a spectrum ranging from 2
to 4.5 eV (corresponding to 300 to 750 nm) with intervals of 0.1 eV (or 15 nm).
Data were taken at an angle of incidence of 70°, and the compensator was set at
45°. Data were fitted by a regression analysis to a film-on-substrate model as
described by their thickness and their complex refractive indexes. First, a
background before monolayer deposition for the gold coated substrate was
recorded. Secondly, after the monolayer deposition, we used a 2-layer model
(substrate/SAM) to fit the measured data and to determine the SAM thickness. We
employed the previously measured optical properties of the gold coated substrate
(background), and we fixed the refractive index of the organic monolayer at 1.50.
The usual values in the literature for the refractive index of organic monolayers
are in the range 1.45−1.50.16 We can notice that a change from 1.50 to 1.55 would
result in less than 1 Å error for a thickness less than 30 Å. We estimated the
accuracy of the SAM thickness measurements at ± 2 Å.
!7
X-ray Photoelectron Spectroscopy - XPS was performed with a Physical
Electronics 5600 spectrometer fitted in an UHV chamber with a residual pressure
of 2×10 -- 10 Torr. High resolution spectra were recorded with a monochromatic Al
Kα X-ray source (hν = 1486.6 eV), a detection angle of 45° as referenced to the
sample surface, an analyzer entrance slit width of 400 µm and with an analyzer
pass energy of 12 eV. Semi-quantitative analysis were completed after standard
background subtraction according to Shirley's method.17 Peaks were decomposed
by using Voigt functions and a least-square minimization procedure and by
keeping constant the Gaussian and Lorentzian broadenings for each component of
a given peak.
Kelvin Probe - Contact potential difference (CPD) was measured on large area
samples with Kelvin Probe S (DeltaPhi Besocke, Jülich, Germany), with a
vibrating gold electrode (work function 5.1 eV) in a home-built Faraday cage
under Ar (argon) atmosphere.
Kelvin Probe Force Microscopy - KPFM measurements were carried out at
room temperature with a Dimension 3100 from Veeco Inc., purged with a flow of
dry nitrogen atmosphere. We used Pt/Ir tip (PPP-EFM-50 from Nanosensors) with
spring constant of ca. 3 N/m and a resonance frequency of ca. 70 kHz.
Topography (tapping mode AFM) and KPFM data were recorded using a standard
two-pass procedure, in which each topography line acquired in tapping mode is
followed by the acquisition of KPFM data in a lift mode, with the tip scanned at a
distance z ~ 100 nm above the sample so as to discard short range surface forces
!8
and be only sensitive to electrostatic forces. DC and AC biases (VDC + VAC
sin(ωt)) are applied to the cantilever with VAC = 2 V. Experimentally, the contact
potential difference (CPD) is measured using a feedback loop which sets to zero
the cantilever oscillation amplitude by adjusting the tip DC bias VDC. The work
function (WF) of the sample was deduced from the CPD following the relation,
WF = Wtip - e.CPD, where e is the elementary charge and Wtip the work function
of the KPFM tip. Since we focused on the WF modifications, the exact value of
Wtip is not an issue, and we simply have δWF = - e.δCPD. We recorded images (1
μm x 1 μm) at 3-4 different zones on the sample, from which we constructed the
CPD histograms.
Conducting Atomic Force Microscope - Conducting atomic force microscopy
(C-AFM) was performed under a flux of N2 gas (ICON, Bruker), using a tip probe
in platinum/iridium (SCM-PIC v2 from Bruker). The tip loading force on the
surface was fixed ≤ 50 nN to avoid a too important strain-induced deformation of
the monolayer (≲ 0.3 nm).18 A square grid of 10×10 was defined with a pitch of
100 nm. At each point, the I-V curve is acquired leading to the measurements of
100 I-V traces. This process was repeated 3 times at different places on the
sample, and the 300 I-V traces were used to construct the current-voltage
histograms. The bias was applied on the TSAu substrate and the tip was grounded
through the input of the current amplifier. Note that around 0V the currents are
very weak and in many cases at the limit of detection (0.1 pA). Thus it is likely
!9
that any "asymmetry" in the IV around 0V induced by the existing surface
potential (see KPFM) in the SAMs is not observable.
Computational modeling - The adsorption of Lpa-GGH onto Au(111) surface,
the corresponding interaction properties with Cu2+ ions as well as the calculation
of the work function were theoretically addressed at the Density-Functional
Theory (DFT) level. We used mixed Gaussian plane wave (GPW) methods with
the standard implementation in the CP2K package.19 Here, the Kohn-Sham
orbitals are expanded into linear combinations of contracted Gaussian type
orbitals and complemented by a plane-wave basis set in order to compute the
electronic charge density. In all calculations, the PBE (Perdew, Burke, and
Ernzerhof) exchange-correlation functional was used,20 and its corresponding
norm-conserving pseudo-potential GTH (Goedecker, Teter and Hutter).21 Finally,
a DZVP (double zeta for valence electrons plus polarization functions) basis set
complemented with a plane-wave basis set energy cut-off 350 Ry was employed
and dispersion corrections were included through the standard D2-Grimme
parameterization.22
RESULTS
GGH self-assembled monolayers
The synthesis of the Lpa-GGH (lipoic acid Gly-Gly-His) tripeptide was
reported by some of us in a previous work.14 Formation of low- and high-density
tripeptide monolayers (LDP and HDP respectively) were prepared by dipping
!10
cleaned TSAu substrates (see Methods) either in a 10 μM or 250 μM solution of
Lpa-GGH in DI water for 2h, then rinsed copiously with DI water and dried under
dry N2 stream. The change in the concentration resulted in a LDP and HDP
monolayers (see below).
For the Cu2+ exposure, both LDP and HDP monolayers were dipped
overnight into a solution of 10 μM Cu(NO3)2 and 640 μM HNO3 in DI water, then
rinsed copiously with DI water and dried under dry N2 stream. To avoid any
measurement bias and irreproducible results, we used two protocols (namely A
and B, Fig. S1 in the Supporting Information) to study the effect of Cu2+ exposure
on electron transport and work function. We started by preparing two samples in
the same Lpa-GGH solution. Sample A was characterized after the SAM
formation and then exposed to Cu2+ and measured again. As a control experiment,
sample B was immediately dipped in the Cu2+ solution and then measured.
Protocol B checks that the measurements before Cu2+ exposure have not induced
any contamination during the measurements of the as-prepared sample. Protocol A
ensures that the comparison without and with Cu2+ was made on the same sample.
Thus, if we observe the same trends (at least qualitatively) between samples A and
B, we are reasonably sure that we have no measurement bias, no contamination
and no strong irreproducibility. We mainly report data obtained with protocol A
(unless specified), and the control data (protocol B) are given in the supporting
information.
!11
Physical characterization of the GGH SAM
First, the TSAu-Lpa-GGH SAMs were characterized by spectroscopic
ellipsometry to determine their thicknesses. The high-density peptide (HDP)
monolayers have a thickness of 2.2 ± 0.2 nm in good agreement with the
calculated length (2.3 nm) of the Lpa-GGH molecule in its extended conformation
on Au (see Modeling section). After exposure to Cu2+, the thickness is not
significantly modified 2.2 ± 0.2 nm. These values are consistent with formation of
a highly dense SAM on the TSAu surface. The low-density peptide (LDP)
monolayers have a thickness of 1.8 ± 0.2 nm indicating an average tilt angle of
~40° to the surface normal. After exposure to Cu2+, the thickness is not
significantly modified 1.7 ± 0.2 nm. We refer, in the following of the text, to the
two SAMs as HDP and LDP, respectively.
XPS analysis of the HDP SAMs. We observed Au4f 7/2 and Au4f 5/2 (80.0
and 87.7 eV, respectively), C1s (complex signal containing 3 peaks at 284.6, 286.4
and 288.3 eV), N1s (400.2 eV), O1s (531.7 eV). Figure S2 shows S2p3/2 and S2p1/2 at
161.9 and 163.1 eV, respectively (exclusively sulfur atoms bound to Au)
indicating that dithiolane linkers are fully chemisorbed on gold through dithiolate
bonds. The ratios of measured atomic concentrations were in agreement with the
expected values (table S1). After Cu2+ exposure, the same elements are present
and the Cu2p3/2 and Cu2p1/2 peaks (932.6 and 952.4 eV, respectively) are clearly
observed (figure S3). These peaks are significantly shifted compare to typical XPS
Cu2+ peak (933.6 eV), which is attributed to copper chelation. These results
!12
confirm the formation of the Lpa-GGH SAM and its complexation with Cu2+
They are in good agreement with our previous report.14
Tapping mode-AFM images of the bare TSAu surface, of the HDP SAMs
before and after the Cu2+ exposure were recorded during the KPFM measurements
(see Methods). These images (Figure 2) reveal a rather homogeneous SAM
formation and no modification after the Cu2+ exposure. The root mean square
(rms) of roughness is a parameter to quantify this homogeneous SAM formation.
We have measured the same rms roughness for the TSAu, the HDP SAM (0.40 ±
0.02 nm) and the HDP+Cu2+ SAM (0.39 nm ± 0.01 nm). These results again
confirm the formation of highly dense Lpa-GGH SAMs.
Electrical characterization of the GGH SAMs
For the electrical measurements at the nanoscale, we used two instruments in
parallel, one for the C-AFM and one for the KPFM measurements (see Methods).
The objective was to avoid any variations in the set-up experiments during the
complete sequence of measurements. Since the tip for the C-AFM and KPFM
measurements were not the same, using the same instrument and changing the tip
back and forth makes difficult to recover the same parameters. The parameters of
these two instruments were fixed and kept constant for the complete set of
measurements (to ensure a reliable comparison between reference sample (TSAu-
Lpa-GHH SAMs before Cu2+ exposure) and after the Cu2+ exposure for both
protocols A and B. Thus, measurements of a given set of samples were done in the
same session, same configuration, conditions, and with the same tip.
!13
We measured the work function modification of the HDP SAM upon Cu2+
exposure by macroscopic Kelvin probe (KP, see Methods) and at the nanoscale by
Kelvin probe force microscopy (KPFM). The contact potential difference (CPD)
is reduced by about 70 meV (considering the mean of the CPD distribution, see
figure S4) after the Cu2+ exposure, which corresponds to an increase of the WF by
70 meV (Fig. 3). The macroscopic KP measurements give an increase of WF by
75 meV after Cu2+ exposure of HDP surface (Fig. 3). KPFM measurements for the
LDP SAM shows an increase of the WF by about 30 meV (CPD values in Figure
S4, Supporting Information) in agreement with the trend observed for the denser
SAM, albeit the WF increase is smaller (see Fig. 3). The same trends were
observed for the control experiments (protocol B) for both LDP and HDP SAMs
(figure S4 in the Supporting Information). The change in WF for peptide layer
after exposure to Cu2+ is explained by peptide-copper complexation, which is
attributed to peptide conformational changes. These conformational changes alter
the dipole of the layer and hence have a significant influence on the WF. The KP
and KPFM similar values proves that the microscopic and macroscopic KP
analysis are in agreement thus indicates the homogeneity of the layer and supports
effect of conformational changes.
The current-voltage (I-V) curve of the HDP SAM and the HDP+Cu2+
SAMs were measured by C-AFM and the current histograms at 1 V constructed
form 300 I-V measurements are shown in Figure 4. The histograms are fitted by
log-normal distributions. The parameters, log-mean current (log μ) and log-
!14
standard deviation (log σ) are given in Table 1. The Cu2+ exposure induces a slight
decrease of log μ (at 1V) from -11.12 (i.e. 7.6x10-12 A) to -11.90 (1.26x10-12 A).
The same trend was observed for the control experiment, see Figure S5. Thus, the
Cu2+ exposition induces a slight decrease of the current through the HDP SAM.
However, in that case of the LDP SAM, the C-AFM measurements reveal an
increase of the current after Cu2+ exposure (Fig. 5, Table 1), from log μ (at 2V) =
-11.77 (i.e. 1.7x10-12 A) to -10.77 (1.7x10-11 A), with again a similar trend (log μ
(at 2V) -10.13 (7.4x10-11 A)) for the control experiments (Figure S6).
TSAu-Lpa-GGH
TSAu-Lpa-GGH+Cu2+
HDP
LDP
log µ
-11.12
(7.6×10-12 A)
-11.90
(1.26×10-12 A)
log σ
0.7
log µ
-11.77
(1.7×10-12 A)
log σ
0.61
0.52
-10.77
0.75
(1.7×10-11 A)
Table 1. Fitted parameters, log-mean current (log μ), and log-standard deviation
(log σ) of the normal distributions shown in Figs. 4, 5, S5 and S6 (Supporting
Information).
Modeling
We have first addressed for the sake of reference the electronic structure and the
corresponding work function changes for the case studied in a previous
!15
publication,14 where only structural issues were discussed. As we showed, Lpa-
GGH covalently bonds to the Au(111) surface by breaking the di-sulfide bond
and, upon Cu2+ chelation, the molecule experiences a conformational change from
a nearly linear configuration to a square-planar coordination where four N atoms
englobe the Cu2+ ion. In the different panels of Fig. 6 various components of the
projected electronic density of states (PDOS) are displayed for both geometrical
conformations of the Lpa-GGH without and with Cu2+ binding. Shown are
projections on the Au substrate (grey background), on the Lpa-GGH molecule,
and on the Cu and N atoms belonging to the molecule. The states close to the
Fermi energy EF have a strong contribution from the N atoms and, upon chelation
of Cu2+, also Cu-derived states emerge there (see the lower panel of Fig. 6). As a
result, there is considerable hybridization of N-based and Cu-based orbitals. The
hybridization takes place mostly between the N p-states and Cu d-states, with
some additional weight coming from the O p-states (As shown in Figure S7 of the
Supporting Information). We also remark that the state slightly above the Fermi
level has a small electronic occupation and may be interpreted as a metal (Cu)-
induced gap state, where gap refers to the molecular HOMO-LUMO gap of the
isolated GGH molecule. We may expect that this state should contribute to the
electronic transport in the linear response (low-bias) limit, thus providing a
transport channel that would be absent otherwise. Overall, the Cu2+ binding shifts
the states below the Fermi level closer to it.
!16
To shed further light into the experimental findings related to SAMs with
low and high packing density, we have also considered different scenarios for the
Cu2+ ion binding in the Lpa-GGH/Au(111) system; they are labeled as
conformations I-IV in Fig. 7. They differ from each other in the number of N
atoms (1 to 4) involved in the binding to Cu2+, conformation C-IV corresponding
to the tetradentate N-donor square planar complex previously discussed.14 The
other complexes, C-I (monodentated), C-II (bidentated), and C-III (tridentated)
turned out to be energetically less favorable when considering the Cu-binding to
single GGH molecules. The binding process within an HDP monolayer may favor
such complexes because of steric hindrance that prevents the formation of the
more stable square planar complex C-IV. The corresponding work functions (WF)
of these systems have been calculated by using the relation WF = -e F(∞) − EF,
where EF is the bulk Fermi energy, e is the electron charge, and F(∞) is the
asymptotic electrostatic potential in vacuum. F(∞) is extracted from the in-plane-
! Φ(z) =(1/ A)
dxdyΦ(x, y,z)
A∫∫
averaged potential
with A being the area of the
surface unit cell. We have found that the WF of the bare Au(111) surface
(WF=5.32 eV) decreases after attachment of Lpa-GGH to a value of 3.98 eV (see
the inset in Fig. 7). Upon ion binding, the WF increases again smoothly from Case
IV to Case I, attaining in the latter case its largest value of ~5.62 eV, i.e. situations
with a HDP SAM may yield larger WFs as for LDP.
The change in the work function after ion binding obviously correlates
with the modification of the interface dipole Δμ, as follows from Helmholtz
!17
!ΔWF =(e/ µ0A)Δµ
. In the most stable situation, case IV, we found a
equation:
strong modification of the interface dipole, manifesting in a change from +7.3
Debye for Lpa-GGH on Au(111) to -1.3 Debye after binding of Cu2+. The
molecular dipole moment μ is computed according to the relation
eµ
ε0A
V(+∞) −V(−∞) =
, where V(+∞) and V(-∞) are the asymptotic electrostatic
potentials on both sides of the SAM, e is the electron charge, A is the surface area,
and ε0 the vacuum dielectric constant. These are easily obtained from our
electronic structure calculations, since the electrostatic potential reaches its
asymptotic values within a distance of few Å from the metal-molecule interface.23
These changes of the dipole moment relate to the charge rearrangement at the
molecule/surface system resulting from ion binding and the associated
conformational change. The relatively strong change in the WFs of roughly 1 eV
after ion binding found in our calculations may be related to the fact that no
depolarization effects were taken into account in our calculations. In Figures S8
and S9 (Supporting Information) the PDOS for all cases are shown, projected on
individual atomic subsets as well as orbital-resolved. Here, we see that very close
to the Fermi energy states arising from Cu(d)-N(p) state hybridization are
dominating. In configuration IV, there is additionally a split state slightly above
the Fermi energy, which contains contributions from N p-, O p-, and Cu p-states
(see panel d) in Figure S8).
!18
DISCUSSION
The increase of the WF after Cu2+ exposition measured by KP and KPFM are in
good agreement (around 70 meV, see Fig. 3) and qualitatively supported by theory
(increase by 1 eV if we consider Cu2+ ions surrounded by a ring with 4 N atoms,
Fig. 7). The experimental values are lower than theory, which can be explained by
the fact that the depolarization effect24-25 in the SAM, which is known to reduce
the dipole per molecule,26-28 were not taken into account in the calculation (single
molecule).
For the current-voltage measurements by C-AFM, we first consider that
after Cu2+ exposure, the Lpa-GGH peptide undergoes a conformational change
forming tetradentate N-donor square planar complex (C-IV complex) of the Cu2+
ion as reported in a previous work.14 According to the DFT calculations of C-IV
(Fig. 6), the HOMO moves closer (at ~ -0.55 eV) to the metal Fermi energy EF,
and an additional, partially occupied state appears just above EF (at ~ +0.26 eV).
Consequently, we would expect an increase of the current through the molecular
junction, since these additional molecular states can fall in the energy window
opened by applying a voltage between the two electrodes of the TSAu/SAM/C-
AFM tip junction. This scenario is consistent with the C-AFM measurements on
the less dense SAM (Fig. 5). We note that this effect (increase of electron
transport) resembles that recently observed when "doping" a polypeptide (7-
alanine) by substituting with a single tryptophan unit introduces an energy level
near the Fermi energy.4 This is also consistent with the observation that
!19
incorporating chelated metal atoms in a molecular chain significantly improves
the conductance of the molecular junction.29
We have observed (Fig. 4) only a slight decrease of the current for the
denser SAM. We explain this feature by assuming that, in the present case, the
steric hindrance prevents a complete folding and only a partial coordination (i.e.
cases I to III depicted in Fig. 7) is allowed. In that case, the DFT calculations (see
panels a-c in both Figs. S6 and S7 in the Supporting Information) show that no
additional state appears above the Fermi energy, and we do not expect a clear
increase of the current in the molecular junction. However, even in these cases,
the electronic structures calculated for the Cu2+ induced partial conformational
changes (forming C-I, C-II, and C-III complexes) show that the HOMO level is
moving closer to the Au Fermi energy as in the C-IV complex case (Figs. S6 and
S7 in the supporting information). In principle, this feature alone can also induce
an increase of the current through the molecular junction, which is not observed
for the denser sample (Fig. 4). This point deserves more detailed calculations of
the electron transfer probability and I-V curves taken into account the molecule/
tip (PtIr) interface (for example, the contact with high WF tip can counterbalance
the shift of the HOMO level due to charge transfer at the molecule/tip interface),
while the presence of an additional electron transmission channel (as in C-IV)
induces the observed increase of current.
Finally, we assume that the surface roughness determines the ability of the
peptide to acquire a full conformational change to form C-IV complex (square
!20
planar coordination). In this study we used ultra flat gold while in our previous
study14 a gold substrate with rougher morphology was used. We assumed that a
denser monolayer could be assembled on flat TSAu substrate (rms roughness of 0.4
nm, Fig. 2) while the rougher surface (evaporated Au surface with rms roughness
of ~0.9 nm) leads to a less packed monolayer. The lower density GGH monolayer
allows the full conformational change to yield C-IV complex while this complex
is less likely to form in the high-density GGH monolayer on TSAu substrate. Our
TM-AFM (Fig. 2) and ellipsometry measurements show that there is no
significant morphological or topographic changes induced by Cu2+ binding thus
suggests that the C-IV complex can't be formed on the high-density GGH
monolayer.
CONCLUSION
The LDP assembled on gold results in an increase of work function and
enhanced conductivity in response to Cu2+ complexation. On the other hand, the
HDP assembled on gold results in larger increase in work function and decrease in
electrical current passing through this monolayer. We attribute these differences to
the ability of the peptide to adopt a square planar coordination with copper in the
LDP and the steric hindrance in the HDP that prevents such conformational
changes. Our DFT calculations proved that the significant increase of the current
through the LDP-Cu monolayer can result from the movement of the molecular
orbitals closer to the gold Fermi energy of the electrodes and an additional
!21
electronic state that appears above the gold Fermi energy that is unique to the C-
IV complex.
ASSOCIATED CONTENT
Supporting Information. The following information is available in the
Supporting Information : protocol of measurements, XPS data, CPD values from
KPFM measurements, current-voltages curves and hisotgrams for the HDP and
LDP control SAMs (protocol B), additional PDOS calculated for the Lpa-GGH
and Lpa-GGH SAMs with different conformations of the chelated Cu2+ ion.
AUTHOR INFORMATION
Corresponding authors.
[email protected]
[email protected]
[email protected]
ORCID
Evgeniy Mervinetsky : 0000-0002-4373-1263
Israel Alshanski : 0000-0002-9310-1921
Stephane Lenfant : 0000-0002-6857-8752
Gianaurelio Cuniberti : 0000-0002-6574-7848
Mattan Hurevich : 0000-0002-1038-8104
!22
Shlomo Yitzchaik : 0000-0001-5021-5139
Dominique Vuillaume : 0000-0002-3362-1669
Notes
The authors declare no competing financial interest.
Acknowledgements
The authors would like to thank RECORD-IT project. This project has received
funding from the European Union's Horizon 2020 research and innovation
program under grant agreement No 664786. SY is the Binjamin H. Birstein Chair
in Chemistry. This work has also been partly supported by the German Research
Foundation (DFG) within the Cluster of Excellence "Center for Advancing
Electronics Dresden". We acknowledge the Center for Information Services and
High Performance Computing (ZIH) at TU Dresden for computational resources.
!23
Figure 1. Schematic representation of conformational changes of the Lpa-GGH
monolayers in response to Cu2+ chelation. Top: sterically hindered chelation for
high-density peptide (HDP) monolayer; Bottom: unperturbed chelation for lower
density peptide (LDP) monolayer.
!24
!
Figure 2. Tapping Mode-AFM images of (a) the naked TSAu substrate, rms
roughness of 0.40 ±0.02 nm, (b) the HDP SAM, rms roughness of 0.40 ±0.02 nm,
!25
!
and (c) HDP SAM after Cu2+ exposure, rms roughness of 0.39 ±0.01 nm. Scale
bars are 200 nm.
Figure 3. Work function variations (average values) for the HDP SAM exposed to
Cu2+ measured by KP and KPFM and for LDP SAM measured by KPFM.
!26
"
Figure 4. 2D histograms of the current-voltage curves (300 I-V traces, tip loading
force 50 nN) and corresponding current histograms at 1V for: (a) HDP SAM, (b)
HDP SAM exposed to Cu2+. The current histograms at 1 V are fitted (black lines)
with log-normal distributions (fitted parameters in Table 1). In the I-V histograms,
the white lines are the sensitivity limit of the C-AFM system (~0.1 pA).
!27
Figure 5. 2D histograms of the current-voltage curves (300 I-V traces, tip loading
force 29 nN) and corresponding current histograms at 2V for the less densely
packed SAM (at 10 μM): (a) LDP SAM, (b) LDP SAM exposed to Cu2+. The
current histograms at 2 V are fitted (black lines) with log-normal distributions
(fitted parameters in Table 1). In the I-V histograms, the white lines are the
sensitivity limit of the C-AFM system (~0.1 pA).
!28
Figure 6. Projected Density of States (PDOS) on various atomic subsets for the
Lpa-GGH molecule chemisorbed on Au(111) (upper panel) and upon binding a
Cu2+ ion in the C-IV complex (lower panel). In this later case, electronic states
shortly above the Fermi energy (set at zero) result from the hybridization between
Cu d-states and N p-states (see also Figure S7 in Supporting Information). The
corresponding structural conformations are inserted in both panels.
!29
!
Figure 7. Calculated work functions for the four different conformations
considered in this study C-I to C-IV (Cu coordinated to 1, 2, 3 and 4 N atoms,
respectively) and described in the main text. Independently of the studied
conformation, the WF always increases upon Cu2+ chelation. Inset: Variation of
the electrostatic potential along the z-direction for the different relevant cases:
bare Au(111), Lpa-GGH on Au(111), and configuration C-IV: Lpa-GGH/Cu2+ on
Au(111). Indicated are also the computed values of the work functions for the
three cases. After Cu2+ chelation the WF increases by 1.06 eV for the most stable
configuration (C-IV). The horizontal arrows indicate the change in molecular
length related to the molecular conformational change (going from 2.3 nm down
to 1.7 nm).
!30
!
REFERENCES
Amdursky, N.; Marchak, D.; Sepunaru, L.; Pecht, I.; Sheves, M.; Cahen,
1.
D. Electronic Transport Via Proteins. Adv. Mater. 2014, 26, 7142-7161.
Fereiro, J. A.; Yu, X.; Pecht, I.; Sheves, M.; Cuevas, J.-C.; Cahen, D.
2.
Tunneling Explains Efficient Electron Transport Via Protein Junctions.
Proceedings of the National Academy of Sciences 2018, 115, E4577-E4583.
Ardoña, H. A. M.; Tovar, J. D. Peptide Π-Electron Conjugates: Organic
3.
Electronics for Biology? Bioconjugate Chem. 2015, 26, 2290-2302.
Guo, C.; Yu, X.; Refaely-Abramson, S.; Sepunaru, L.; Bendikov, T.; Pecht,
4.
I.; Kronik, L.; Vilan, A.; Sheves, M.; Cahen, D. Tuning Electronic Transport Via
Hepta-Alanine Peptides Junction by Tryptophan Doping. Proceedings of the
National Academy of Sciences 2016, 113, 10785-10790.
Tao, K.; Makam, P.; Aizen, R.; Gazit, E. Self-Assembling Peptide
5.
Semiconductors. Science (New York, NY) 2017, 358, eaam9756.
Yitzchaik, S.; Gutiérrez, R.; Cuniberti, G.; Yerushalmi, R. Diversification
6.
of Device Platforms by Molecular Layers: Hybrid Sensing Platforms, Monolayer
Doping, and Modeling. Langmuir 2018, 34, 14103-14123.
!31
Scullion, L.; Doneux, T.; Bouffier, L.; Fernig, D. G.; Higgins, S. J.;
7.
Bethell, D.; Nichols, R. J. Large Conductance Changes in Peptide Single
Molecule Junctions Controlled by Ph. J. Phys. Chem. C 2011, 115, 8361-8368.
Xiao, X.; Xu, B.; Tao, N. Conductance Titration of Single Peptide
8.
Molecules. J Am Chem Soc 2004, 126, 5370-5371.
Uji, H.; Morita, T.; Kimura, S. Molecular Direction Dependence of Single-
9.
Molecule Conductance of a Helical Peptide in Molecular Junction. Phys. Chem.
Chem. Phys. 2013, 15, 757-760.
10. Kitagawa, K.; Morita, T.; Kimura, S. Electron Transfer in Metal-Molecule-
Metal Junction Composed of Self-Assembled Monolayers of Helical Peptides
Carrying Redox-Active Ferrocene Units. Langmuir 2005, 21, 10624-10631.
11. Kitagawa, K.; Morita, T.; Kimura, S. Molecular Rectification of a Helical
Peptide with a Redox Group in the Metal-Molecule-Metal Junction. The journal
of physical chemistry B 2005, 109, 13906-13911.
12. Devillers, C. H.; Boturyn, D.; Bucher, C.; Dumy, P.; Labbé, P.; Moutet, J.-
C.; Royal, G.; Saint-Aman, E. Redox-Active Biomolecular Architectures and
Self-Assembled Monolayers Based on a Cyclodecapeptide Regioselectively
Addressable Functional Template. Langmuir 2006, 22, 8134-8143.
!32
13. Mondal, P. C.; Fontanesi, C.; Waldeck, D. H.; Naaman, R. Spin-Dependent
Transport through Chiral Molecules Studied by Spin-Dependent Electrochemistry.
Acc Chem Res 2016, 49, 2560-2568.
14. Mervinetsky, E.; Alshanski, I.; Hamo, Y.; Sandonas, L. M.; Dianat, A.;
Buchwald, J. x. F. r.; Gutiérrez, R.; Cuniberti, G.; Hurevich, M.; Yitzchaik, S.
Copper Induced Conformational Changes of Tripeptide Monolayer Based
Impedimetric Biosensor. Sci. Rep. 2017, 7, 9498.
15. Weiss, E.; Chiechi, R.; Kaufman, G.; Kriebel, J.; Li, Z.; Duati, M.; Rampi,
M.; Whitesides, G. Influence of Defects on the Electrical Characteristics of
Mercury-Drop Junctions: Self-Assembled Monolayers of N-Alkanethiolates on
Rough and Smooth Silver. J. Am. Chem. Soc 2007, 129, 4336-4349.
16. Ulman, A. An Introduction to Ultrathin Organic Films : From Langmuir-
Blodgett to Self-Assembly; Academic press: Boston, 1991.
17. Shirley, D. A. High-Resolution X-Ray Photoemission Spectrum of the
Valence Bands of Gold. Phys. Rev. B 1972, 5, 4709-4714.
18. Engelkes, V. B.; Daniel Frisbie, C. Simultaneous Nanoindentation and
Electron Tunneling through Alkanethiol Self-Assembled Monolayers. The Journal
of Physical Chemistry B 2006, 110, 10011-10020.
!33
19. Hutter, J.; Iannuzzi, M.; Schiffmann, F.; VandeVondele, J. Cp2k: Atomistic
Simulations of Condensed Matter Systems. Wiley Interdisciplinary Reviews:
Computational Molecular Science 2014, 4, 15-25.
20. Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient
Approximation Made Simple. Physical Review Letters 1996, 77, 3865-3868.
21. Goedecker, S.; Teter, M.; Hutter, J. Separable Dual-Space Gaussian
Pseudopotentials. Physical Review B 1996, 54, 1703-1710.
22. Grimme, S. Semiempirical Gga-Type Density Functional Constructed with
a Long-Range Dispersion Correction. Journal of Computational Chemistry 2006,
27, 1787-1799.
23. Rusu, P. C.; Brocks, G. Surface Dipoles and Work Functions of
Alkylthiolates and Fluorinated Alkylthiolates on Au(111). J Phys Chem B 2006,
110, 22628-22634.
24. Sessler, G. M. Electric Fields and Forces Due to Charged Dielectrics. J.
Appl. Phys. 1972, 43, 405-408.
25. Knapp, A. G. Surface Potenatials and Their Meausrement by the Diode
Method. Surface Science 1973, 34, 289-316.
!34
26. Peor, N.; Sfez, R.; Yitzchaik, S. Variable Density Effect of Self-Assembled
Polarizable Monolayers on the Electronic Properties of Silicon. J Am Chem Soc
2008, 130, 4158-4165.
27. Sfez, R.; Peor, N.; Yitzchaik, S. Experimental Evidence of Molecular
Cooperative Effect in a Mixed Parallel and Antiparallel Dipole Monolayer ? J.
Phys. Chem. C 2010, 114, 20531-20538.
28. Krzeminski, C.; Delerue, C.; Allan, G.; Vuillaume, D.; Metzger, R. Theory
of Electrical Rectification in a Molecular Monolayer. Physical Review B 2001, 64,
085405.
29. Tuccitto, N.; Ferri, V.; Cavazzini, M.; Quici, S.; Zhavnerko, G.;
Licciardello, A.; Rampi, M. A. Highly Conductive Approximately 40-Nm-Long
Molecular Wires Assembled by Stepwise Incorporation of Metal Centres. Nat
Mater 2009, 8, 41-46.
!35
TOC graphics
!36
!
Electron transport through tripeptides
self-assembled monolayers
Evgeniy Mervinetsky,1,2 Israel Alshanski,1,2 Stephane Lenfant,3 David Guerin,3 Leonardo
Medrano Sandonas,4,6 Arezoo Dianat,4 Rafael Gutierrez,4
Gianaurelio Cuniberti,4,5,6 Mattan Hurevich,1,2
Shlomo Yitzchaik1,2, & Dominique Vuillaume3
1) Institute of Chemistry, The Hebrew University of Jerusalem, Safra campus, Givat Ram,
2) Center for Nanoscience and Nanotechnology, The Hebrew University of Jerusalem, Safra
3) Institute for Electronics Microelectronics and Nanotechnology, CNRS, Univ. Lille, 59652,
Jerusalem, 91904, Israel.
campus, Jerusalem, 91904, Israel.
Villeneuve d'Ascq, France.
Dresden, Germany.
4) Institute for Materials Science and Max Bergmann Center of Biomaterials, TU Dresden, 01069
5) Dresden Center for Computational Materials Science, TU Dresden, 01062 Dresden, Germany.
6) Center for Advancing Electronics Dresden, TU Dresden, 01062 Dresden, Germany
Supporting information
Figure S1. Schematic description of the sample preparation and measurement protocols.
"
Figure S2. The S2p signal by XPS of HDP SAM (blue) before and (red) after Cu2+
exposure exhibited only sulfur atoms bound to gold (S2p3/2 at 161.9 eV and S2p1/2 at 163.1
eV) indicating that molecules are fully chemisorbed on gold.
Figure S3. The Cu2p signal (Cu2p1/2 at 952.4 eV and Cu2p3/2 at 932.6 eV), shifted compare
to free Cu2+ (933.6 eV), was only observed in the SAM exposed to Cu2+.
!
!
HDP
HDP + Cu2+
BE (eV)
Acorrected
BE(eV)
Acorrected
4023
1451
1125
319
-
C1s
N1s
O1s
284.6, 286.4, 288.3
400.2
531.7
S2p3/2-1/2
161.9, 163.1
Cu2p3/2-1/2
-
N/S = 4.5 (3)
N/O = 1.3 (1.5)
O/S = 3.5 (2)
C/N = 2.7 (3)
C/O = 3.6 (4.5)
284.6, 286.4, 288.3
400.2
531.6
161.9, 163.1
932.6, 952.4
N/S = 4.6 (3)
N/O = 1.0 (1.5)
O/S = 4.3 (2)
C/N = 2.9 (3)
C/O = 3.0 (4.5)
3665
1276
1209
277
43
Table S1. Binding energy and concentration of elements by XPS in HDP SAM before
and after Cu2+ exposure. Corrected area corresponds to peak area divided by the atomic
sensitivity factor. Atomic ratios measured by XPS are compared with theoretical values in
brackets.
Figure S4. (a) CPD measured by KPFM on HDP, HDP + Cu2+ protocol A, and HDP +
Cu2+ protocol B. All values are averaged from 4 KPFM images. We deduce an increase of
WF by 52±37 meV (protocol A) and 91±20 meV (protocol B). The average WF increase,
reported in Fig. 3 (main text) is 71.5 mV (b) CPD measured by KPFM on LDP, LDP +
Cu2+ by protocol A and LDP + Cu2+ protocol B. All values are averaged from 3 KPFM
images. We deduce an increase of WF by 29±14 meV (protocol A) and 28±16 meV
(protocol B)
!
Figure S5. 2D histograms of the current-voltage curves (300 I-V traces, tip loading force
50 nN) and corresponding current histograms at 1V for HDP SAM exposed to Cu2+
(protocol B). The current histograms at 1 V are fitted (black lines) with log-normal
distributions (fitted parameters in Table 1). In the I-V histograms, the white line is the
sensitivity limit of the C-AFM system (~0.1 pA).
Figure S6. 2D histograms of the current-voltage curves (300 I-V traces, tip loading force
29 nN) and corresponding current histograms at 2V for LDP SAM exposed to Cu2+
(protocol B). The current histograms at 1 V are fitted (black lines) with log-normal
distributions (fitted parameters in Table 1). In the I-V histograms, the white lines is the
sensitivity limit of the C-AFM system (~0.1 pA)
Figure S7. Orbital resolved projected density of states for Lpa-GGH/Au(111) before
(upper panel) and after (lower panel) Cu2+ ion binding for the C-IV complex (see also the
main text). The inset in the lower panel shows the spatial charge density distribution of
the state closest to the Fermi energy, which is derived from the hybridization of Cu d- and
N p-states with some contribution from O p-states, too.
!
Figure S8. Projected density of states for the four possible configurations studied in the
main text, in which Cu2+ can bind to Lpa-GGH/Au(111). Panels a-d correspond to the
conformations shown in Fig. 7 of the main text (a) C-I, (b) C-II, (c) C-III and (d) C-IV.
We show the DOS projected on the Lpa-GGH+Cu2+ complex as well as on the Cu atom
and all the N atoms of the molecule. With increasing coordination of the Cu to the N
atoms in the ring region, the spectral contribution at the Fermi energy is continuously
reduced from a to d. Notice that in the latter case almost no Cu and N weight is observed
at the Fermi energy.
!
Figure S9. Orbital resolved projected density of states (PDOS) of the possible
configurations a-d (see Fig. 7 in the main text) for the study of Cu2+ ion location in the
Lpa-GGH/Au(111) system. Shown are the contributions from Cu s-, p- and d-states, and
from the p-states by O, N, and S.
!
|
1903.10312 | 1 | 1903 | 2019-02-03T13:43:15 | Ferroelectric multiple-valued logic units | [
"physics.app-ph"
] | Employing many-valued logic (MVL) data processing allows to dramatically increase the performance of computing circuits. Here we propose to employ ferroelectrics for the material implementation of MVL units basing on their ability to pin the polarization as a sequence of multi-stable states. Two conceptual ideas are considered. As the first system, we suggest using the strained ferroelectric films that can host the polarization states, allowing the effective field-induced multilevel switching between them. As the second one, we propose to employ the ferroelectric nano-samples that confine the topologically protected polarization textures, which may be used as MVL structural elements. We demonstrate that these systems are suitable for engineering of the 3-, 4- and even 5-level logic units and consider the circuit design for such elements. | physics.app-ph | physics | Ferroelectric Multiple-Valued Logic Units
I. Lukyanchuk,1,2,* E. Zaitseva,3 V. Levashenko,3 M. Kvassay,3
S. Kondovych,1 Yu. Tikhonov,4 L. Baudry,5 and A. Razumnaya 4
1 Laboratoire de Physique de la Matière Condensée, Université de Picardie Jules Verne, 80039
Amiens, France
2 Landau Institute for Theoretical Physics, Moscow, Russia
3 Faculty of Informatics and Management Science, University of Zilina, Zilina 01026, Slovakia
4 Faculty of Physics, Southern Federal University, Rostov-on-Don 344090, Russia
5 Institute of Electronics, Microelectronics and Nanotechnology (IEMN)-DHS Départment, UMR
CNRS 8520, Université des Sciences et Technologies de Lille, 59652 Villeneuve d'Ascq Cedex,
France
Employing many-valued logic (MVL) data processing allows to dramatically increase the
performance of computing circuits. Here we propose to employ ferroelectrics for the
material implementation of MVL units basing on their ability to pin the polarization as a
sequence of multi-stable states. Two conceptual ideas are considered. As the first system,
we suggest using the strained ferroelectric films that can host the polarization states,
allowing the effective field-induced multilevel switching between them. As the second one,
we propose to employ the ferroelectric nano-samples that confine the topologically-
protected polarization textures which may be used as MVL structural elements. We
demonstrate that these systems are suitable for engineering of the 3-, 4- and even 5-level
logic units and consider the circuit design for such elements.
KEYWORDS: Ferroelectrics, switching, hysteresis, topological structures, multiple-valued
logic, vortices, domains, skyrmions.
1
1. Introduction
Permanently growing size and amount of processed information and increasing of the
computation speed impose the emergence of new technologies for efficient data treatment. A non-
binary architecture of the computing circuits is considered as an alternative to the conventional binary
2-valued logic. Binary elements are based on the principle that a data bit is always either "zero" or
"one". However, already in the 1970s, the researchers pointed out the limitations of this technology.
The first one is the interconnect problem, related to both on-chip and between-chip connections. The
difficulties of placement and routing of the digital logic elements are escalating with an increase of
the computation capability per chip. In fact, the area of interconnections may consist of more than
70% of active logic elements [1]. However, the further increase of the on- and off-chip connections
faces the mechanical, thermal, and electrical restrictions [2,3].
Another limiting factor is the necessity of increasing the clock speed of switching between
different binary states [4] that finally determines the computer performance. During past decades, the
clock speed had doubled almost every year. Usually, the limitation of the clock speed is bypassed by
packing some cores into a chip, which has resulted in multi-core processors. However, this approach
does not greatly improve performance because of the limiting amount of the binary data that need to
be transferred.
These factors point on the necessity to design the principally new computing circuits using the
Multiple-Valued Logic (MVL) architecture. The primary advantage of MVL is the ability to encode
more information per variable ("multi-valued bit") [1,5]. However, in this case, one should go beyond
the conventional material technology of semiconducting transistor, having only two stable states,
"on" or "off". Hence, the development of new MVL computing circuits is based on two inter-related
advances, the technological one that includes the development of new non-silicon multi-valued logic
gates and the mathematical one that embraces the design of new computation methods and algorithms.
In this work, we suggest to use the ferroelectric materials as a platform for implementation of the
MVL elements. The idea is based on the capacity of ferroelectrics to hosts multiple polarization states
having nearly the same energy. Forming the logical levels of the MVL unit, these states can be
switched by the electric field. The most evident way that we consider in Section 2 is to use the
degeneracy of polarization orientation in the pseudo-cubic crystal lattice of perovskite oxides. As a
matter of fact, the strained films of pseudo-cubic ferroelectric oxides can host a variety of logically
different multilevel hysteresis loops, holding two, three, or even four polarization states [6,7].
Importantly, these logical gates are operational at room temperature, and their logic is tunable by
strain and temperature which is important for prospective future "in silico" applications.
Even more fascinating opportunity can be provided by a variety of topological structures of
polarization, confined in the nanoscale ferroelectric samples: nanodots, and nanopillars (nanorods).
In Section 3 we consider the switching between different types of these structures: domain patterns
[8,9], singular vortices [10-12], coreless vortices and skyrmions [13-16] that allows to realize even
more lively logical states for MVL elements, opening this the unprecedented horizons for the domain-
and topological structures -- provided nanoelectronics [17].
The design of the logic operations and MVL circuits will be considered in Section 4. Our findings
enable developing a platform for the emergent MVL technology of information processing and target
the further challenges posed by needs of quantum and neuromorphic computing.
2
2. Multilevel hysteresis switching in ferroelectrics films
In [7] we have demonstrated that MVL cells can be realized using the substrate-deposited thin films
of ferroelectric perovskite oxides. The model ferroelectric material, PbTiO3, having the pseudo-cubic
structure is viewed as a promising material since the technology of thin-film deposition of PbTiO3 is
fairly well controlled and it can operate at room temperature. Importantly, the polarization states of
PbTiO3 films crucially depend on the strain, imposed by the substrate that can have both the tensile and
compressive character. Accordingly, the polarization can have either in- or out-of-plane orientations with
respect to the plane of the film. The shown in Fig.1a strain-temperature um-T phase diagram of the
substrate-deposited single-domain PbTiO3 films was demonstrated [18] to host three ferroelectric phases:
the c-phase with the out-of-plane polarization orientation, the aa-phase with the in-plane polarization
orientation along the face diagonal of the pseudo-cubic lattice cell and the r-phase in which polarization
is tilted with respect to the film plane and lies in the vertical diagonal sectional plane, formed by the c-
and aa- directions.
The key point here is that staying in one of the thermodynamically stable states, the system can host
other metastable states that are the legacy of the phases, stable in other parts of the um-T phase diagram.
Application of the external electric field permits to switch between these states, realizing hence the
complicate hysteresis loop with various branches. Depending on the landscape of the energy profile and
on the protocol of the field variation one can get the loop of a rather complex structure. Some examples
of the hysteretic transitions between c- aa- and r-state having two, three, and four branches are shown in
Fig. 1b. (Fig. 1a shows the location of these loops at the um-T phase diagram.) Panels of Fig. 1b display
also the switching maps of the corresponding MVL elements.
Figure 1. Multilevel hysteresis switching in films of PbTiO3 (a) Strain-temperature phase diagram of strained
PbTiO3 film [18]. Bubble notations correspond to the location of the hysteresis loops, shown on the panel (b). (b)
Example of hysteresis loops inPbTiO3 film [7]. Switching topology depends on the location of the loop in panel (a).
States "±1" and "±2" correspond to the r- and c-phases with "up"/"down"-oriented polarization respectively, state
"0" correspond to the aa-phase. (C) Usual 2-level binary systems, realizing in c-phase. (IX) The 3-level loop in
aa- phase where each state can be reached from all others (IV,V) Examples of 4-level loops in r-phase. Insets
show the logical switching map of corresponding MVL elements.
3
Panel C of Fig, 1b shows the 2-branch hysteresis loop, realizing the conventional 2-bit logic operation.
This standard situation is realized in the c-phase. The aa-phase in the vicinity of the first-order transition
from c-phase hosts a stable aa-state and also two metastable c-states having the "up" and "down" out-
of-plane-oriented polarization. The corresponding 3-branch hysteresis loop (panel IX) allows for the
stack-wise access to all the logical levels, realizing hence the 3-level MVL unit. We consider the
operational modes for such element in Section 4 in more detail. Finally, the r-phase has two "up" and
"down" out-of-plane-oriented polarization components and also two metastable states corresponding to
the "up"- and "down"-oriented polarization states of the c-phase. The corresponding hysteresis with 4
branches can have far more complex logical structure. Two examples of the 4-branch loops and
corresponding switching maps of 4-level MVLs are shown at panels IV and V. A more exhaustive
description of all the possibilities is given in Ref. [7].
3. Switching of confined topological structures
Confinement that imposes additional symmetries on the system can stabilize exotic topological
states bringing novel functionalities, which do not exist in bulk materials [19,20]. Topological
structures in nano-size ferroelectric samples [21], nanodots, and nanopillars (nanorods) (Fig. 2a) are
of special interest for applications because they can be relatively easily controlled and manipulated
by electric fields. Ferroelectric topological excitations can be reduced in size to atomic scales, in
addition to being tunable through lattice strain. The depolarizing charge associated with the
topological excitation permits coupling with the incident electromagnetic field and mutual
electrostatic cross-talk.
Importantly, confined topological structures can be manipulated and switched by electric fields
with a critical threshold field Ec < 10 mV/nm and characteristic read/write times < 50 ps. Variety of
different topologically-protected polarization textures with nearly the same energy arising in the same
sample permits the field-induced switching between them, realizing hence the multivalued hysteresis
loops, similar to those, described in the previous Section.
The topological class of the confined polarization texture depends on the degeneracy and
anisotropy of configuration space of the order parameter, which is the vector of polarization. Figure
2 summarizes the state-of-the-art of the study of topological excitations in ferroelectrics and the field-
induced switching between them. They are conventionally classified according to the uniaxial
anisotropy of the system, imposed by strains. For simplicity, we do not consider the effect of the
higher-symmetry cubic anisotropy, appropriate for the perovskite oxide ferroelectrics.
(i) Uniaxial easy-axis anisotropy, according to magnetic terminology (Fig. 2b). Polarization has
only two, equilibrium orientations "up" and "down" with two-point degeneracy Z2 in the
configuration space. The Kittel domain structure is formed to minimize the depolarization
energy, similar to the ferroelectric thin films and superlattices [22-24]. However, in this case, the
confined domain configuration has more elaborated geometry. A wealth of switching paths
between them is possible [9]. We show, as an example, the hysteresis loop with 5 branches that
realizes the 5-level MVL cell.
(ii) Uniaxial easy-plane anisotropy (Fig. 2c). The in-plane orientation of polarization has the circular
degeneracy S2 in configuration space. Calculations, based on the Landau-Ginzburg formalism
coupled with the electrostatic equations show that a variety of the vortex-like solutions can have
4
approximately the same minimal energy [11]. The residual degeneracy of the system is
characterized by the clock- and counter-clockwise polarisation rotation. The vortices are formed
to vanish the depolarization energy that would be huge in case of the uniform polarization. The
field, applied perpendicular to the vortex plane results in the 3-branch hysteresis loop that
switches between the vortex-states and the "up"- and "down"-polarized monodomain states.
Hence, the 3-level MVL cell is realized.
(iii) Almost-isotropic system with spherical S3 degeneracy of the order parameter (Fig. 2d). The
exotic skyrmion topological excitation having the structure of vortices with the non-singular core
can be confined in cylindrical nanodot [13]. Besides orientation of polarization "up" and "down"
inside of the central core region the skyrmions are characterized by the left-hand and right-hand
chirality, hence are optically active. Switching of the skyrmions by electric field [14] results in
the 4-branch hysteresis loop with the "up"- and "down"-oriented skyrmion states and with "up"-
and "down"-oriented uniformly polarized states. This loop corresponds to the 4-level MVL cell.
Many other confined topological structures in ferroelectrics, including the recently reported
polarization textures in spherical nanodots [25] and two-domain structure in ferroelectric nano-
capacitors with negative capacitance [26] can also be thought as a platform for MVL cells.
(a)
(b)
(c)
(d)
Figure 2. Multilevel switching between polarization topological structures in nano-samples. (a)
Sketch of the PbTiO3 nanopillars embedded in SrTiO3 matrix and of PbTiO3 nanodot deposited on SrTiO3-
substrate. (b) In case of the uniaxial easy-axis anisotropy, the "up"- and "down"-oriented polarization
domains are formed [9] Top view of polarization distribution in domains and example of 5-level switching
loop between them are shown. Red and blue colors denote the "up"- and "down"- oriented polarization. (c)
In the case of uniaxial easy-axis anisotropy, the in-plane polarization vortices are formed [11]. Multiple
vortex states are shown. The blue color corresponds to the singular vortex cores where polarization
vanishes. The 3-level switching occurs between "up"- and "down"-oriented polarized monodomain states
and the in-plane vortex state. (d). In case of weak anisotropy, the coreless vortices - skyrmions with "up"-
and "down"-oriented cores can be formed in cylindrical nanodot [14]. The switching hysteresis in this case
have 4 branches.
5
3D, quasi-isotropic Skyrmions4-level logicP 'up'Skyrmion 'up'P 'down'Skyrmion 'down' 1D, easy axis polarization Domain Walls5-level logicPE 2D, easy plane polarization in-plane Vortices P 'up'P 'down'Vortex 3-level logicSwitchable nanodot PbTiO3PtSrTiO3PPbTiO3SrTiO3Nanopillars 4. Multi-valued logic circuit design
There are different techniques for logic circuits design based on MVL elements [27-29]. One of these
techniques is Programmable Logic Arrays (PLA) development that permits to implement any logic
function (combinational logic circuits) [28]. PLA is a kind of programmable logic device that has two
parts (Fig. 3): memory (M) and a logic block (L). The ferroelectrics elements can be used for the memory
block implementation according to the technological aspect.
x1 x2 … xn
M L
a(0)
a(1)
…
n
a(3
-1)
f(x)
Figure 3. Structure of the proposed PLA for implementation of MVL function.
The regular structure of PLA causes the use of canonical and regular mathematical representation of
MVL function implemented by this circuit. One of possible representations is generalised Reed-Muller
expansion (GRME). The ferroelectrics technologies considered above allows implementing of 3-level
switching elements. The mathematical interpretation of this elements in MVL is possible by the 3-valued
function f(x) = f(x1, x2, . . . , xn) of n variables x1, x2, . . . , xn. This function is defined for set m = {0, 1, 2}
as mapping {0, 1, 2}n {0, 1, 2}. The GRME of 3-valued function f(x) on n variables is determined by
the following equation [30-31]:
3𝑛−1
𝐴(𝒙) = ∑ 𝑎(𝑖)𝑥1
𝑖1𝑥2
𝑖2 … 𝑥𝑛
𝑖𝑛 (𝑚𝑜𝑑 3)
(1)
were a(i) are the coefficients, a(i) {0, 1, 2}; ij is the j-th digit of 3-valued representation of parameter
i, i = (i1,i2,...,in)3 and j = 1, …, n. The GRME of 3-function on one variable according to (1) is:
𝑖=0
2
𝐴(𝑥) = ∑ 𝑎(𝑖)𝑥𝑖 =
𝑎(0) + 𝑎(1)𝑥 + 𝑎(2)𝑥2 = 𝑎(0) + (𝑎(1) + 𝑎(2)𝑥)𝑥 (𝑚𝑜𝑑 3).
(2)
𝑖=1
The PLA for a 3-valued function based on GRME representation has 3n inputs. These inputs are the
memory's inputs to program the GRME coefficients values. The change of these inputs values (re-
programming of GRME coefficient) results in new 3-valued function (new circuits based on PLA). The
coefficients are read and transmitted to the logic block. The logic block has a set of n external inputs for
variables and 3n inputs from the memory block. The logic block has a number of sum-product's
homogeneous sub-blocks which are linked together to give output.
Let us consider this PLA for 3-valued functions of tree variable (n = 3). The detailed structure of the
PLA for 3-valued function on 3 variables is shown in Fig. 4. This PLA consists of memory and logic
blocks and has 3 inputs for function's variables x1, x2 and x3 and 27 inputs to programme the coefficients
of GRME. The memory block includes 27 memory cells that can be implemented based on ferroelectric
technology. The logic block consists of 3 levels. The first level includes 9 homogeneous sub-blocks that
are connected to 3 sub-blocks of the second level. The third level consists of one sub-block. All sub-
6
block are homogeneous and each of them has 3 inputs from the previous level, 1 input for variable and
1 output. The logical structure of the sub-block is caused by the GRME for one variable (2). The sub-
block is the parallel structure with 3 modulo adders and 2 modulo multipliers that implement the
calculation of (2). The modulo adder and multiplier can be elaborated based on different technologies,
for example, these devices based on CMOS technology is considered in [32].
The logic block for the 3-valued function of n variable is homogeneous recurrent and parallel. It
consists of n level. The s-th level (s = 1, …, n) is connected to the (n-s+1)-th variable and includes (3n-s)
homogeneous sub-blocks each of them calculates GRME on one variable (2). The outputs of blocks of
the s-th level are inputs of blocks of the (s+1)-th level. The first level is connected to the memory of PLA
to read the coefficients.
Figure 4. The detailed structure of the PLA for implementation of 3-valued function on 3 variables
7
x3 x2 x1 a(0) a(1) a(2) a(3) a(4) a(5) a(6) a(7) a(8) a(9) a(10) a(11) a(12) a(13) a(14) a(15) a(16) a(17) a(18) a(19) a(20) a(21) a(22) a(23) a(24) a(25) a(26) M L
Acknowledgments
This work was supported by the French-Slovak bilateral collaborative program PHC-STEFANIK, by the
Slovak Research and Development Agency under the contract No. SK-FR-2017-0003 and by the H2020-RISE-
ENGIMA action.
References
1. S. L. Hurst, Multiple-Valued Logic. Its Status and its Future. IEEE Trans on Computers 33 (12), 1160 --
1179 (1984).
2. P. Beckett, Towards a Reconfigurable Nanocomputer Platform, 7th ACSAC Proc., 141-150 (2002). (P.
Beckett, Towards a Reconfigurable Nanocomputer Platform, Australian Computer Science
Communications 24, 141-150 (2002).
3. C. Vudadha, and M.B. Srinivas, Design Methodologies for Ternary Logic Circuits, 48th IEEE Int. Symp.
on Multiple-Valued Logic Proc., 192 -- 197 (2018).
4. B. Choi, and K. Shukla, Multi-Valued Logic Circuit Design and Implementation, Int. Journal of
Electronics and Electrical Engineering 3 (4), 256 -- 262 (2015).
5. A. P. Surhonne, D. Bhattacharjee, and A. Chattopadhyay, Synthesis of Multi-Valued Literal using
Łukasiewicz logic, 48th IEEE Int. Symp. on Multiple-Valued Logic Proc., 204 -- 209 (2018).
6. L. Baudry, I. A. Luk'yanchuk, and A. Razumnaya, Dynamics of Field Induced Polarization Reversal in
Strained Perovskite Ferroelectric Films with c-oriented Polarization, Phys. Rev. B 91, 144110 (2015).
7. L. Baudry, I. Lukyanchuk, and V. M. Vinokur, Ferroelectric symmetry-protected multibit memory cell,
Sci. Rep. 7 1 -- 7 (2017).
8. I. Lukyanchuk, P. Sharma, T. Nakajima, S. Okamura, J.F. Scott, and A. Gruverman, High-Symmetry
Polarization Domains in Low-Symmetry Ferroelectrics, Nano Lett. 14, 6931 (2014).
9. P.-W. Martelli, S. M. Mefire, and I. Luk'yanchuk, Multidomain switching in the ferroelectric nanodots,
Europhys. Lett. 111, 50001 (2015).
10. I. I. Naumov, L. Bellaiche and H. Fu, Unusual phase transitions in ferroelectric nanodisks and nanorods,
Nature 432, 737 (2004)
11. L. Lahoche, I. Luk'yanchuk, and G. Pascoli, Stability of vortex phases in ferroelectric easy-planes nano-
cylinders, Integrated Ferroelectrics 99, 60 (2008).
12. A. K. Yadav, C. T. Nelson, S. L. Hsu, Z. Hong, J. D. Clarkson, C. M. Schlepütz, A. R. Damodaran, P.
Shafer, E. Arenholz, L. R. Dedon, D. Chen, A. Vishwanath, A. M. Minor, L. Q. Chen, J. F. Scott, L. W.
Martin, and R. Ramesh, Nature 530, 198 (2016).
13. L. Baudry, A. Sene, I. Luk'yanchuk, and L. Lahoche, Vortex state in thin films of multicomponent
ferroelectrics, Thin Solid Films 519, 5808 (2011).
14. L. Baudry, A. Sené, I.A. Luk'yanchuk, L. Lahoche, and J.F. Scott, Polarization vortex domains induced
by switching electric field in ferroelectric films with circular electrodes, Phys. Rev. B 90, 024102 (2014).
15. Y. Nahas, S. Prokhorenko, L. Louis, Z. Gui, I. Kornev, and L. Bellaiche, Discovery of stable skyrmionic
state in ferroelectric nanocomposites, Nat. Comm. 6, 1 (2015).
16. M.A.P. Gonçalves, C. Escorihuela-Sayalero, P. García-Fernández, J. Junquera and Jorge Íñiguez,
Theoretical guidelines to create and tune electric skyrmions, Preprint arXiv:1806.01617 (2018).
17. G. Catalan, J. Seidel, R. Ramesh, and J. F. Scott, Domain wall nanoelectronics, Rev. Mod. Phys. 84, 119
(2012).
18. N. A. Pertsev, A. G. Zembilgotov, and A. K. Tagantsev, Effect of Mechanical Boundary Conditions on
Phase Diagrams of Epitaxial Ferroelectric Thin Films, Phys. Rev. Lett., 80, 1988 (1998).
19. N. D. Mermin, The topological theory of defects in ordered media, Rev. Mod. Phys. 51, 591 (1979).
8
20. V. P. Mineev, Topologically stable defects and solitons in ordered media, in: I.M. Khalatnikov (Ed.),
Soviet Scientic Reviews, Section A: Physics Reviews, Vol. 2, Harwood, London, 1980, pp. 173 -- 246.
21. J. Seidel, Topological structures in ferroic materials: domain walls, vortices and skyrmions (Springer,
2016).
22. A. M. Bratkovsky, and A. P. Levanyuk, Abrupt Appearance of the Domain Pattern and Fatigue of Thin
Ferroelectric Films, Phys. Rev. Lett. 84, 3177 (2000).
23. V. A. Stephanovich, I. A. Luk'yanchuk and M. G. Karkut, Domain-enhanced interlayer coupling in
ferroelectric/paraelectric superlattices, Phys. Rev. Lett. 94, 047601 (2005).
24. I. Lukyanchuk, A. Sené, and V. M. Vinokur, Electrodynamics of ferroelectric films with negative
capacitance, Phys. Rev. B 98, 024107 (2018)
25. J. Mangeri, Y. Espinal, A. Jokisaari, S.A. Pamir, S. Nakhmanson, and O. Heinonen, Topological phase
transformations and intrinsic size effects in ferroelectric nanoparticles, Nanoscale 9, 1616 (2017).
26. I. Luk'yanchuk, Y. Tikhonov, A. Sene, A. Razumnaya and V. M. Vinokur, Harnessing ferroelectric
domains for negative capacitance, Preprint arXiv:1811.12103 (2018).
27. A.I. Reis, and R. Drechsler, Advanced Logic Synthesis (Springer, 2018).
28. H. M. Munirul, and M. Kameyama, Ultra-Fine-Grain Field-Programmable VLSI Using Multiple-Valued
Source-Coupled Logic, 34th Int. Symp. Multiple-Valued Logic Proc., 26-30 (2004).
29. Z. Zilic, and Z.G. Vranesic, Multiple-valued logic in FPGAs, 36th Midwest Symposium on Circuits and
Systems Proc., 2, 1553 - 1556 (1993).
30. D. H. Green, Ternary Reed-Muller Switching Functions with Fixed and Mixed Polarity, Int. J. Electronics
67, 761-775 (1989).
31. G.A. Kukharev, V.P. Shmerko, and E.N. Zaitseva, Multiple-Valued Data Processing Algorithms and
Systolic Processors (Nauka i Technika, Minsk, Belarus, 1990).
32. K. Tarun, and M. S. Hashmi, Multiple valued current mode logic circuits, Int. Conf. Multimedia, Signal
Processing and Communication Technologies Proc., 65 -- 69 (2017).
9
|
1907.11127 | 1 | 1907 | 2019-07-25T15:12:11 | Polarization-dependent and Valley-protected Lamb Waves in Asymmetric Pillared Phononic Crystals | [
"physics.app-ph"
] | We present the realization of the topological valley-protected zero-order antisymmetric (A0) or symmetric (S0) and zero-order shear-horizontal (SH0) Lamb waves at different domain walls based on topologically distinct asymmetric double-sided pillared phononic crystals. The elastic periodic structures have either the triangular or the honeycomb symmetry and give rise to a double-negative branch in the dispersion curves. By artificially folding the doubly negative branch, a degenerate Dirac cone is achieved. Different polarization-dependent propagation along the same primary direction along the constituent branches are presented. Moreover, divergent polarization-dependent phenomena along different primary directions along a given branch are also reported. By imposing two large space-inversion symmetry (SIS) breaking perturbations the topological phase transition is obtained. We show that the Berry curvature becomes strongly anisotropic when the wave vector gets away from the valleys. Further, we demonstrate the unidirectional transport of A0, S0, and SH0 Lamb waves at different domain walls in straight or Z-shape wave guides. In the large SIS breaking case, we show negligible reflection at the zigzag outlet of the straight wave guide and occurrence of weak inter-valley scattering at the bending corners of the Z-shape wave guide. For a larger strength of SIS breaking, the edge states are gapped and strong reflection at the zigzag outlet and bending corners is observed. The topological protection cannot be guaranteed any more in that case. | physics.app-ph | physics | Polarization-dependent and Valley-protected Lamb Waves in
Asymmetric Pillared Phononic Crystals
Wei Wang1, Bernard Bonello1*, Bahram Djafari-Rouhani2, and Yan Pennec2
1Sorbonne Université, UPMC Université Paris 06 (INSP -- UMR CNRS 7588),
4, place Jussieu 75005 Paris, France
2Institut d'Electronique, de Micro-électronique et de Nanotechnologie (IEMN -- UMR CNRS 8520),
Université de Lille Sciences et Technologies, Cité Scientifique, 59652 Villeneuve d'Ascq Cedex, France
*corresponding author: [email protected]
Abstract:
We present the realization of the topological valley-protected zero-order antisymmetric (A0) or symmetric
(S0) and zero-order shear-horizontal (SH0) Lamb waves at different domain walls based on topologically
distinct asymmetric double-sided pillared phononic crystals. The elastic periodic structures have either the
triangular or the honeycomb symmetry and give rise to a double-negative branch in the dispersion curves.
By artificially folding the doubly negative branch, a degenerate Dirac cone is achieved. Different
polarization-dependent propagation along the same primary direction along the constituent branches are
presented. Moreover, divergent polarization-dependent phenomena along different primary directions along
a given branch are also reported. By imposing two large space-inversion symmetry (SIS) breaking
perturbations the topological phase transition is obtained. We show that the Berry curvature becomes
strongly anisotropic when the wave vector gets away from the valleys. Further, we demonstrate the
unidirectional transport of A0, S0, and SH0 Lamb waves at different domain walls in straight or Z-shape
wave guides. In the large SIS breaking case, we show negligible reflection at the zigzag outlet of the straight
wave guide and occurrence of weak inter-valley scattering at the bending corners of the Z-shape wave guide.
For a larger strength of SIS breaking, the edge states are gapped and strong reflection at the zigzag outlet
and bending corners is observed. The topological protection cannot be guaranteed any more in that case.
I. INTRODUCTION
Topologically protected edge states provides an fascinating approach to manipulate the propagation of
waves [1 -- 10] since it allows for the one-way propagative edge states immune to defects and disorders at the
domain walls between two topologically distinct configurations. The foundation of such a compelling
phenomenon originates from the notion of topological phase which was first investigated in quantum
1
systems [11 -- 13]. Soon afterwards, its great potential for the unidirectional transport with negligible
attenuation has been exploited for other types of waves, including acoustic and elastic waves, and several
designs of phononic crystals have been proposed for this purpose [14 -- 16].
Three kinds of topological phononic crystals have been reported, each of which exhibiting transport
properties analogous to quantum Hall effect (QHE), quantum spin Hall effect (QSHE), and quantum valley
Hall effect (QVHE) respectively. To mimic QHE, active components must be involved in order to break the
time-reversal symmetry (TRS). In elastic systems this can be achieved with rotating gyroscopes and non-
inertial platforms [2,17,18] or with flow circulators in acoustic systems [1,19,20] but at the price of great
challenges in both the manufacturing and assembly. Emulating QSHE with TRS preserved only involves
passive components, whereas this generally requires intricate designs to achieve a double degenerate Dirac
cone [21 -- 24]. As regards QVHE, it exploits the valley degree of freedom, which has been proved to be
another controllable degree of freedom for an electron in a two-dimensional honeycomb lattice of graphene,
and was quickly introduced in acoustic and elastic phononic crystals. Comparatively to the other two effects,
it allows to significantly reduce the geometrical complexity and the topological phase transition can be
simply achieved by breaking the space-inversion symmetry (SIS) in the unit cell. As a result, a great deal of
configurations emulating QVHE have been proposed to theoretically and experimentally demonstrate the
unidirectional transport of elastic waves both in discrete models [18,25 -- 27] and in continuum structures
[28 -- 33]. Among these, the plate-type structures have generated much interests due to their advantages in
controlling Lamb waves that are widely used in engineering applications, including nondestructive
evaluation and structural health monitoring. For instance, the propagation of valley-protected antisymmetric
Lamb waves in phononic plates filled with geometrical indentations [31] or triangular inclusions [33] has
been reported. Another very promising platform is the pillared phononic crystal (PPnC) [14,34,35] built by
regularly erecting cylinders on a thin plate [36,37]. These attached pillars act then as resonators which
bending, compressional, and torsional modes can easily couple with the symmetric (S), the antisymmetric
(A), or the shear-horizontal (SH) Lamb waves propagating in the plate respectively.
In this work, in analogy to QVHE, we have carried out the valley-protected A0, S0, and SH0 Lamb waves at
different domain walls constructed upon topologically distinct asymmetric double-sided PPnCs. Firstly, we
describe a PPnC arranged in a triangular lattice, the Brillouin zone (BZ) of which features an isolated branch
where both the effective mass density and the elastic modulus are negative. By offsetting the unit cell and
rearranging the lattice in the honeycomb symmetry, the doubly negative branch is artificially folded, giving
rise to a degenerate Dirac cone and to different polarization-dependent propagation along the same primary
direction in the BZ, for each constituent branches. Moreover, we report on divergent polarization-dependent
phenomena along the different primary directions on the same branch. Secondly, the topological phase
2
transition is achieved by imposing a perturbation to the height of some pillars in the unit cell. Two SIS
breaking situations with different strengths are considered. The Berry curvature in the vicinity of the valleys
is numerically evaluated. This parameter becomes strongly anisotropic when the wave vector departs from
the valleys. Thirdly, we investigate the unidirectional transport of Lamb waves along different domain walls
either in a straight or in a Z-shaped wave guide. We show that the propagation of SH0 Lamb wave is
topologically protected at one domain wall where the propagation of A0 (S0) Lamb wave is forbidden
because of the mismatch in the spatial parity. The opposite phenomenon is observed at the other domain
wall. In the case of a moderate SIS breaking, the reflection at the zigzag outlet of the straight wave guide
can be neglected and weak inter-valley scattering occurs at the bending corners of the Z-shape wave guide.
For larger strength of SIS breaking, the edge states are gapped. A strong reflection occurs at the zigzag
outlet of a straight wave guide and at the bending corners of a Z-shaped wave guide, which signifies the
failure of the topological protection.
II. CONSTRUCTING A DEGENERATE DIRAC CONE AND ITS POLARIZATION-DEPEDENT
PROPAGATION
A. Doubly negative branch in triangular lattice
In this section, we describe the procedure we used to construct a degenerate Dirac cone, which is a
prerequisite for topological edge states to occur. It is based on the differentiated manner the eigenmodes of
the pillar couple with the Lamb modes. Actually, in the low frequency regime, both A0 and S0 Lamb waves
propagating in the plate can be altered by the bending and the compressional vibration of the pillars [38],
whereas SH0 mode, because of its in-plane polarization perpendicular to the propagation direction, can
easily couple with the torsional resonance of the pillars [39]. Moreover, if the resonators in an asymmetric
double-sided square lattice PPnC are thoughtfully designed, the bending, the compressional, and the
torsional modes fall within the same frequency interval [39] and an isolated doubly negative branch occurs.
The propagation along ΓX direction for a mode in this branch strongly depends on the polarization: it is
allowed for an incident SH0 Lamb wave, while it is not for A0 or S0 Lamb mode.
In analogy, we put forward an asymmetric double-sided PPnC arranged in a triangular lattice. The structure
and the unit cell are displayed in Fig. 1(a). It features distinct pillars denoted as pillar A and pillar B
concentrically connected to a thin plate. Both the plate and the pillars are made of steel whose Young's
modulus, Poisson's ratio and mass density are E = 200GPa, v = 0.3 and ρ = 7850kg/m3 respectively. The
lattice constant and the thickness of the plate are a = 231μm and e = 100μm. Pillar A is designed to have
3
both the bending and the compressional resonances within a common frequency interval where the effective
mass density turns negative and the propagation of A0 or S0 modes is forbidden. The diameter and height of
pillar A are dA = 120μm and hA = 268μm. On the other hand, the dimensions of pillar B (dB = 140μm and
hB = 160μm) are optimized for the torsional resonance to occur in a narrow frequency interval inside this
band gap. The propagation is allowed in this frequency interval, which shows that the doubly negative
property is achieved therein. This could be formally demonstrated by calculating the band structure. The
result, obtained by applying periodic conditions on the four lateral boundaries of the plate and solving the
eigenvalue equations, is displayed in Fig. 1(b). As expected, a band gap opens up in between 4.176 and
4.643MHz and an isolated branch with a negative-slope (i.e. the group velocity is different from zero) takes
place in between 4.408 and 4.486MHz. This branch is not present in the dispersion curves if one considers
the single-sided PPnC constructed solely by pillar A or pillar B (not shown here). Therefore, at any
frequency in between 4.408 and 4.486MHz, the system simultaneously exhibits negative effective mass
density and shear modulus, whereas only the effective mass density is negative outside this band. The
eigenmodes at points labelled C, D and E in Fig. 1(b) are displayed in Fig. 1(c). Clearly, the eigenmodes at
points D and E are respectively the second-order bending resonance (relating to the components 11 and 22
of the effective mass density matrix) and the first-order compressional resonance (relating to 33) of pillar
A [40]. The eigenmode at point C is the first-order torsional resonance of pillar B and relates to the negative
effective shear modulus. Actually, the weighting of the deformation around z-axis of pillar B, associated to
this mode can be defined by:
,
where U and V are the displacement field and the volume of the unit cell respectively. The result, depicted
by the color scale in Fig. 1(b), bears out that the torsion around z-axis is the dominant deformation for the
modes on the doubly negative branch, especially when the wave vector is close to a critical point in the BZ.
As a consequence, only SH0 Lamb waves may couple with a mode in the isolated branch. This is further
confirmed by the spectra of the transmission of A0, S0, and SH0 modes along K1 and ΓM1, represented by
the black, red and blue solid lines in Figs. 1(d) and 1(e) respectively: the doubly negative branch acts as a
forbidden band for incident A0 or S0 Lamb waves, whereas an incident SH0 Lamb wave propagates with
negligible attenuation along both primary directions. This suggests that one may form a degenerate Dirac
cone with such a doubly negative branch, and in turn obtain a structure suitable to support the topologically
protected propagation of SH0 Lamb waves.
4
2Pillar BPillar Bcurl curl curl zdVdVUUU (a)
(b)
(c)
(d)
(e)
Figure 1: (a) Schematic of the asymmetric double-sided PPnC arranged in triangular lattice and (b) the
corresponding band structure. (c) Eigenmodes at points labelled C, D and E in (b); the black arrows are for the
deformation vectors. Transmission spectra of incident A0 (black), S0 (red) and SH0 (blue) Lamb waves propagating
along (d) ΓK1 and (e) ΓM1 directions.
5
B.Artificial folding in honeycomb lattice
To construct a degenerate Dirac cone based on the aforementioned doubly negative branch, we considered
the arrangement shown in Fig. 2(a). Two pairs of asymmetric double-sided pillars are assembled into a
honeycomb unit cell, the distance between nearest neighbors being the same as for the triangular lattice
depicted in Fig. 1(a). The lattice constant is set to a = 400μm. Actually, the honeycomb unit cell can be
described as the merging of two triangular unit cells with an offset along M2 direction between them.
These two sub unit cells form equidistant Bragg plane only along K2 direction, leading to a perfect artificial
folding [41] and to a degenerate Dirac cone at point K2 of the BZ. Additionally, a thorough hole with
diameter d = 240μm is drilled at each corner of the unit cell. These additional holes allow to soften the plate
[42] without noticeable alteration of the band structure.
The corresponding band structure in the range 3 − 5.5MHz is displayed in Fig. 2(b). Two eigenmodes at
4.318MHz create a degenerate Dirac cone at the corner K2. The group velocity is 56.14ms-1 and a partial
band gap opens up in between 4.302 and 4.344MHz along ΓM2 direction. Figure 2(c) depicts the
deformations at points labelled as F, G, H1 and H2 in Fig. 2(b). At point F, both pillars BL and BR in the
unit cell undergo in-phase torsional resonance. This, together with the group velocity which is negative
around this point, suggests that the effective shear modulus negative is negative. As for the eigenmode at
point G, both pillars BL and BR exhibit out-of-phase torsion which directly comes out from the band folding.
Concerning the degenerate eigenmodes at points H1 and H2, torsional motions around z-axis in pillar BL as
well as in pillar BR are predominant. It can be shown [18] that, once the degenerate Dirac cone is lifted by
imposing SIS breaking perturbation, the new eigenmodes at the bounding are well approximated by a linear
combination of the degenerate eigenmodes. It can be therefore speculated that the torsion resonances of both
pillars BL and BR still play an important role in the perturbed configuration. The transmission spectra along
ΓK2 for incident A0 (black), S0 (red) and SH0 (blue) Lamb waves are shown in Fig. 2(d). The corresponding
pattern is shown on the top. The excitation lays on the left of the supercell and two perfectly matched layers
(PMLs) are placed at both ends to eliminate any reflected wave. The propagation of an incident SH0 Lamb
wave is allowed in the upper band (in between 4.318 and 4.353MHz, upper cyan region) but it is not for
incident A0 or S0 Lamb waves. The opposite polarization-dependent propagation can be observed at
frequency the lower band (in between 4.141 and 4.318MHz, lower cyan region). Moreover, in between
4.302 and 4.344MHz (magenta region), any Lamb mode can propagate, whatever the symmetry is, owing
to the combination of in-phase and out-of-phase deformations. The situation is pretty different along M2,
as evidenced by Fig. 2(e). Clearly in that case, only SH0 mode at a frequency either in the upper or in the
lower band, can propagate. This can be recognized by considering the symmetry about xz-plane of the
supercell drawn on top of each figure.
6
(a)
(b)
(c)
(d)
(e)
Figure 2: (a) Schematic of the asymmetric double-sided PPnC arranged in honeycomb lattice and (b) the
corresponding band structure. (c) Eigenmodes at points labelled F, G, H1 and H2 in (b); the black arrows are for the
deformation vectors. Transmission spectra of incident A0 (black), S0 (red) and SH0 (blue) Lamb waves propagating
along (d) K2 and (e) ΓM2 directions.
7
III. VALLEY-PROTECTED TOPOLOGICAL LAMB WAVES
A. Topological phase transition
Mimicking QVHE requires that SIS perturbation is imposed to the honeycomb unit cell in order to lift the
degenerate Dirac cone and to induce the topological phase transition. A non-zero valley Chern number can
be obtained this way, along with the occurrence of a reopened band gap at both valleys. In practice, this can
be done by varying either the height or the diameter of the asymmetric double-sided pillars. In doing so, the
point group C3v is broken because of the violation of the mirror-symmetry about the mid plane of the unit
cell, while the symmetry C3 is preserved.
As discussed in the previous section, the torsional vibration of the lower pillars significantly contributes to
the formation of the constituent branches of the degenerate Dirac cone. Therefore, we first consider the
perturbation in the height of the lower pillars (pillars BL and BR) according to hBL = hB + Δh and hBR = hB
Δh. The band structure of the perturbed PPnC with Δh = 1µm, hereafter referred as PPnC-I, is displayed
in Fig. 3(a). As expected, the degenerate Dirac cone is lifted and an omnidirectional band gap ranging from
4.301 to 4.335MHz reopens. The vortex chirality at the valley K2 is revealed by the phase distribution of
the out-of-plane displacement on the top surface of the plate, displayed in Fig. 3(b). In this figure, the top
and bottom images represent the phase field at the higher (4.335MHz) and at the lower (4.301MHz)
bounding frequencies respectively. In the top image, the phase field is centered on pillar AL and gradually
decreases counterclockwise, while it keeps a constant value at the location where pillar AR is standing: this
can be recognized as the valley pseudospin up state. In the bottom image, a uniformly distributed phase
field is observed at the position of pillar AL, whereas it gradually decreases clockwise from the center of
pillar AR, which corresponds to the valley pseudospin down state. It should be noted that these valley
pseudospin states will be inverted at the valley K2' if considering TRS preservation in the system. Note also
that each valley state can be selectively excited by imposing the proper chirality to fit with the pseudospin
state of the desired valley [43 -- 46]. The width of the topological band gap is fairly proportional to the
magnitude of the height perturbation h. This may be checked in Fig. 3(c) where the pseudospin up and
down states at valley K2 are drawn as red and black circles respectively. In the investigated range, the band
gap firstly closes, reopens when Δh crosses zero, whilst exchanging the frequency order of the up and down
states which are signatures of the topological phase transition.
The configuration where the perturbation of the height Δh = −1µm, hereafter referred as PPnC-II, is exactly
the space-inverted counterpart of PPnC-I. Obviously both configurations have the same band structure but
nonetheless, they are absolutely distinct from the topological point of view. As a matter of fact, perturbing
the height of the upper pillars (pillars AL and AR) according to hAL = hA + Δh and hAR = hA Δh leads to
8
the pseudospin up and down states displayed in Fig. 3(c) as black and red crosses. The comparison with the
previous case clearly shows that the band gap is more sensitive to the height perturbation of the lower pillars.
(a)
(b)
(c)
Figure 3: (a) Band structure of PPnC-I with the height perturbation Δh = 1µm. (b) Phase distribution of the out-of-
plane displacement on the top surface of the plate at the higher (top panel) and lower (bottom panel) bounding of
the lifted Dirac cone. (c) Evolution of the pseudospin up and down states at the valley K2 against the height
perturbation Δh in the lower (black and red circles) and upper (black and red crosses) pillars.
B. Berry curvature and valley Chern number
In this subsection we calculate the integral over a portion of the BZ of the Berry curvature of PPnC-I, which
is the invariant characterizing the topological nature of the structure after lifting the degenerate Dirac cone.
The Berry curvature of the nth band at a given wave vector k can be calculated from:
,
(1)
9
nnnikkkukukz where un(k) is the periodic part of the displacement field in the unit cell for the wave vector k = (kx, ky).
Because of the TRS conservation, the topological invariant known as the Chern number calculated by
integrating the Berry curvature over the whole BZ should be zero. Concerning SIS breaking in the perturbed
configuration, the Berry curvature of the lower bounding of the lifted Dirac cone at the valley K2/K2' reads
[32,33,47]:
,
(2)
where k = k -- kK2/K2' is the relative wave vector with respect to the valley K2/K2', vg is the group velocity
at the degenerate Dirac cone; m stands for the effective mass and is proportional to the frequency interval
between the pseudospin up and down states. It is also directly related to the height perturbation Δh and
represents the strength of SIS breaking. In the small SIS breaking regime, the Berry curvature is strongly
localized around the valleys K2/K2' and the valley Chern number Cv quickly converges to a non-zero
quantized value which can be theoretically derived as being [27,47]
.
(3)
Equation (3) establishes that Cv only depends on the sign of the effective mass m. On the other hand, Eq. (2)
shows that the distribution of the Berry curvature broadens as the effective mass m increases and
interferences between two opposite valleys might occur. The assumption of a small SIS breaking is no
longer valid and a large SIS breaking regime should be considered instead. In that case, when directly
integrating Eq. (2) in the vicinity of the valleys [31] to calculate the valley Chern number, a huge deviation
from the theoretical values 1/2 might arise because of the destructive interference between two opposite
valleys of the Berry curvature. Therefore, to identify the strength of SIS breaking in PPnC-I, we have
computed the distribution of the Berry curvature around the valleys [33,47]. In Fig. 4(a), the black dotted
line represents the numerical results along K2 direction when the wave vector varies from kx = 2π/3a to kx
= 10π/3a. For comparison, the theoretical values predicted by Eq. (2) is displayed as a red solid line. The
Both methods yield consistent results in between kx = 4π/3a and kx = 8π/3a, i.e. along the high symmetry
boundaries K2-M2-K2' of the BZ. From kx = 2π/3a to kx = 4π/3a the numerical results decrease more quickly
than the theoretical estimate.
We have also computed the anisotropy around the valley K2. The result is depicted in Fig. 4(b) as black,
red, and blue solid lines which represent the absolute values of the Berry curvature for the wave vector at a
distance k = 0.05K2, 0.1K2, and 0.2K2 away from the valley K2, respectively. The Berry
10
2232222ggmvvmkk211K2/K2'sgn22vCdmkk= curvature displays a circular shape when the wave vector is very close to the valley K2 that can be therefore
considered as isotropic. However, it turns into a triangular shape as the distance k increases up to 0.1K2
and even it changes to the clover-like shape when k = 0.2K2 and exhibits then strong variations if the
wave vector deviates from the high symmetry boundaries K2-M2-K2' in the BZ. This behavior relates to
the steep dispersion curve around the lifted Dirac cone. The Berry curvature around the valley K2 when the
components kx and ky vary in the intervals [π/a,2π/a] and [√3π/3a ,√3π/3a] respectively, is displayed in
Fig. 4(c). Integrating the Berry curvature over these intervals yields to a value for the valley Chern number
significantly different from what is expected from theory since we found Cv= 0.22 instead of 1/2. This
clearly indicates that the perturbation Δh = 1µm in PPnC-I corresponds to a large SIS breaking case.
Generally, a larger SIS breaking case would occur if increasing the height perturbation. Figure 4(d) depicts
both the numerical (black dotted line) and theoretical (red solid line) Berry curvature, when the height
perturbation is set to Δh = 2µm. The profile of the Berry curvature is then much broader and the integral of
the Berry curvature takes the value −0.08. In between kx = 4π/3a and kx = 8π/3a, the Berry curvature
associated with the valley K2 and K2' tends to connect directly (see the slope) suggesting the occurrence of
very strong inter-valley scattering.
(a)
(b)
11
(c)
(d)
Figure 4: Numerical (black dotted line) and theoretical (red solid line) values of the Berry curvature along K2
direction at the wave vectors varying from kx = 2π/3a to kx = 10π/3a and ky= 0 when the height perturbation is set to
be (a) Δh = 1µm and (d) Δh = 2µm. (b) Absolute values of the Berry curvature at the wave vectors with a distance
away from the valley K2 of k = 0.05K2 (black), 0.1K2 (red) and 0.2K2 (blue) respectively. (c) Distribution
of the Berry curvature around the valley K2 at the wave vectors varying from kx = π/a to kx = 2π/a and from ky =
√3π/3a to ky = √3π/3a.
C. Valley-protected Lamb waves at the domain walls
The above analysis indicates the existence of the valley-protected edge states at the domain walls formed
by the topologically distinct PPnC-I and PPnC-II. To verify this, we consider a three-layer ribbon supercell
shown in Fig. 5(a), consisting of six unit cells of PPnC-I sandwiching eight unit cells of PPnC-II. Two
zigzag domain walls are formed and zoomed in the side views, namely LDW and SDW, with the height of
adjacent lower pillars at the interface expended and shortened by 1µm respectively. The dispersion curves
of this ribbon supercell in the frequency interval ranging from 4.25 to 4.4MHz, are displayed in Fig. 5(b).
The black dotted lines represent the bulk modes; the red and blue solid lines are the edge states occurring at
LDW and SDW respectively. It should be noted that the projection of the valleys K2 and K2' onto these two
zigzag domain walls are kx = 2π/3a and kx = 2π/3a. At the valley K2, the group velocity of the edge state
localized at LDW is negative as can be foreseen if considering the change of the valley Chern number from
PPnC-I to PPnC-II, namely
. In contrast, the group velocity is turning positive
at the valley K2'. The eigenmodes at points M (4.316MHz) and N (4.323MHz) for the pillars on the rear
face of the supercell are displayed in the left and right panels of Fig. 5(c), with the red arrows denoting the
deformation vectors. Adjacent pillars on this face exhibit antisymmetric torsion about LDW interface while
adjacent pillars on the upper face display antisymmetric bending (not shown here). In contrast, the torsion
12
PPnC-ΙPPnC-ΙΙK2K21vvCC of the pillars on the rear face and the bending of the pillars on the upper face are symmetric about SDW
interface.
It has been reported [3,48] that the antisymmetric edge state leads to a deaf band because of the mismatch
in the spatial parity between the eigenmode and the deformation caused by the incident wave. We
demonstrate in what follows that this edge state may be actually excited in our system if considering an
incident SH0 Lamb wave.
(a)
(b)
(c)
Figure 5: (a) Top view of the three-layer ribbon supercell constructed by placing six unit cells of PPnC-I at two
ends and eight unit cells of PPnC-II in the middle and zoomed side views of two zigzag domain walls. (b) Dispersion
curves of the ribbon supercell. (c) Bottom views of the eigenmodes at points labelled M and N in (b).
13
The straight wave guide featuring 24×20 unit cells shown in Fig. 6(a) was designed to investigate the
unidirectional transport of the valley-protected edge states. The upper and lower domains are made of PPnC-
I and PPnC-II that forms a LDW-type interface between them. PMLs enclose the structure to eliminate any
reflected waves. In order to excite the K2'-polarized edge state [32], two elastic sources with a phase
difference of π/3 and separated by a distance of a, are placed at the red point, where the edge state features
antisymmetric deformation about the interface. An incident SH0 Lamb wave at 4.314MHz is launched by
applying y-axis polarized traction forces. The amplitude of the out-of-plane displacement on the top surface
of the plate is plotted in Fig. 6(b) plots. Consistently with the positive group velocity at the valley K2', the
wave propagates along positive x-axis only, toward the right zigzag termination where the K2'-polarized
SH0 Lamb wave gets both positively and negatively refracted. To quantitatively estimate the ability of the
structure to prevent from backscattering, one may introduce the amplitude ratio between the out-of-plane
displacement at the two ends, namely
. It is equal to 0.064 here. Although it slightly deviates
from zero, we can still consider that we are in a large SIS breaking case and that the backscattering wave
are actually suppressed at the right zigzag termination.
For comparison, Fig. 6(c) depicts the mapping of amplitude of the out-of-plane displacement when z-axis
polarized forces are applied in order to excite A0 Lamb wave at the same frequency. In that case, the elastic
energy remains highly localized around the sources and cannot propagate along the domain wall because
the generated field does not match the antisymmetric deformation of the edge state [see left panel of Fig.
5(c)]. It is therefore a forbidden band for an incident A0 Lamb mode. The same conclusion holds if S0 Lamb
mode is excited.
(a)
(b)
(c)
14
LeftRightAA Figure 6: (a) Schematic of the straight wave guide featuring LDS-type interface constructed by placing PPnC-I and
PPnC-II in the upper and lower domains respectively. The red point represents the position of two phase-matched
sources. Plots of the amplitude of the out-of-plane displacement on the top surface of the plate under the excitation
of the K2'-polarized (b) SH0 and (c) A0 Lamb waves at 4.314MHz.
Inverting the positions of PPnC-I and PPnC-II sketched in Fig. 6(a) allows to construct another straight
wave guide featuring this time a SDW-type interface. Indeed, at this domain wall, the edge state displays a
symmetric deformation about the interface. As previously did, K2'-polarized SH0 and A0 Lamb waves at
frequency 4.325MHz are excited on the place marked by the red dot in Fig. 6(a). The amplitude of the out-
of-plane displacement on the top surface of the plate are displayed in Figs. 7(a) and 7(b) respectively.
whereas SH0 Lamb wave remain localized around the sources, the excited A0 Lamb wave propagate along
negative x-axis. This behavior can be well understood when considering the spatial parity between the elastic
field generated by the sources and the symmetric deformation of the edge state at this domain wall [see right
panel of Fig. 5(c)]. In that case, only A0 Lamb mode matches the required symmetric displacement field and
the propagation along negative x-axis is consistent with the negative group velocity of the edge state at the
valley K2'.
(a)
(b)
Figure 7: Plots of the amplitude of the out-of-plane displacement on the top surface of the plate under the excitation
of the K2'-polarized (a) SH0 and (c) A0 Lamb waves at 4.325MHz. The straight wave guide is constructed by
inverting the position of PPnC-I and PPnC-II shown in Fig. 6(a).
For the sake of completeness, we have investigated the propagation of the edge state in a Z-shape wave
guide featuring two 60 sharp bending corners, as drawn in Fig. 8(a). The wave guide is created by placing
PPnC-I and PPnC-II in the upper and lower domains respectively. According to the discussion above, only
SH0 Lamb wave can propagate along the domain wall. A right-going K2'-polarized SH0 Lamb wave at
4.314MHz is launched from the red dot in Fig. 8(a). The resulting out-of-plane displacement field is
displayed in Fig. 8(b). Refracted waves appear at the left outlet which indicates the occurrence of the inter-
valley scattering of the edge state at the bending corners. Comparatively, the amplitude measured at the left
outlet is much smaller than what is measured at the right outlet, suggesting weak inter-valley scattering.
Actually, the amplitude ratio between the two outlets is 0.279, that is much larger than what is observed in
15
the straight wave guide. This shows that, despite the weak inter-valley scattering occurring at the bending
corners, most of the injected energy can propagate through the Z-shape wave guide within this large SIS
breaking frame.
(a)
(b)
Figure 8: Schematic of the Z-shape wave guide constructed by placing PPnC-I and PPnC-II in the upper and lower
domains respectively. The red point represents the position of two phase-matched sources. (b) Plot of the amplitude
of the out-of-plane displacement on the top surface of the plate under the excitation of the K2'-polarized SH0 Lamb
wave at 4.314MHz in the Z-shape wave guide.
This last conclusion encourages to verify whether the pillared structure still allows for topological protection
in an even larger SIS breaking situation. To this end, we have considered a ribbon supercell featuring the
height perturbation Δh = 2µm, that gives rise to the dispersion curves displayed in Fig. 9(a) where the red
and blue solid lines represent the edge state occurring at LDW and SDW respectively. These two edge states
have similar profiles as the ones in Fig. 5(b), but are totally gapped. The topological protection has been
examined by the propagation of the K2'-polarized SH0 Lamb wave at 4.304MHz both in the straight and in
the Z-shape wave guides. The results are shown in Figs. 9(b) and 9(c) respectively, where the waves
reflected from the zigzag outlet and the bending corners can be clearly observed. The amplitude ratios are
0.228 for the straight wave guide and goes to 0.943 for the Z-shape wave guide. Therefore, it can be
concluded that the topological protection of the edge state cannot be guaranteed any more in this larger SIS
breaking case with gapped edge states.
(a)
16
(b)
(c)
Figure 9: (a) Dispersion curves of the three-layer ribbon supercell with height perturbation Δh = 2µm. Plots the
amplitude of the out-of-plane displacement on the top surface of the plate under the excitation of the K2'-polarized
SH0 Lamb wave at frequency 4.304MHz in both the (b) straight and (c) Z-shape wave guides.
IV. Conclusion
In this work, the valley-protected topological propagation of A0, S0, and SH0 Lamb waves at different
domain walls constructed by topologically distinct asymmetric double-sided PPnCs is numerically
demonstrated. A degenerate Dirac cone is achieved by artificially folding the doubly negative branch that
occurs in the dispersion curves of both the triangular and the honeycomb lattices. At its constituent branches,
different polarization-dependent propagation along the same primary direction of BZ are observed that are
directly related to the in-phase and out-of-phase deformation in the honeycomb unit cell. Moreover, on a
given branch, divergent polarization-dependent phenomena along different primary directions in the BZ are
also reported. Afterwards, two large SIS breaking perturbations are imposed on the height of the lower
pillars that features torsional motion to lift the degenerate Dirac cone, then realizing the topological phase
transition. We observe that the Berry curvature becomes strongly anisotropic when the wave vector deviates
from the valleys. Finally, the unidirectional transport of Lamb waves at different domain walls in the straight
and Z-shape wave guides are discussed. The propagation of SH0 Lamb wave is topologically protected at
one domain wall where the propagation of A0 or S0 Lamb waves is forbidden because of the mismatch in
the spatial parities. The contrary phenomena can be observed at the other domain wall. In the large SIS
breaking case, the reflection at the zigzag outlet of the straight wave guide can be neglected and the weak
inter-valley scattering occurs at the bending corners of the Z-shape wave guide. When the strength of SIS
breaking becomes larger, the edge states are gapped and strong reflection at the zigzag outlet of the straight
wave guide and the bending corners of the Z-shape wave guide is observed. The topological protection
cannot be guaranteed any more in that case.
17
References:
[1]
A. B. Khanikaev, R. Fleury, S. H. Mousavi, and A. Alù, Topologically robust sound propagation in
an angular-momentum-biased graphene-like resonator lattice, Nat. Commun. 6, 8260 (2015).
[2]
P. Wang, L. Lu, and K. Bertoldi, Topological Phononic Crystals with One-Way Elastic Edge Waves,
Phys. Rev. Lett. 115, 104302 (2015).
[3]
J. Lu, C. Qiu, L. Ye, X. Fan, M. Ke, F. Zhang, and Z. Liu, Observation of topological valley transport
of sound in sonic crystals, Nat. Phys. 13, 369 (2017).
[4]
Z. Zhang, Q. Wei, Y. Cheng, T. Zhang, D. Wu, and X. Liu, Topological Creation of Acoustic
Pseudospin Multipoles in a Flow-Free Symmetry-Broken Metamaterial Lattice, Phys. Rev. Lett. 118,
084303 (2017).
[5]
A. Foehr, O. R. Bilal, S. D. Huber, and C. Daraio, Spiral-Based Phononic Plates: From Wave
Beaming to Topological Insulators, Phys. Rev. Lett. 120, 205501 (2018).
[6] M. Miniaci, R. K. Pal, B. Morvan, and M. Ruzzene, Experimental Observation of Topologically
Protected Helical Edge Modes in Patterned Elastic Plates, Phys. Rev. X 8, 031074 (2018).
[7]
S.-Y. Yu, C. He, Z. Wang, F.-K. Liu, X.-C. Sun, Z. Li, H.-Z. Lu, M.-H. Lu, X.-P. Liu, and Y.-F.
Chen, Elastic pseudospin transport for integratable topological phononic circuits, Nat. Commun. 9,
3072 (2018).
[8]
S. Li, D. Zhao, H. Niu, X. Zhu, and J. Zang, Observation of elastic topological states in soft materials,
Nat. Commun. 9, 1370 (2018).
[9] M. Serra-Garcia, V. Peri, R. Süsstrunk, O. R. Bilal, T. Larsen, L. G. Villanueva, and S. D. Huber,
Observation of a phononic quadrupole topological insulator, Nature 555, 342 (2018).
[10] H. He, C. Qiu, L. Ye, X. Cai, X. Fan, M. Ke, F. Zhang, and Z. Liu, Topological negative refraction
of surface acoustic waves in a Weyl phononic crystal, Nature 560, 61 (2018).
[11] B. A. Bernevig and S.-C. Zhang, Quantum Spin Hall Effect, Phys. Rev. Lett. 96, 106802 (2005).
[12] C. L. Kane and E. J. Mele, Quantum Spin hall effect in graphene, Phys. Rev. Lett. 95, 226801 (2005).
[13] M. Z. Hasan and C. L. Kane, Colloquium: Topological insulators, Rev. Mod. Phys. 82, 3045 (2010).
[14] Y. Jin, D. Torrent, and B. Djafari-Rouhani, Robustness of conventional and topologically protected
18
edge states in phononic crystal plates, Phys. Rev. B 98, 054307 (2018).
[15] C. Brendel, V. Peano, O. Painter, and F. Marquardt, Snowflake phononic topological insulator at the
nanoscale, Phys. Rev. B 97, 020102 (2018).
[16]
J. Cha, K. W. Kim, and C. Daraio, Experimental realization of on-chip
topological
nanoelectromechanical metamaterials, Nature 564, 229 (2018).
[17] L. M. Nash, D. Kleckner, A. Read, V. Vitelli, A. M. Turner, and W. T. M. Irvine, Topological
mechanics of gyroscopic metamaterials, Proc. Natl. Acad. Sci. 112, 14495 (2015).
[18] Y. Chen, X. Liu, and G. Hu, Topological phase transition in mechanical honeycomb lattice, J. Mech.
Phys. Solids 122, 54 (2019).
[19] Z. G. Chen and Y. Wu, Tunable Topological Phononic Crystals, Phys. Rev. Appl. 5, 054021 (2016).
[20] X. Ni, C. He, X. C. Sun, X. P. Liu, M. H. Lu, L. Feng, and Y. F. Chen, Topologically protected one-
way edge mode in networks of acoustic resonators with circulating air flow, New J. Phys. 17, 053016
(2015).
[21] C. Sugino, Y. Xia, S. Leadenham, M. Ruzzene, and A. Erturk, A general theory for bandgap
estimation in locally resonant metastructures, J. Sound Vib. 406, 104 (2017).
[22]
J. Chen, H. Huang, S. Huo, Z. Tan, X. Xie, and J. Cheng, Self-ordering induces multiple topological
transitions for elastic waves in phononic crystals, Phys. Rev. B 98, 014302 (2018).
[23] R. Chaunsali, C. W. Chen, and J. Yang, Subwavelength and directional control of flexural waves in
zone-folding induced topological plates, Phys. Rev. B 97, 054307 (2018).
[24] S. H. Mousavi, A. B. Khanikaev, and Z. Wang, Topologically protected elastic waves in phononic
metamaterials, Nat. Commun. 6, 8682 (2015).
[25] Y. Wu, R. Chaunsali, H. Yasuda, K. Yu, and J. Yang, Dial-in Topological Metamaterials Based on
Bistable Stewart Platform, Sci. Rep. 8, 112 (2018).
[26] H. Chen, H. Nassar, and G. L. Huang, A study of topological effects in 1D and 2D mechanical
lattices, J. Mech. Phys. Solids 117, 22 (2018).
[27] R. K. Pal and M. Ruzzene, Edge waves in plates with resonators: An elastic analogue of the quantum
valley Hall effect, New J. Phys. 19, 025001 (2017).
19
[28]
J. Vila, R. K. Pal, and M. Ruzzene, Observation of topological valley modes in an elastic hexagonal
lattice, Phys. Rev. B 96, 134307 (2017).
[29] E. Riva, D. E. Quadrelli, G. Cazzulani, and F. Braghin, Tunable in-plane topologically protected
edge waves in continuum Kagome lattices, J. Appl. Phys. 124, 164903 (2018).
[30] H. J. Lee, H. W. Kim, and Y. Y. Kim, Far-field subwavelength imaging for ultrasonic elastic waves
in a plate using an elastic hyperlens, Appl. Phys. Lett. 98, 241912 (2011).
[31] H. Zhu, T.-W. W. Liu, and F. Semperlotti, Design and experimental observation of valley-Hall edge
states in diatomic-graphene-like elastic waveguides, Phys. Rev. B 97, 174301 (2018).
[32] T. W. Liu and F. Semperlotti, Tunable Acoustic Valley-Hall Edge States in Reconfigurable Phononic
Elastic Waveguides, Phys. Rev. Appl. 9, 014001 (2018).
[33]
J. Wang and J. Mei, Topological valley-chiral edge states of Lamb waves in elastic thin plates, Appl.
Phys. Express 11, 057302 (2018).
[34]
J.-J. Chen, S.-Y. Huo, Z.-G. Geng, H.-B. Huang, and X.-F. Zhu, Topological valley transport of
plate-mode waves in a homogenous thin plate with periodic stubbed surface, AIP Adv. 7, 115215
(2017).
[35] Z. Wang, S.-Y. Yu, F.-K. Liu, Y. Tian, S. Kumar Gupta, M.-H. Lu, and Y.-F. Chen, Slow and robust
plate acoustic waveguiding with valley-dependent pseudospins, Appl. Phys. Express 11, 107310
(2018).
[36] Y. Pennec, B. Djafari-Rouhani, H. Larabi, J. O. Vasseur, and A. C. Hladky-Hennion, Low-frequency
gaps in a phononic crystal constituted of cylindrical dots deposited on a thin homogeneous plate,
Phys. Rev. B 78, 104105 (2008).
[37] Y. Pennec, B. Djafari Rouhani, H. Larabi, A. Akjouj, J. N. Gillet, J. O. Vasseur, and G. Thabet,
Phonon transport and waveguiding in a phononic crystal made up of cylindrical dots on a thin
homogeneous plate, Phys. Rev. B 80, 144302 (2009).
[38] Y. Jin, B. Bonello, R. P. Moiseyenko, Y. Pennec, O. Boyko, and B. Djafari-Rouhani, Pillar-type
acoustic metasurface, Phys. Rev. B 96, 104311 (2017).
[39] W. Wang, B. Bonello, B. Djafari-Rouhani, Y. Pennec, and J. Zhao, Exploring rotational resonance
in elastic metamaterial plates to realize doubly negative property, ArXiv 1801.06771 (2018).
20
[40] W. Wang, B. Bonello, B. Djafari-Rouhani, Y. Pennec, and J. Zhao, Double-Negative Pillared Elastic
Metamaterial, Phys. Rev. Appl. 10, 64011 (2018).
[41] S. Yves, R. Fleury, F. Lemoult, M. Fink, and G. Lerosey, Topological acoustic polaritons: Robust
sound manipulation at the subwavelength scale, New J. Phys. 19, 075003 (2017).
[42] O. R. Bilal and M. I. Hussein, Trampoline metamaterial: Local resonance enhancement by
springboards, Appl. Phys. Lett. 103, 111901 (2013).
[43]
J. Lu, C. Qiu, M. Ke, and Z. Liu, Valley Vortex States in Sonic Crystals, Phys. Rev. Lett. 116,
093901 (2016).
[44] X.-D. Chen, F.-L. Shi, H. Liu, J.-C. Lu, W.-M. Deng, J.-Y. Dai, Q. Cheng, and J.-W. Dong, Tunable
Electromagnetic Flow Control in Valley Photonic Crystal Waveguides, Phys. Rev. Appl. 10, 044002
(2018).
[45] Y. Yang, H. Jiang, and Z. H. Hang, Topological Valley Transport in Two-dimensional Honeycomb
Photonic Crystals, Sci. Rep. 8, 1588 (2018).
[46] Z. Zhang, Y. Cheng, and X. Liu, Achieving acoustic topological valley-Hall states by modulating
the subwavelength honeycomb lattice, Sci. Rep. 8, 16784 (2018).
[47] X. Wu, Y. Meng, J. Tian, Y. Huang, H. Xiang, D. Han, and W. Wen, Direct observation of valley-
polarized topological edge states in designer surface plasmon crystals, Nat. Commun. 8, 1304 (2017).
[48] S. Y. Huo, J. J. Chen, H. B. Huang, and G. L. Huang, Simultaneous multi-band valley-protected
topological edge states of shear vertical wave in two-dimensional phononic crystals with veins, Sci.
Rep. 7, 10335 (2017).
21
|
1808.04461 | 1 | 1808 | 2018-06-21T09:42:31 | Short Circuit Synchronized Electric Charge Extraction (SC-SECE): a tunable interface for wideband vibration energy harvesting | [
"physics.app-ph",
"physics.class-ph"
] | In this paper, we present a new harvesting interface, called Short Circuit Synchronous Electric Charge Extraction (SC-SECE). The SC-SECE strategy includes a tunable short-circuit time thanks to two tuning parameters, $\phi_S$ and $\Delta\phi$. $\phi_S$ stands for the phase between the mechanical displacement extrema and the energy harvesting event. $\Delta\phi$ stands for the angular time spent in the short-circuit phase. The theoretical analysis and modelling of this short-circuit influences are derived in this paper. When associated with highly coupled harvesters, it is shown that both the harvested power and bandwidth are greatly improved. These results have been numerically validated and they demonstrate the potential of this strategy for extending the bandwidth of piezoelectric vibration energy harvesters. | physics.app-ph | physics | Journées Nationales sur la Récupération et le Stockage d'Energie (JNRSE) 2018, Besançon, May 14th-15th 2018
Short Circuit Synchronized Electric Charge
Extraction (SC-SECE): a tunable interface for
wideband vibration energy harvesting
Adrien MOREL1,2*, Adrien BADEL2, Pierre GASNIER1, David GIBUS1,2, Gaël PILLONNET1
1 Univ. Grenoble Alpes, CEA, LETI, MINATEC, F-38000 Grenoble, France
2 Univ. Savoie Mont Blanc, SYMME, F-74000 Annecy, France
*[email protected]
Abstract -- In this paper, we present a new harvesting interface,
called Short Circuit Synchronous Electric Charge Extraction (SC-
SECE). The SC-SECE strategy includes a tunable short-circuit
time thanks to two tuning parameters, 𝝓𝑺 and 𝜟𝝓. 𝝓𝑺 stands for
energy harvesting event. 𝜟𝝓, stands for the angular time spent in
the phase between the mechanical displacement extrema and the
the short-circuit phase. The theoretical analysis and modelling of
this short-circuit influences are derived in this paper. When
associated with highly coupled harvesters, it is shown that both the
harvested power and bandwidth are greatly improved. These
results have been numerically validated and they demonstrate the
potential of this strategy for extending the bandwidth of
piezoelectric vibration energy harvesters.
I. INTRODUCTION
In order to make small systems and sensors autonomous
scavenging ambient energy has been widely investigated in the
last two decades as an alternative to batteries. Piezoelectric
energy harvesters (PEH) are of particular interest in closed
confined environments, where there are few solar radiations and
thermal gradients.
In order to maximize the energy harvested from
piezoelectric harvesters, the electrical interface is a key point to
consider. Several non-linear synchronous strategies, such as
Synchronous Electric Charge Extraction
(SECE) and
Synchronized Switch Harvesting on Inductance (SSHI) have
been introduced [1], and have been implemented using discrete
components [2,3] or dedicated ASIC [4,5]. Those strategies
exhibit high performance for lowly coupled and/or highly
damped piezoelectric harvesters. However, for highly coupled
and/or lowly damped piezoelectric harvesters, these strategies
may overdamp the mechanical resonator, leading to low
performances. To face this challenge, researchers started to
propose new tunable strategies inducing lower damping [6], and
even tuning the PEH resonant frequency thanks to the important
influences of the electrical interface on the mechanical
resonator [7].
In this paper, we introduce a strategy based on the
SECE interface, which introduces a tunable short-circuit phase.
As detailed extensively in this paper, this short-circuit allows to
reduce the damping induced by the electrical interface.
Furthermore, for high coupling harvesters, it allows to tune the
PEH resonant frequency, leading to an enhanced harvesting
bandwidth.
1
II. THEORETICAL MODELLING
A. Linear PEH modelling
𝑖= α𝑥−𝐶-𝑣-
A linear PEH under a periodic excitation can usually be
modelled by a system of linear differential equations, given by
(1).
𝑀𝑥+𝐷𝑥+𝐾)*𝑥+α𝑣-= −F=−𝑀𝑦
𝑥𝑡 = 𝑋7cos(θ)=𝑋7cos (ω𝑡)
where 𝑦, 𝐹 and 𝑥 stand for the ambient displacement, the
respectively. 𝑀,𝐾)*,𝐷,𝐶- and 𝛼 stand for the equivalent mass
force applied on the PEH, and the tip mass displacement,
(1)
the piezoelectric material, and
In order to find the harvested power expression, we have to
of the PEH, its short-circuited stiffness, its mechanical damping,
the capacitance of
the
piezoelectric force coefficient, respectively. Fig.1 shows the
electrical circuit modelling these equations.
B. Expression of the piezoelectric voltage
greatly simplify the calculations. The piezoelectric harvester is
either working in open or short-circuit, as illustrated in Fig.2.
Thus, the piezoelectric voltage during a single vibration period
is given by (2).
solve (1) and find the mechanical displacement magnitude 𝑋7.
As vC is a variable in (1), finding a linear expression of 𝑣- would
𝑥θdθ
,∀θ∈ ]ϕQ+Δϕ−π,ϕQ]
F
0,∀θ∈ ]ϕQ,ϕQ+Δϕ]
GHIJGKL
𝑥θdθ
,∀θ∈ ]ϕQ+Δϕ,ϕQ+π]
F
0,∀θ∈ ]ϕQ+π,ϕQ+Δϕ+π]
GHIJG
αCC
vCθ =
αCC
ϕQ∈[0,𝜋] corresponds to the angular phase between the
Δϕ∈[0,𝜋] stands for the angular time spent in the short-circuit
shown in Fig.2. As expressed by (2), vC is not sinusoidal. Only
phase. A system implementing this extraction strategy is
depicted in Fig.1, and an example of the voltage waveform is
harvesting process and the precedent displacement extremum.
(2)
This power Phijk, divided by the maximum harvestable
power 𝑃[im=𝐷KX𝑀𝑦c/8 [6] has been computed in Figure 3
with optimized parameters (𝜙a,𝛥𝜙) as a function of the
normalized vibration frequency 𝛺7=𝜔/𝜔r with
normalized squared coupling coefficients 𝑘7c and a mechanical
quality factor of 𝑄7=25. These normalized parameters have
bandwidth gain becomes more important as 𝑘7c is increased.
been extensively described in [6,7]. We can observe that the
three
1
0.8
0.6
0.4
0.2
/"0=0.86
/"0=0.43
/"0=0.29
.
,
"
'
/
-
*
,
+
'
=
"
*
)
(
'
0
0.8
1
Ω"=$/$&
1.2
1.4
1.6
Fig. 3. Frequency responses of highly coupled piezoelectric harvesters with
optimized resistive loads (dashed lines) and the proposed SC-SECE interface
(straight lines)
III. CONCLUSION
In this paper, we present the SC-SECE strategy, which is
based on a combination of the SECE interface and a tunable
short-circuit. Numerical results show that this strategy could be
used to both enhance the harvested power and the harvesting
bandwidth of highly coupled PEH Experiments confirming
these theoretical predictions are currently being run, and will be
presented during the conference.
REFERENCES
[1] E. Lefeuvre et al., "A comparison between several vibration-powered
piezoelectric generators for standalone systems", Sensors and Actuators
A: Physical, vol. 126, no. 2, pp. 405 -- 416, Feb. 2006.
[2] G. Shi et al., "An efficient self powered synchronous electric charge
extraction interface circuit for piezoelectric energy harvesting systems"
Journal of Intelligent Material Systems and Structures, vol. 27, no. 16, pp.
2160 -- 2178, Sep. 2016.
[3] Y. Wu et al., "Piezoelectric vibration energy harvesting by optimized
synchronous electric charge extraction" Journal of Intelligent Material
Systems and Structures 24(12): 1445 -- 1458, 2012.
[4] A. Quelen et al., "A 30nA Quiescent 80nW to 14mW Power Range
Shock-Optimized SECE-based Piezoelectric Harvesting Interface with
420% Harvested Energy Improvement", IEEE International Solid State
Circuit Conference, 2018.
[5] T. Hehn et al., "A Fully Autonomous Integrated Interface Circuit for
Piezoelectric Harvesters", IEEE Journal of Solid-State Circuits, vol. 47,
no. 9, pp. 2185 -- 2198, Sep. 2012.
[6] A. Morel et al., "Regenerative synchronous electrical charge extraction
for highly coupled piezoelectric generators", IEEE Midwest symposium
of circuits and systems (MWSCAS) 2017, 2017.
[7] A. Morel et al., "Short Circuit Synchronous Electric Charge
Extraction(SC-SECE) Strategy
for Wideband Vibration Energy
Harvesting", IEEE International Symposium of Circuits And Systems
(ISCAS) 2018, 2018.
Journées Nationales sur la Récupération et le Stockage d'Energie (JNRSE) 2018, Besançon, May 14th-15th 2018
the first harmonic of vC is considered in order to analytically
solve (1) while simplifying the calculations.
Vibration
Mechanical oscillator
Piezoelectric material
Electrical interface
0
-
,
1/)*+
/%&
.
/0
2&
%&
S"
$
S#
Energy
storage
#$,&
34
1
2
3
3
2
1
0
!
()
()++(−!
Fig. 2. Displacement (blue) and voltage (red) waveforms of the SC-SECE
From (2), we eventually find the expression of the Fourier
Fig. 1. System view of the proposed SC-SECE strategy [7]
Open Circuit phase
Short Circuit phase
Resonant transfert
(/0 /')
(/0 /')
(/0 /')
Δ.
()+!
()++(
"
'!
series coefficients aX and bX associated with 𝑣-:
aX=αX[πCC[π−𝛥𝜙+sin2𝜙)+2𝛥𝜙
+sin2𝜙)2
2
+2cos𝜙)+𝛥𝜙 sin𝜙)]
bX=−αX[πCC cos𝜙a+cos𝜙a+∆𝜙 c
Hence, the first harmonic of vC, vCX, is expressed by (4).
vCX=x aX𝑋7−jbX𝑋7 =xaX∗−jbX∗
where 𝑥 is the mechanical displacement in the Fourier
domain. aX∗ and bX∗ are the first Fourier coefficients aX and bX
divided by the displacement amplitude, 𝑋7.
(3)
(4)
C. Expression of the harvested power
𝑀𝑦
Due to the filtering effect of the resonator, we consider that
the first voltage harmonic may impact the PEH dynamics. We
can hence substitute the voltage expression (4) in (1) to find the
displacement amplitude. Solving (1) in the Fourier domain, the
mechanical displacement amplitude can be expressed by (5).
𝑋7=
𝐾)*−𝑀ωc+αaX∗ c+ ω𝐷+αbX∗ c
one stored in 𝐶- when θ=ϕQ. Thus, from (1), we can derive the
Phijk=ωαc2π𝐶-X[c cos𝜙)+𝛥𝜙 +cos𝜙)
For each vibration's semi-period, the harvested energy is the
harvested power expression:
(5)
c
(6)
2
|
1707.06472 | 1 | 1707 | 2017-07-20T12:30:52 | Velocity-map imaging for emittance characterization of multiphoton-emitted electrons from a gold surface | [
"physics.app-ph"
] | A velocity-map-imaging spectrometer is demonstrated to characterize the normalized transverse emittance of photoemitted electron bunches. The two-dimensional (2D) projected velocity distribution images of photoemitted electrons are recorded by the detection system and analyzed to obtain the normalized transverse emittance. With the presented distribution function of the electron photoemission angles a mathematical method is implemented to reconstruct the three-dimensional (3D) velocity distribution curve. As a first example, multiphoton emission from a planar Au surface is studied via irradiation at a glancing angle by intense 45 fs laser pulses at a central wavelength of 800 nm. The reconstructed energy distribution agrees very well with the Berglund-Spicer theory of photoemission. The normalized transverse emittance of the intrinsic electron bunch is characterized to be 0.52 and 0.05 $\pi \cdot mm \cdot mrad$ in $X$- and $Y$-directions, respectively. | physics.app-ph | physics | Velocity-map imaging for emittance characterization
of multiphoton-emitted electrons from a gold surface
Hong Ye,1, 2 Sebastian Trippel,1, 3, ∗ Michele Di Fraia,1, 3, † Arya Fallahi,1
Oliver D. Mücke,1, 3 Franz X. Kärtner,1, 2, 3 and Jochen Küpper1, 2, 3
1Center for Free-Electron Laser Science, Deutsches Elektronen-Synchrotron DESY, Notkestrasse 85, 22607 Hamburg, Germany
2Department of Physics, University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany
3The Hamburg Center for Ultrafast Imaging, University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany
(Dated: July 21, 2017)
A velocity-map-imaging spectrometer is demonstrated to characterize the normalized transverse
emittance of photoemitted electron bunches. The two-dimensional (2D) projected velocity distribution
images of photoemitted electrons are recorded by the detection system and analyzed to obtain
the normalized transverse emittance. With the presented distribution function of the electron
photoemission angles a mathematical method is implemented to reconstruct the three-dimensional
(3D) velocity distribution curve. As a first example, multiphoton emission from a planar Au surface
is studied via irradiation at a glancing angle by intense 45 fs laser pulses at a central wavelength of
800 nm. The reconstructed energy distribution agrees very well with the Berglund-Spicer theory of
photoemission. The normalized transverse emittance of the intrinsic electron bunch is characterized
to be 0.52 and 0.05 π · mm · mrad in X- and Y -directions, respectively.
PACS numbers: 79.60.Bm, 41.20.Cv, 85.60.Ha, 79.20.Ws
I.
INTRODUCTION
Time-resolved imaging of both transient molecular
structure and condensed phase dynamics with picometer-
femtosecond spatiotemporal resolution has recently be-
come possible with the advent of x-ray free-electron lasers
(XFELs) [1–7]. The high x-ray brilliance, coherence, and
ultrashort pulse durations available from these sources
are the key properties [8] that open up unprecedented
opportunities for new science. Therefore, precise control
of the x-ray pulse characteristics, including spectral cover-
age and temporal and spatial beam profiles are of utmost
importance for advanced applications. These parameters
are directly influenced by the properties of the electron
bunch generating the x-ray pulses. Therefore, the ac-
curate characterization of the electron beam quality is
indispensable for assessing available approaches in order
to enable improvements of the underlying electron beam
technology. In addtion, high quality electron bunches are
instrumental in experiments where materials are studied
using electron diffractive imaging [9–12].
The key measure in electron beam quality is electron
beam emittance, i. e., the transverse phase-space distri-
bution of the generated electron bunches. To quantify
electron beam emittance as a function of photocathode
composition and emission mechanisms, we demonstrate
a velocity-map-imaging (VMI) spectrometer that allows
us to directly access the transverse momentum distri-
bution of photoemitted electrons, enabling the measure-
ment of normalized transverse emittance from various
∗ Email: [email protected]
† Present address: Elettra-Sincrotrone Trieste S.C.p.A., 34149,
Basovizza, Trieste, Italy
cathodes. Usually, emission mechanisms are classified as
thermionic emission, photoemission, or tunneling emission
under extraordinarily high electric fields. More recently,
nanostructured and plasmonic photocathodes used with
multiphoton or strong-field optical emission have been
used as improved electron sources [13–18]. Both, the
experimental characterization and the theoretical descrip-
tion of the electron emittance from such cathodes is highly
important, which motivates the direct VMI measurements
developed here.
As a first proof-of-principle example, we report on quan-
titative measurements of multiphoton emission from a
400 nm thick Au thin film at room temperature, which
was excited with 45-fs laser pulses centered at 800 nm.
These measurements additionally allowed us to bench-
mark the performance of this new experimental setup.
Quantum-yield-dependent measurements were performed
by recording the events of electrons impinging on the
detector when varying the average laser power and the po-
larization angle, respectively. These experimental results
confirm that four-photon emission occurs from the planar
Au surface.
In our experiments the 2D transverse ve-
locity/momentum distribution of photoemitted electrons
was directly imaged onto the detector. An experimental
3D energy distribution was reconstructed from the mea-
sured 2D VMI data using a mathematical algorithm (vide
infra) and compared to the theoretically derived 3D-space
energy distribution from the Berglund-Spicer photoemis-
sion model [19–22]. The very good agreement of our
experimental results with the theoretical model demon-
strates the applicability of VMI for the characterization
of the normalized transverse emittance of photoelectron
emitters.
arXiv:1707.06472v1 [physics.app-ph] 20 Jul 2017
2
FIG. 1. Schematic of the velocity-map-imaging (VMI) spec-
trometer consisting of three parallel electrodes, R: repeller, E:
extractor, G: ground. The sample is mounted on the top of
the holder, which can be retracted from this main chamber
into a load-lock chamber.
II. EXPERIMENTAL SETUP
The velocity-mapping technique maps the velocity co-
ordinates of particles onto a 2D detector without, to first
order, the influence of the spatial coordinates. To achieve
this, a configuration of electrostatic lenses, in the simplest
case using three parallel electrodes, is employed to spa-
tially tailor the electric fields [23–25]. The electric fields
can be also used to image and magnify the spatial coordi-
nates suppressing the effect of velocity coordinates, which
is then referred to as spatial-map imaging (SMI) [23]. The
spectrometer demonstrated here aims to characterize the
electron emittance via characterizing the average spread
of electron coordinates in position-and-momentum phase
space.
The schematic of the spectrometer is shown in Fig. 1.
The sample is mounted on the top of the sample holder,
which can be retracted into a load-lock chamber. The
load lock is designed for exchanging the sample without
breaking the ultra-high vacuum (UHV) of the imaging
system. When performing the electrostatic imaging ex-
periments, the sample holder is transferred into the main
chamber and brought in contact with the repeller plate
to make sure they are at the same electric potential.
The main chamber, maintained at 10−9 mbar, contains
a stack of three cylindrically symmetric plates, labelled
repeller (R), extractor (E) and ground (G) electrodes in
Fig. 1. They are arranged in parallel, separated by 15 mm,
and, with applied potentials, serve as the electrostatic
lens. This is followed by a ∼0.5 m drift tube, which ends
with a detector assembly consisting of a double micro-
channel plate (MCP, Chevron configuration), a phosphor
screen (P46), and a CMOS camera (Optronis CL600×2)
for recording images of the electron distributions. The full
configuration is shielded against stray fields by a µ-metal
tube. A 800-nm 45-fs Ti:Sapphire laser amplifier with a 3-
kHz repetition rate was used to illuminate the sample at a
glancing incidence angle of ∼84 ◦, with a laser focal inten-
sity spot size of ∼17×160 µm2 root-mean-square (RMS)
on the sample. In our experiments, single-shot electron-
distribution images are read out at a repetition rate of
1 kHz, limited by the camera-acquisition frame rate. The
average number of electrons emitted per pulse is on the
order of one or less, thus, space charge effects, which were
reported before [26], are excluded.
To calibrate and optimize the spectrometer field con-
figuration for both SMI and VMI, a fixed potential of
-6 kV was applied to both the repeller plate and the
sample holder while the ground plate being grounded;
see supplementary information for details. While scan-
ning the extractor voltage from -5.8 kV to -4.3 kV, we
observed the focusing behavior of the electron bunch de-
pending on the extractor voltage [27]. This behavior is
revealed by the RMS of the electron bunch size in X- and
Y -directions on the detector shown in Fig. S1 (supple-
mentary information). The SIMION [28] software is used
to simulate the electric field configuration and to calcu-
late the electron trajectories from a 2D Gaussian source
with σX = 140 µm and σY = 15 µm, yielding an RMS
behavior curve that fits the experimental results. SMI is
obtained at the minimum RMS size, i. e., at an extrac-
tor voltage of -5560 V, corresponding to a magnification
factor of 7.5. From the measured SMI data, the RMS
size is analyzed to be σX = 158 µm and σY = 20 µm,
which is in good agreement with the simulated electron
bunch size and the laser focal spot size. The extractor
voltage for VMI conditions is found at -4790 V according
to the SIMION simulations and the calibration factor of
velocity-per-pixel is 8014 m/s/pixel on the detector. The
details of the simulations and experimental calibration
are described in the supplementary material. In order
to minimize field distortions, the sample front surface
should be placed in the same plane as the repeller front
surface. However, samples of different thickness lead to a
position offset with reference to the repeller front, which
strongly influences the field configuration. Therefore, the
extractor voltage for operating in SMI and VMI mode
are optimized by voltage adjustments of [50,−50] V and
[400,−200] V, respectively, to correct for a position offset
of [−0.5, 0.5] mm. In this case, re-adjusting the potential
right after exchanging a sample is necessary, but quick
(vide infra).
III. EXPERIMENTAL RESULTS
Fig. 2 a shows the photoemitted electron yield as a
function of incident laser energy on a logarithmic scale.
The data, shown in red, were measured and averaged over
four measurement sequences and the error bars show the
corresponding standard deviations of the photoemitted
electron counts due to laser fluctuations. The blue line
reflects the results of a linear regression analysis that
yielded a slope of cx ≈ 3.94, with a standard error of 0.04
and a coefficient of determination R2 ≈ 0.999.
The Fowler-Dubridge model for the n-th order pho-
laser
incidence angle
sample holder
detector
electron beam
X
Z
Y
R
E
G
3
the electron yield mostly depends on the bulk absorption
coefficient, expressed as term (1 − R)n in the Fowler-
Dubridge model [30]. R is calculated by Fresnel equations
with n1 = 1 and n2 = 0.189 + i4.71 [32] at an incidence
angle of 84 ◦. The plotted (1−R)4 curve fits very well with
the data, which proves again the 4-th order multiphoton
process.
A velocity-map image from a planar Au surface is shown
in the inset of Fig. 3 a. The image was integrated over
6×104 laser shots with an energy of ∼50 nJ, correspond-
ing to a peak intensity of 4×1010 W/cm2 on the cathode.
Generally, in laser-induced multiphoton emission the emit-
ted electron velocity vectors exhibit cylindrical symmetry
along the direction normal to the sample surface. There-
fore, the center of mass (COM) of the image is set as
coordinate origin. The corresponding angle-integrated
radial velocity distribution of the projected electrons is
plotted in Fig. 3 a as black line. To allow for comparison
with the theoretical model, the 3D velocity/energy dis-
tribution is required. Introducing a novel mathematical
method similar to the Onion Peeling algorithm [33], we
are able to reconstruct the momentum/energy distribu-
tion when the angular distribution of emitted electrons is
known. Fortunately, for multiphoton emission, the inten-
sity of photoemitted electrons at various angles θ can be
derived from the Berglund-Spicer model [20] as
I(θ) ∝ ℵ2 cos θ ·
1
1 + α l(E) ·
(2)
1
p1 − ℵ2 · sin2 θ
where α is the optical absorption coefficient, l(E) is the
electron-electron scattering length for an electron of ki-
netic energy E, and ℵ expresses the electron analogy of
refraction at the vacuum-metal boundary [34]. For a small
ℵ (our case, ℵ = 0.275), i. e., an incident photon energy
nhν comparable to the work function W , the equation can
be simplified to I(θ) ∝ cos θ [35, 36]. Therefore, the 3D
velocity distribution can be reconstructed as is described
in detail in the supplementary information.
The reconstructed velocity distribution is plotted as
blue line in Fig. 3 a, and the smoothed energy distri-
bution shown in Fig. 3 b. The energy distribution of
the emitted electrons shows an energy spread of ∼1 eV,
which corresponds to the energy difference between a
four-1.55 eV-photon excitation and the Au work function
of 5.31 eV.
IV. DISCUSSION
The Berglund-Spicer three-step model is employed as
the analytic expression for the kinetic energy distribution
of the photoemitted electrons. As the model is derived
for single-photon emission, it is implied in our analysis
that the electrons at an initial energy state E0 absorb
sufficient number of photons instantaneously, rather than
sequentially, to be pumped to a higher energy state E =
E0 + nhν. The kinetic energy distribution for single-
photon emission [19] is adapted to multiphoton emission
FIG. 2. Counts of the photoemitted electrons as function of
(a) laser average power and (b) laser polarization angle. The
experimental data for polarization angles > π is of reduced
quality due to laser drifting, etc., within the errorbar plot in
(a).
toelectric current can be written in a generalized form
as [29]
J ∝ A(1 − R)n I n F(cid:18) nhν − eφ
kT
(cid:19)
(1)
0
the Fowler function.
where n is the number of photons, h is the Planck constant,
A is the Richardson coefficient, R the reflection coefficient
from the metal surface, I the incident light intensity, φ the
ln(1 + e−(y+x)) dy
metal work function, and F (x) =R ∞
The experimental data in Fig. 2 a follow a power law
with a slope of ∼4, in agreement with a 4-photon emission
process according to the nonlinear photoelectric effect,
which indicates that simultaneous absorption of 4 photons
(photon energy 1.55 eV at 800 nm) has to take place to
overcome the metal work function W [30], which is re-
ported as 5.31–5.47 eV for Au [31]. As shown in Fig. 2 b,
varying the laser polarization angle, the photoemitted
electron intensity reaches a maximum when the laser is
p-polarized (electric field normal to the sample surface),
and appears minimum when it is s-polarized. For mul-
tiphoton emission at a certain incident light intensity,
yfit= 3.94x -7.74
(a)
104
103
102
Numberofelectrons
20
30
40
50
Laser energy (nJ)
60
70 80
(b)
p-polarization
s-polarization
Norm.intensity
4
FIG. 3. (a) Projected 2D (black curve) and reconstructed
3D (blue curve) radial velocity distribution of the measured
velocity-map image that is shown in the inset. (b) Recon-
structed kinetic-energy distribution and its simulation using
the Berglund-Spicer model assuming an electron temperature
of 6000 K. The colorbar represents the probabilities of photo-
electron kinetic energies due to the photon-energy spectrum
of the laser. The inset shows the density of states calculated
for bulk Au, which is used in the Berglund-Spicer model sim-
ulation. The blue area depicts the four-photon-ionization
range.
as
N (E) dE ∝
K C(E) α
α + 1/l(E)
dE
nhν
nhν − 1 + ln
×1 + 4 E − Ef
E − Ef! (3)
where Ef is the Fermi energy of Au. C(E) = 0.5×(1 −
pW/E) for E > W is a semiclassical threshold func-
tion. l(E) is the electron-electron scattering length, which
is proportional to E−3/2. The absorption efficient α is
calculated from the extinction coefficient k = 4.71 as
α = 4πk/λ and taken as a constant α = 7.7×105 cm−1
independent of electron energy. K is a correction factor
related to both C(E) and α l(E), which is between 0.5 to
1. To evaluate Equation 3, the probability of a photon
carrying energy hν is calculated from the measured laser
spectrum in the range from 760 to 850 nm. To overcome
the barrier of 5.31 eV, an electron is assumed to always ab-
sorb four photons (n ≡ 4). Absorption of various photon
energies leads to slight difference of the quantum yield at a
certain emitted kinetic energy as one can see from Fig. 3 b.
The main consequence of absorbing photons with various
energies is the spectral/intensity broadening, which is
illustrated by the color coding in Fig. 3 b, but with an
essentially unchanged spectral shape. We mention that
(3) only includes the emitted electrons that experience
none or one electron-electron scattering process during
transport to the metal-vacuum surface. Electron-electron
scattering is dominant over electron-phonon scattering
and reshapes the energy distribution on a fast timescale,
i. e., during an ultrashort laser pulse.
The density of states (DOS), i. e., the number of states
available for electrons at a certain energy level, is shown
in the inset of Fig. 3 b. During the photoemission pro-
cess, an energy state E0 is first occupied by an elec-
tron, which is then excited to a higher energy state
E, which was empty. As fermions, electrons obey the
Pauli exclusion principle.
In thermal equilibrium, the
possibility of electrons to occupy an available energy
state is given by the Fermi-Dirac (FD) distribution fFD.
However, excitation of a metal with ultrashort strong
laser pulses initially creates a nonequilibrium distribu-
tion that then thermalizes via electron-electron scattering
towards a Fermi-Dirac distribution. In gold, this ther-
malization occurs on a timescale of hundreds of femtosec-
onds [37, 38]. Subsequently, the electrons cool down by
dissipating energy into the lattice via electron-phonon
scattering occurring on a longer picosecond timescale. In
the following discussion, where we employ the Berglund-
Spicer model in our analysis, we assume that the elec-
tronic system can be described by a Fermi-Dirac distri-
bution with quasi-equilibrium electron temperature Te.
Hence, the appropriate densities of states and FD dis-
tributions are multiplied with the energy distribution
as N (E)dE fFD(E0) DOS(E0) (1 − fFD(E)) DOS(E),
resulting in the spectrum shown in Fig. 3 b.
The best fit with our reconstructed experimental en-
ergy distribution is obtained for an electron temperature
of 6000 K. This is comparable to previously observed
electron temperatures of 7000 K in surface-enhanced mul-
tiphoton emission from copper [39]. The high energy tail
of the spectrum indicates that very "hot" electrons are
photoemitted by the femtosecond laser pulse, consistent
with the high excess energy deposited into the electronic
system. For the energy tail up to 4 eV, except for the high
temperature, another process that might need to be taken
into account is above-threshold photoemission (ATP), i. e.,
the absorption of one (or more) extra photon, occurring
together with the four-photon process [40]. Moreover,
for our experimental conditions, we can neglect tunnel
ionization, which could result in high energy emitted elec-
trons. Since we estimate the absorbed peak intensity
for the recorded image, Fig. 3 a, to be ∼4×109 W/cm2
012345
10
6
f
E+nhν
f
E+W
Vacuumlevel
f
E
Fermilevel
4-photon
process
0
5
Energy (eV)
Berglund-Spicer, T=6000K
2D energy distribution
3D energy distribution
Smoothed 3D distribution
2D velocity distribution
3D velocity distribution
0
VX (m/s)
1.1
×106
×106
-1.1
1.1
-1.1
0
VY(m/s)
0.5
1
1.5
Velocity (m/s)
2
×106
E+W-nhν
f
80
40
0
-5
ofbulkAu(a.u.)
Densityofstates
120
0
0
12
3
45
Emitted kinetic energy (eV)
(b)
(a)
1.2
1
0.8
0.6
0.4
0.2
0
0
Norm.counts
1.2
1
0.8
0.6
0.4
0.2
Norm.counts
taking into account Fresnel losses. This implies a Keldysh
parameter γ =pW/2Up ≈ 17 (cid:29) 1, which is well in the
multiphoton emission regime; here, Up ∝ λ2I is the pon-
deromotive energy with laser wavelength λ and intensity
I.
Since both, the measured quantum yield and the mo-
mentum distribution, are in quantitative agreement with
the Fowler-Dubridge and Berglund-Spicer models, as one
would expect from multiphoton emission from a planar
Au cathode, the VMI spectrometer has successfully been
implemented as a tool to characterize the photoemitted
electrons from cathodes, especially to directly measure
the transverse momentum distribution. Assuming there
is no correlation between the location of emission and the
transverse momentum, the RMS normalized emittance n
is defined as [34]
nζ = phζ 2ihpζ
mc
2i
, with ζ ∈ {X, Y }
(4)
2i is the momen-
where hζ 2i is the spatial spread and hpζ
tum spread of the electron bunch. From the velocity map
image shown in the inset of Fig. 3 a, the RMS normal-
ized emittance of the planar Au photocathode irradiated
by 45-fs 800-nm laser pulses with a focal spot size of
σX = 161 µm and σY = 17 µm is characterized to be
nX = 0.52 π · mm · mrad and nY = 0.05 π · mm · mrad
in the X and Y -directions, respectively. To decrease the
intrinsic normalized emittance, in principle one needs
to decrease either the emission area or the momentum
spread. The former can be intuitively decreased by an
extremely tight focal spot size or sharp tip surface, which
geometrically limits the emission area. For reducing of
the momentum/energy spread, choosing a proper mate-
rial with appropriate work function and irritated by a
laser beam with matched photon energy, for example the
photoemission of Cu under 266-nm laser irradiation, is
expected to help. Further reduction is expected when
entering the strong-field emission regime, where the elec-
trons are considered to adiabatically tunnel through the
surface barrier with zero initial momentum and are then
driven by the instantaneous optical field [13, 41]. Under
these conditions electrons are expected to be emitted with
a relatively small divergence angle and significantly lower
transverse momentum spread.
V. CONCLUSIONS
We demonstrated an electron spectrometer with VMI
and SMI capabilities, which intuitively allows for the
5
measurement of the normalized transverse emittance of
photocathodes. i. e., through the direct observation of the
transverse position and momentum distributions. We ver-
ified and benchmarked the capabilities of the instrument
in a proof-of-concept experiment, in which we character-
ized the photoemitted electrons from a 400 nm thin Au
film. For ultrashort femtosecond laser pulses with a peak
intensity lower than 1012 W/cm2 at 800 nm central wave-
length, which would correspond to γ = 1, multiphoton
emission is shown to be the dominant contribution to the
entire electron current.
We intend to utilize this new setup for the emittance
characterization of electron bunches strong-field emitted
from nanotips under optical field irradiation. Such devices
should show superior emittance [14, 17]. Moreover, the
small radius of the sharp tips realize a field enhancement,
which dramatically lowers the laser power required for
entering the strong-field regime and thus avoids damaging
of the cathodes. Our ongoing work aims at the charac-
terization of electron emission from nanostructured array
emitters, which are predicted to provide high-current low-
emittance coherent electron bunches in the strong-field
emission regime.
VI. ACKNOWLEDGMENTS
We gratefully acknowledge helpful discussions with Jens
S. Kienitz and Nele Müller, the expert technical support of
Thomas Tilp, and Nicolas Tancogne-Dejean for providing
the computed density-of-states data of bulk Au.
Besides DESY, this work has been supported by the
excellence cluster "The Hamburg Center for Ultrafast
Imaging – Structure, Dynamics and Control of Matter
at the Atomic Scale" (CUI, DFG-EXC1074), the priority
program QUTIF (SPP1840 SOLSTICE) of the Deutsche
Forschungsgemeinschaft, the European Research Coun-
cil under the European Union's Seventh Framework Pro-
gramme (FP7/2007-2013) through the Consolidator Grant
COMOTION (ERC-Küpper-614507) and the Synergy
Grant AXSIS (ERC-Kaertner-609920), the Helmholtz
Association "Initiative and Networking Fund", and the
accelerator on a chip program (ACHIP) funded by the
Betty and Gordon Moore foundation.
[1] H. N. Chapman, A. Barty, M. J. Bogan, S. Boutet,
S. Frank, S. P. Hau-Riege, S. Marchesini, B. W. Woods,
S. Bajt, W. H. Benner, London W. A., E. Plönjes,
M. Kuhlmann, R. Treusch, S. Düsterer, T. Tschentscher,
J. R. Schneider, E. Spiller, T. Möller, C. Bostedt,
M. Hoener, D. A. Shapiro, K. O. Hodgson, D. van der
Spoel, F. Burmeister, M. Bergh, C. Caleman, G. Huldt,
M. M. Seibert, F. R. N. C. Maia, R. W. Lee, A. Szöke,
6
N. Timneanu, and J. Hajdu, "Femtosecond diffractive
imaging with a soft-x-ray free-electron laser," Nat. Phys.
2, 839–843 (2006).
[2] Anton Barty, Sebastien Boutet, Michael J. Bogan, Stefan
Hau-Riege, Stefano Marchesini, Klaus Sokolowski-Tinten,
Nikola Stojanovic, Ra'anan Tobey, Henri Ehrke, Andrea
Cavalleri, Stefan Düsterer, Matthias Frank, Sasa Bajt,
Bruce W. Woods, M. Marvin Seibert, Janos Hajdu, Rolf
Treusch, and Henry N. Chapman, "Ultrafast single-shot
diffraction imaging of nanoscale dynamics," Nat. Photon.
2, 415 (2008).
[3] L Young, E P Kanter, Bertold Kraessig, Y Li, A M March,
S T Pratt, R Santra, S. H. Southworth, N Rohringer,
L F DiMauro, Gilles Doumy, C A Roedig, N Berrah,
L Fang, M Hoener, P H Bucksbaum, J P Cryan, S Ghimire,
J M Glownia, D A Reis, J D Bozek, C Bostedt, and
M Messerschmidt, "Femtosecond electronic response of
atoms to ultra-intense x-rays," Nature 466, 56 (2010).
[4] Anton Barty, Carl Caleman, Andrew Aquila, Nicusor Tim-
neanu, Lukas Lomb, Thomas A White, Jakob Andreas-
son, David Arnlund, Saša Bajt, Thomas R M Barends,
Miriam Barthelmess, Michael J Bogan, Christoph Bost-
edt, John D Bozek, Ryan Coffee, Nicola Coppola, Jan
Davidsson, Daniel P Deponte, R Bruce Doak, Tomas
Ekeberg, Veit Elser, Sascha W Epp, Benjamin Erk, Hol-
ger Fleckenstein, Lutz Foucar, Petra Fromme, Heinz
Graafsma, Lars Gumprecht, Janos Hajdu, Christina Y
Hampton, Robert Hartmann, Andreas Hartmann, Günter
Hauser, Helmut Hirsemann, Peter Holl, Mark S Hunter,
Linda Johansson, Stephan Kassemeyer, Nils Kimmel,
Richard A Kirian, Mengning Liang, Filipe R N C Maia,
Erik Malmerberg, Stefano Marchesini, Andrew V Mar-
tin, Karol Nass, Richard Neutze, Christian Reich, Daniel
Rolles, Benedikt Rudek, Artem Rudenko, Howard Scott,
Ilme Schlichting, Joachim Schulz, M Marvin Seibert,
Robert L Shoeman, Raymond G Sierra, Heike Soltau,
John C H Spence, Francesco Stellato, Stephan Stern,
Lothar Strüder, Joachim Hermann Ullrich, X Wang,
Georg Weidenspointner, Uwe Weierstall, Cornelia B Wun-
derer, and Henry N Chapman, "Self-terminating diffrac-
tion gates femtosecond x-ray nanocrystallography mea-
surements," Nat. Photon. 6, 35–40 (2012).
[5] Benedikt Rudek, Sang-Kil Son, Lutz Foucar, Sascha-
W. Epp, Benjamin Erk, Robert Hartmann, Marcus
Adolph, Robert Andritschke, Andrew Aquila, Nora
Berrah, Christoph Bostedt, Nicola Bozek, Johnand Cop-
pola, Frank Filsinger, Hubert Gorke, Tais Gorkhover,
Heinz Graafsma, Lars Gumprecht, Andreas Hartmann,
Günter Hauser, Sven Herrmann, Helmut Hirsemann, Peter
Holl, Andre Hömke, Loic Journel, Christian Kaiser, Nils
Kimmel, Faton Krasniqi, Kai-Uwe Kühnel, Michael Maty-
sek, Marc Messerschmidt, Danilo Miesner, Thomas Möller,
Robert Moshammer, Kiyonobu Nagaya, Bjorn Nilsson,
Guillaume Potdevin, Daniel Pietschner, Christian Reich,
Daniela Rupp, Gerhard Schaller, Ilme Schlichting, Carlo
Schmidt, Florian Schopper, Sebastian Schorb, Claus-
Dieter Schröter, Joachim Schulz, Marc Simon, Heike
Soltau, Lothar Strüder, Kiyoshi Ueda, Georg Weidens-
pointner, Robin Santra, Joachim Ullrich, Artem Rudenko,
and Daniel Rolles, "Ultra-efficient ionization of heavy
atoms by intense x-ray free-electron laser pulses," Nat.
Photon. 6, 858–865 (2012).
[6] Anton Barty, Jochen Küpper, and Henry N. Chapman,
"Molecular imaging using x-ray free-electron lasers," Annu.
Rev. Phys. Chem. 64, 415–435 (2013).
[7] Benjamin Erk, Rebecca Boll, Sebastian Trippel, De-
nis Anielski, Lutz Foucar, Benedikt Rudek, Sascha W
Epp, Ryan Coffee, Sebastian Carron, Sebastian Schorb,
Ken R Ferguson, Michele Swiggers, John D Bozek,
Marc Simon, Tatiana Marchenko, Jochen Küpper, Ilme
Schlichting, Joachim Ullrich, Christoph Bostedt, Daniel
Rolles, and Artem Rudenko, "Imaging charge transfer in
iodomethane upon x-ray photoabsorption." Science 345,
288–291 (2014).
[8] Zhirong Huang and Kwang Je Kim, "Review of x-ray free-
electron laser theory," Phys. Rev. ST Accel. Beams 10,
1–26 (2007).
[9] H Ihee, VA Lobastov, UM Gomez, BM Goodson, R Srini-
vasan, CY Ruan, and Ahmed H Zewail, "Direct imaging of
transient molecular structures with ultrafast diffraction,"
Science 291, 458–462 (2001).
[10] B J Siwick, J R Dwyer, R E Jordan, and R J Dwayne
Miller, "An atomic-level view of melting using femtosecond
electron diffraction," Science 302, 1382–1385 (2003).
[11] Max Gulde, Simon Schweda, Gero Storeck, Manisankar
Maiti, Hak Ki Yu, Alec M Wodtke, Sascha Schäfer, and
Claus Ropers, "Ultrafast low-energy electron diffraction
in transmission resolves polymer/graphene superstructure
dynamics," Science 345, 200–204 (2014).
[12] Jie Yang, Markus Guehr, Xiaozhe Shen, Renkai Li,
Theodore Vecchione, Ryan Coffee, Jeff Corbett, Alan Fry,
Nick Hartmann, Carsten Hast, Kareem Hegazy, Keith
Jobe, Igor Makasyuk, Joseph Robinson, Matthew S Robin-
son, Sharon Vetter, Stephen Weathersby, Charles Yoneda,
Xijie Wang, and Martin Centurion, "Diffractive imaging
of coherent nuclear motion in isolated molecules," Phys.
Rev. Lett. 117, 153002 (2016).
[13] Michael Krüger, Markus Schenk, and Peter Hommelhoff,
"Attosecond control of electrons emitted from a nanoscale
metal tip." Nature 475, 78–81 (2011).
[14] G Herink, D R Solli, M Gulde, and C Ropers, "Field-
driven photoemission from nanostructures quenches the
quiver motion." Nature 483, 190–193 (2012).
[15] Anna Mustonen, Paul Beaud, Eugenie Kirk, Thomas
Feurer, and Soichiro Tsujino, "Five picocoulomb electron
bunch generation by ultrafast laser-induced field emission
from metallic nano-tip arrays," Astrophys. Lett. & Comm.
99, 103504 (2011).
[16] William P Putnam, Richard G Hobbs, Phillip D Keathley,
Karl K Berggren, and Franz X Kärtner, "Optical-field-
controlled photoemission from plasmonic nanoparticles,"
Nat. Phys. 13, 335–339 (2016).
[17] Soichiro Tsujino, Prat Das Kanungo, Mahta Monshipouri,
Chiwon Lee, and R.J. Dwayne Miller, "Measurement of
transverse emittance and coherence of double-gate field
emitter array cathodes," Nat. Commun. 7, 13976 (2016).
[18] F. X. Kärtner, F. Ahr, A. L. Calendron, H. Çankaya,
S. Carbajo, G. Chang, G. Cirmi, K. Dörner, U. Dorda,
A. Fallahi, A. Hartin, M. Hemmer, R. Hobbs, Y. Hua,
W. R. Huang, R. Letrun, N. Matlis, V. Mazalova, O. D.
Mücke, E. Nanni, W. Putnam, K. Ravi, F. Reichert, I. Sar-
rou, X. Wu, A. Yahaghi, H. Ye, L. Zapata, D. Zhang,
C. Zhou, R. J D Miller, K. K. Berggren, H. Graaf-
sma, A. Meents, R. W. Assmann, H. N. Chapman, and
P. Fromme, "Axsis: Exploring the frontiers in attosecond
x-ray science, imaging and spectroscopy," Nucl. Instrum.
Meth. A 829, 24–29 (2016).
[19] C. N. Berglund and W. E. Spicer, "Photoemission studies
7
[39] M. Aeschlimann, C. A. Schmuttenmaer, H. E. Elsayed-Ali,
R. J. D. Miller, J. Cao, Y. Gao, and D. A. Mantell, "Ob-
servation of surface enhanced multiphoton photoemission
from metal surfaces in the short pulse limit," J. Chem.
Phys. 102, 8606–8613 (1995).
[40] Francesco Banfi, Claudio Giannetti, Gabriele Ferrini,
Gianluca Galimberti, Stefania Pagliara, Daniele Fausti,
and Fulvio Parmigiani, "Experimental evidence of above-
threshold photoemission in solids," Phys. Rev. Lett. 94,
037601 (2005), arXiv:1201.3049 [physics].
[41] P. B. Corkum, "Plasma perspective on strong-field mul-
tiphoton ionization," Phys. Rev. Lett. 71, 1994–1997
(1993).
of copper and silver: Theory," Phys. Rev. 136, A1030
(1964).
[20] C. N. Berglund and W. E. Spicer, "Photoemission studies
of copper and silver: Experiment," Phys. Rev. 136, A1044–
A1064 (1964).
[21] W. F. Krolikowski and W. E. Spicer, "Photoemission
studies of the noble metals. i. copper," Phys. Rev. B 185,
882–900 (1969).
[22] W. F. Krolikowski and W. E. Spicer, "Photoemission
studies of the noble metals. ii. gold," Phys. Rev. B 1,
478–487 (1970).
[23] André T. J. B. Eppink and David H. Parker, "Velocity
map imaging of ions and electrons using electrostatic
lenses: Application in photoelectron and photofragment
ion imaging of molecular oxygen," Rev. Sci. Instrum. 68,
3477–3484 (1997).
[24] A I Chichinin, K H Gericke, S Kauczok, and C Maul,
"Imaging chemical reactions - 3d velocity mapping," Int.
Rev. Phys. Chem. 28, 607–680 (2009).
[25] M. Stei, J. von Vangerow, R. Otto, A. H. Kelkar, E. Car-
rascosa, T. Best, and R. Wester, "High resolution spatial
map imaging of a gaseous target," J. Chem. Phys. 138,
214201 (2013).
[26] Guillaume Petite, Pierre Agostini, Rusty Trainham, Eric
Mevel, and Philippe Martin, "Electron emission from
metals under laser irradiation," Phys. Rev. B 45, 12 210 –
12 217 (1992).
[27] Nele L. M. Müller, Sebastian Trippel, Karol Długołęcki,
and Jochen Küpper, "Electron gun for diffraction exper-
iments on controlled molecules," J. Phys. B 48, 244001
(2015), arXiv:1507.02530 [physics].
[28] Scientific Instrument Services Inc., USA, "Simion 8.1,"
[32] Mikhail N. Polyanskiy, "Refractive index database,"
on
accessed
https://refractiveindex.info,
24. July 2017.
[33] Cameron J. Dasch, "One-dimensional tomography: a com-
parison of Abel, onion-peeling, and filtered backprojection
methods," Applied Optics 31, 1146 (1992).
[34] David H. Dowell and John F. Schmerge, "Quantum effi-
ciency and thermal emittance of metal photocathodes,"
Phys. Rev. ST Accel. Beams 12, 074201 (2009).
[35] R. T. Poole, R. C. G. Leckey, J. G. Jenkin,
and
J. Liesegang, "Photoelectron angular distribution from
gold," J. Elec. Spec. Rel. Phen. 2, 371–376 (1972-1973).
[36] Zeting Pei and C. Neil Berglund, "Angular distribution of
photoemission from gold thin films," Jpn. J. Appl. Phys.
41, L52–L54 (2002).
[37] W. S. Fann, R. Storz, H. W K Tom, and J. Bokor,
"Direct measurement of nonequilibrium electron-energy
distributions in sub-picosecond laser-heated gold films,"
Phys. Rev. Lett. 68, 2834–2837 (1992).
[38] W. S. Fann, R. Storz, H. W K Tom, and J. Bokor,
"Electron thermalization in gold," Phys. Rev. B 46, 13592–
13595 (1992), arXiv:1011.1669 [physics].
(2011), URL: http://simion.com.
[29] J. H. Bechtel, W. L. Smith, and N. Bloembergen, "Four-
photon photoemission from tungsten," Opt. Comm. 13,
56–59 (1975).
[30] A Damascelli, G Gabetta, A Lumachi, L Fini,
and
F Parmigiani, "Multiphoton electron emission from Cu
and W: An angle-resolved study." Phys. Rev. B 54, 6031–
6034 (1996).
[31] David R. Lide, CRC Handbook of Chemistry and Physics,
84th ed. (CRC Press, 2003).
Supplementary information: Velocity-map imaging for emittance characterization
of multiphoton-emitted electrons from a gold surface
Hong Ye,1, 2 Sebastian Trippel,1, 3, ∗ Michele Di Fraia,1, 3, † Arya Fallahi,1
Oliver D. Mücke,1, 3 Franz X. Kärtner,1, 2, 3 and Jochen Küpper1, 2, 3
1Center for Free-Electron Laser Science, Deutsches Elektronen-Synchrotron DESY, Notkestrasse 85, 22607 Hamburg, Germany
2Department of Physics, University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany
3The Hamburg Center for Ultrafast Imaging, University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany
(Dated: July 21, 2017)
FIG. S2. Position dependence of the center of mass of the
electrostatic imaging on the detector on the initial source
position for various extractor potentials from -5800 V to -
4500 V.
incidence irradiation, and the finite kinetic energy of the
electrons.
When increasing the extractor voltage, the electron
bunch diverges. Based on our simulations, the extractor
voltage for VMI is approximately -4790 V. For a full cali-
bration of the spectrometer, the simulations were used to
study the field configuration and the electron trajectories
in those fields for the given electrode configurations and
the particles initial distributions. In Fig. S1 the simulated
RMS of the electron bunch, with electrostatic imaging, at
the detector position is plotted as function of extractor
voltage. The simulations were carried out given an initial
spatial 2D Gaussian distribution of 2000 electrons for each
simulated point. The center of mass (COM) of this distri-
bution was given by (X, Y ) = (0, 0) and a Z-coordinate
matching the sample surface with standard deviations of
σX = 140 µm and σY = 15 µm. The initial momentum
distribution was given by a uniform half sphere with an
uniform kinetic energy distribution of electrons in the
range of [0.1, 0.6] eV.
The COM of the electron distribution as a function of
the initial starting position of the electrons, i. e., the laser
focus position on the sample, was used to experimentally
calibrate the voltage for velocity-map imaging. Fig. S2
shows the COM as function of the laser position for various
voltages together with straight-line fits. A decrease of
I. SPECTROMETER CHARACTERIZATION
The electron spectrometer has been characterized exper-
imentally, accompanied by simulations, in order to deter-
mine the focusing conditions for the SMI and VMI modes;
see Fig. 1 in the main manuscript for the experimental
setup. Fig. S1 shows the measured root-mean-square
(RMS) in the X- and Y -directions of the spatial electron
distribution on the detector as a function of the extractor
voltage, together with the results from SIMION [1] elec-
tric field and particle trajectory simulations. A similar
behavior as in Ref. 2 is observed.
The strongest focusing of the electron bunch onto the
detector is achieved at an extractor potential of -5560 V,
which is thus identified as the SMI voltage. The RMS at
this voltage shows the magnified laser-surface-interaction
area. The slightly different focusing behavior of the elec-
tron bunch in the X and Y -directions is attributed to the
asymmetric initial electron bunch size, due to the glancing
FIG. S1. Experimental (hollow) and simulated (solid) root-
mean-square deviations of electron spatial distributions on the
2D detector versus focusing extractor voltage in both X and
Y -directions. The insets show SMI and VMI detector images
for the indicated positions.
∗ Email: [email protected]
† Present address: Elettra-Sincrotrone Trieste S.C.p.A., 34149,
Basovizza, Trieste, Italy
arXiv:1707.06472v1 [physics.app-ph] 20 Jul 2017
experiment, RMSX
experiment, RMSY
simulation, RMS X
simulation, RMS Y
VMI
hmm)
Y
Numberofelectronspershot
0.02
0.015
0.01
.005
00
X hmm)
24.4
11.7
SMI
25.4
11
hmm)
Y
-5500
-5000
-4500
Extractor voltage hV)
X hmm)
-4000
012345
RMSofelectrostaticimaginghmm)
E:-4500V
E:-4600V
E:-4700V
E:-4750V
E:-4775V
E:-4800V
E:-4825V
E:-4850V
E:-4900V
E:-5000V
E:-5300V
E:-5500V
E:-5550V
E:-5560V
E:-5570V
E:-5800V
1
0
0.5
Initial position (mm)
24
22
20
18
16
14
12
Positionondetector(mm)
10
-0.5
2
FIG. S3. Slope of the experimental laser position dependent
COM of the spatial distribution at the detector as function of
the extractor voltage (red circles) with a quadratic fit (red line).
Black points and the black line indicate the corresponding
simulated results.
FIG. S4. The COM of electrostatic imaging on the detector
as a function of the initial source position for SMI mode, i. e.,
an extractor voltage of -5560 V. The slope in Y -direction is
the spatial magnification factor. The ratio between X- and
Y -directions confirms the incidence angle of the laser beam of
84 ◦.
the slope with decreasing extractor voltage is observed.
Fig. S3 depicts the slope of each measurement in Fig. S2 as
function of the extractor voltage together with a quadratic
fit and corresponding simulation results. The error bar
for the experimental points is given by the first-order
coefficient error of each fitting curve with 95 % confidence
bounds. VMI mode is obtained at the zero crossing
of this curve, i. e., at -4790 V, as for this voltage the
distribution, to first order, becomes independent of the
starting position. The data shows a good agreement with
the simulations, confirming that the extractor voltage
for operating the VMI is -4790 V. From the simulations
the imaging setup is calibrated regarding the transverse
electron velocities to 8014 m/s/pixel on the detector. The
resulting voltages for operation in the SMI and VMI
modes are listed in Table I.
Fig. S4 shows experimental and theoretical COM of the
electron distribution at the detector as function of the
lens position, that is used to focus the laser beam onto the
sample for SMI (E:-5560 V). The straight lines are fits to
the data. The difference in the slope between the X- and
Y -directions is due to the glancing incidence angle θ. The
laser spot position on sample moves 1/ cos θ times farther
in X than in Y when displacing the laser beam the same
distance by a translation stage. For the Y -direction we
obtain a magnification factor of ∼ 7.5 from the fit. For the
X-direction a slope of ∼72.7 is obtained. This results in
a ratio of 9.7 between the two slopes that corresponds to
TABLE I. Voltages (in V) applied for operation in SMI and
VMI mode
Repeller Extractor Ground Sample
-6000
-6000
-5560
-4790
0
0
SMI
VMI
-6000
-6000
an incident angle of 84 ◦. The SIMION simulation results,
also shown in Fig. S4, are in good agreement with the
data.
The focusing conditions for the SMI and VMI mode
depend strongly on the position of the sample inside the
velocity-map imaging spectrometer. Fig. S5 shows the
simulated extractor voltages necessary for SMI and VMI
mode for various sample displacements with respect to
the front surface of the repeller plate. These simulations
show that either the sample position has to be known,
or at least be reproduced, to a very high precision or
calibration measurements have to be performed when a
new sample is inserted into the spectrometer. Fortunately,
with the protocol described in our manuscript this cali-
bration can be done quickly. In addition the dependence
FIG. S5. Top: SMI and bottom: VMI extractor potential for
different position offset from sample front to the repeller front
surfaces. FD stands for flying distance in the figure legend.
experimental data
yfit.=-2.7e-06x2-0.02x-26.5
simulation
yfit.=-3.6e-06x2-0.028x-50
-5800 -5600 -5400 -5200 -5000 -4800 -4600 -4400
Extractor Voltage (V)
02
-2
-4
-6
-8
-10
Slope
experiment, X
slope, X: -72.7
experiment, Y
slope, Y: -7.5
simulation
slope, simulation: -7.62
0
1
Initial position (mm)
0.5
1.5
-1
-0.5
40
35
30
25
20
15
10
5
Positionondetector(mm)
FD=481 mm
FD=496 mm
-0.5
0
0.5
1
FD=481 mm
FD=496 mm
-0.5
0
Sample to Repeller front surface offset (mm)
0.5
1
-5400
-5450
-5500
-5550
-5600
-4200
-4400
-4600
-4800
-5000
Extractor(V)
SMI
Extractor(V)
VMI
of the extractor voltage on the flight distance has been
investigated (red points and lines). Our simulations show
that this uncertainty is uncritical compared to the exact
sample position in the spectrometer.
distribution fi by a transfer matrix M.
ρi = Mfi,
with M given by:
3
(1)
(2)
1 1/4 1/9 1/16 ···
0 3/4 3/9 3/16 ···
5/9 5/16 ···
0
7/16 ···
0
0
.
.
.
...
.
.
.
.
.
.
0
0
.
.
.
II. RECONSTRUCTION ALGORITHM
M =
Our reconstruction algorithm for the conversion of the
2D projected velocity distribution to the 3D distribution
is based on the assumption that the angular distribution
of the photoemitted electrons is known. For our sim-
ulations, a cosine function I(θ) ∝ cos θ [3, 4], derived
from the Berglund-Spicer model [5] as discussed in the
main text, is applied in the algorithm. In addition, it
is assumed that for multiphoton emission the angular
distribution is independent of the modulus of the three
dimensional velocity vector. The 3D velocity distribution
is then obtained from the 2D projected distribution by a
matrix method similar to Onion Peeling [6]. For multi-
photon emission from a planar Au surface, the electrons
are assumed to be photoemitted within a half sphere of
ϕ ∈ [0, 2π], θ ∈ [0, π/2]. The photoemitted electron
distribution has cylindrical symmetry with respect to the
surface normal of the sample.
Fig. S6 a shows a scatter plot for a single 3D velocity vi
distribution given by f (v, θ) = δ(v − vi) cos θ. Fig. S6 b
shows the projection of this distribution onto the 2D
detector surface. It can be derived that the projected
velocity distribution for this special case is
Pi(vx, vy) =Z f (v, θ) dvz =(C for vx,y < vi,
otherwise
0
where C is a constant. As shown in Fig. S6 b, the pro-
jected velocity distribution of f (v, θ) is constant inside
the circular phase-space area of radius vi. Furthermore,
Fig. S6 c shows the radial distribution obtained from the
projected velocity distribution given by
0
ρi(v2D) =Z Pi(vx, vy) dθ2D =(2πC · v2D for v2D < vi
where v2D = pvx
2. In the reconstruction, each
radial distribution ρi(v2D) is built up by a triangle as
sketched in Fig. S6 d. vi is taken equally spaced and form
the intervals confined by the neighboring gray dashed
lines. The 2D projected distribution is related to the 3D
otherwise
2 + vy
The 3D distribution can finally be obtained by inversion
of the measured 2D-projected distribution
fi = M−1ρi.
(3)
FIG. S6. (a-c) Representation of a simulated electron bunch
with a single 3D velocity vi and an angular distribution of a
cosine function: (a) in 3D, forming a spherical surface; (b) in
2D, yielding a uniform distribution in the detector plane; (c)
in 1D, showing a linearly increasing radial velocity v2D with
distance from distribution COM. (d) A conceptual diagram of
the reconstruction algorithm: The area of each red triangle at
the bottom indicates the number of photoemitted electrons
having the same 3D velocity. The corresponding distribution
curve is plotted as blue curve. The black curve is the 2D
projection distribution curve, summing up the number of pho-
toemitted electrons within each interval of the same transverse
velocity. The gray dashed lines indicate the transverse-velocity
intervals used in this projection.
[1] Scientific Instrument Services Inc., USA, "Simion 8.1,"
(2011), URL: http://simion.com.
Radial
(c)
10
5
V2D
300
150
0
0
Counts(a.u.)
2D projection
3D reconstruction
3D
(a)
Projected
(b)
-5
0 5
VX
0 5
-5
Vy
-5
0
VX
5
-5
5
0
Vy
(d)
VZ
Counts(a.u.)
0
0
5
10
V (a.u.)
2D
15
20
photoemission from gold thin films," Jpn. J. Appl. Phys.
41, L52–L54 (2002).
[5] C. N. Berglund and W. E. Spicer, "Photoemission studies
of copper and silver: Experiment," Phys. Rev. 136, A1044–
A1064 (1964).
[6] Cameron J. Dasch, "One-dimensional tomography: a com-
parison of Abel, onion-peeling, and filtered backprojection
methods," Applied Optics 31, 1146 (1992).
2] Nele L. M. Müller, Sebastian Trippel, Karol Długołęcki,
and Jochen Küpper, "Electron gun for diffraction exper-
iments on controlled molecules," J. Phys. B 48, 244001
(2015), arXiv:1507.02530 [physics].
4 [
[3] R. T. Poole, R. C. G. Leckey, J. G. Jenkin,
and
J. Liesegang, "Photoelectron angular distribution from
gold," J. Elec. Spec. Rel. Phen. 2, 371–376 (1972-1973).
[4] Zeting Pei and C. Neil Berglund, "Angular distribution of
|
1904.06255 | 1 | 1904 | 2019-04-12T14:38:13 | 3D photo-responsive optical devices manufactured by advanced printing technologies | [
"physics.app-ph",
"physics.optics"
] | Photonic components responsive to external optical stimuli are attracting increasing interest, because their properties can be manipulated by light with fast switching times, high spatial definition, and potentially remote control. These aspects can be further enhanced by novel architectures, which have been recently enabled by the availability of 3D printing and additive manufacturing technologies. However, current methods are still limited to passive optical materials, whereas photo-responsive materials would require the development of 3D printing techniques able to preserve the optical properties of photoactive compounds and to achieve high spatial resolution to precisely control the propagation of light. Also, optical losses in 3D printed materials are an issue to be addressed. Here we report on advanced additive manufacturing technologies, specifically designed to embed photo-responsive compounds in 3D optical devices. The properties of 3D printed devices can be controlled by external UV and visible light beams, with characteristic switching times in the range 1-10 s. | physics.app-ph | physics | Published in SPIE Proceedings Vol. 10915, Organic Photonic Materials and Devices XXI, 1091503
(February 27, 2019) - DOI: 10.1117/12.2512039
3D photo-responsive optical devices manufactured by advanced
printing technologies
Adam Szukalski,a Sureeporn Uttiya,a Francesca D'Elia,b Alberto Portone,a Dario Pisignanoa,c, Luana
aNEST, Istituto Nanoscienze-CNR, Piazza S. Silvestro 12, I-56127 Pisa, Italy; bNEST, Scuola
Normale Superiore, Piazza San Silvestro 12, I-56127 Pisa, Italy; cDipartimento di Fisica, Università
di Pisa, Largo B. Pontecorvo 3, I-56127 Pisa, Italy
Persanoa, Andrea Camposeo*a
ABSTRACT
Photonic components responsive to external optical stimuli are attracting increasing interest, because their properties can
be manipulated by light with fast switching times, high spatial definition, and potentially remote control. These aspects
can be further enhanced by novel architectures, which have been recently enabled by the availability of 3D printing and
additive manufacturing technologies. However, current methods are still limited to passive optical materials, whereas
photo-responsive materials would require the development of 3D printing techniques able to preserve the optical
properties of photoactive compounds and to achieve high spatial resolution to precisely control the propagation of light.
Also, optical losses in 3D printed materials are an issue to be addressed. Here we report on advanced additive
manufacturing technologies, specifically designed to embed photo-responsive compounds in 3D optical devices. The
properties of 3D printed devices can be controlled by external UV and visible light beams, with characteristic switching
times in the range 1-10 s.
Keywords: 3D printing, optical window, photo-isomerization, stereolithography, fused deposition modeling.
1. INTRODUCTION
A paradigm shift is currently occurring in the design and fabrication of optical and photonic devices, especially as a
consequence of the introduction of additive manufacturing (AM) technologies.1 Conventional photonic and
optoelectronic components (lenses, optical fibers, light sources and detectors) typically feature very simple geometries,
such as curved spherical surfaces, filaments, and planar multilayers, possibly ordered in arrays to build the final devices.
Recent examples include, for instance, arrays of Mach-Zender interferometers, which are designed and assembled in
cascade to perform machine learning tasks.2 The current progress of 3-dimensional (3D) printing and AM technologies is
pushing optical and photonic devices towards new, unexplored architectures,3,4 enabled by the realization of 3D material
configurations with almost no restriction in terms of printed geometry. AM technologies comprise a variety of
fabrication processes which have been classified in seven macro-areas,5 taking in account their basic operating
principles. The starting point in all AM technologies is a digital model of the component to be realized, that is sliced in
various layers. The object is then realized by generating consecutively the various layers, in the so-called layer-by-layer
fashion, either by curing liquid resins, or by extrusion of melted polymers and viscous solutions, or by sintering fine
powders, or by delaminating solid layers.6 One of the most popular 3D printing methods is the fused deposition modeling
(FDM), that is based on the extrusion of a melted polymer filament through a nozzle. The 3D object is built layer-by-
layer by translating the nozzle along the 3D path that is given by the sliced digital model.7 The FDM can be applied to
thermoplastic polymers and composites, it can be used to build large-area components and functional devices,8,9 but it
has a spatial resolution limited by the nozzle size (typically of the order of a few hundreds of micrometers). This is one
of the reasons why FDM has been mainly used for optical components in long wavelength ranges (terahertz and
microwaves).10 Also relevant in this framework is the often weak optical transmittance of parts printed by FDM at
visible wavelengths. In fact, for some applications, such as lab-on-chips11 and device packaging,12 FDM would be highly
advantageous, because it would allow plastic components with moderate optical transmittance (>50%) to be printed in a
fast and low-cost way, and then embedded in functional devices such as supports or windows for optical access to
diagnostic environments. This requires, however, the engineering of FDM to comply with transparent thermoplastic
*[email protected]; phone +39 050 509517.
Published in SPIE Proceedings Vol. 10915, Organic Photonic Materials and Devices XXI, 1091503
(February 27, 2019) - DOI: 10.1117/12.2512039
materials, such as polycarbonate (PC), polystyrene, and poly(methyl methacrylate). This field is almost unexplored,
since most of the research works on FDM-printing with such polymers has been focused on the investigation of the
morphological and mechanical characterization of printed parts.13,14
Another AM technology for printing optical components is the laser stereolithography (STL),15 a method based on the
photo-polymerization of a liquid resin by means of a laser beam, which is scanned to solidify the individual layers
forming the 3D object. The various layers are printed consecutively by moving vertically the object holder. STL has
better spatial resolution than FDM, and it may reach the sub-m scale when performed on materials with nonlinear
absorption phenomena,16 though at the cost of slower printing rate. Nevertheless, even standard UV STL has been used
for successfully printing various optical components, such as aspheric lenses, mirrors and optical guides operating in the
visible range.17-19 In this framework, a current challenge is to develop photo-curable materials and printing methods, that
are suitable to achieve optically-active components, namely systems which can feature some specific optical properties
(i.e. emission at well-defined spectral bands, nonlinear and time-changing properties, etc.). Indeed, photonic devices and
integrated systems are continuously shifting from static and passive systems, basically able to control the propagation of
light,2,20 to active elements capable of reconfiguring themselves in response to external signals.21 This can be currently
obtained by phase-changing materials, whose crystalline structure and physical properties can be reversibly modified by
irradiation with laser pulses.22,23 Other examples include the exploitation of stimuli-responsive molecular compounds,24
which can be incorporated in photonic devices and employed for tuning the emission wavelength of organic lasers,25,26
for making birefringent films and nanofibers,27,28 and for controlling the optical transmittance of smart optical
windows.29,30 To fully exploit the enormous potential of such novel optical materials, the 3D printing technology has to
be addressed to manufacturing with them, without degrading their properties.
Here we report on printing processes for the fabrication of transparent layers by FDM and STL, and photo-responsive
layers by photo-polymerization. By optimization of the process parameters (single-layer thickness and printing speed),
PC optical windows are realized by FDM, with optical transmittance at visible wavelengths larger than 50%. In addition,
optical windows with high transparency (>80%) are printed by STL. We also show the possibility to functionalize photo-
curable layers by a photo-responsive molecular compound. In particular, we succeeded in printing photoactive layers,
suitable for reversible control of the intensity of light propagating through the printed components. This allows the
intensity of a polarized light beam to be controlled by a second light beam, with characteristic switching times of the
order of 1-10 s. These results open interesting perspectives for utilizing 3D printed, transparent and photoactive
components in a wide range of devices, including analytical and sensing optical tools.
2. METHODS
A PC filament (Roboze) was used for printing optical windows by FDM. The E-Shell® 600 (ES600, EnvisionTEC) was
used as matrix for samples printed by STL. The printable photo-responsive resin was obtained by mixing the N-Ethyl-N-
(2-hydroxyethyl)-4-(4-nitrophenylazo)aniline (DR1, Sigma Aldrich) with the ES600 matrix (1% DR1:ES600
weight:weight ratio).
FDM experiments were performed by the ONE+400 system (Roboze), equipped with two independent extruders, with
maximum operational temperature of 450 °C. The maximum printing volume is 20×20×20 cm3 (X×Y×Z, here Z is the
direction parallel to the sample thickness). Samples were printed on a PC substrate, that was kept at a constant
temperature of 80 °C. The fabrication of the optical windows was performed by using an extruder with a nozzle of 400
µm diameter. The extruder temperature was varied in a range of 240-290 °C. A temperature of 250 °C was selected as
the one allowing samples with higher optical transparency to be printed.
The Andromeda (Sharebot) 3D printer was used for STL experiments. This is a laser 3D printer, with a maximum
printing volume of 25×25×25 cm3. Optical windows were printed by a laser power fluence of 8 mW/cm2 and a layer
thickness of 100 μm. The first two layers were printed at a laser scanning rate of 1.5×105 µm/s, whereas the remaining
layers were printed using a higher rate (2.5×105 µm/s). This set of parameters allows a good adhesion to the sample
holder to be achieved, as well as a final printed structure with geometrical features in agreement with the original design.
After the printing process, the printed part was washed in isopropanol for 5 minutes, and dried under a nitrogen flow.
Post-processing surface finishing was performed by coating the samples with a thin and homogenous film of liquid resin
(thickness 200 µm), that is then polymerized by exposure to UV light (maximum power fluence 10 mW/cm2) for two
minutes. The finished part was washed again for 5 minutes in isopropyl alcohol to remove the uncured layer in contact
with air, because oxygen inhibits radical polymerization and leaves the outermost layer uncured.
Published in SPIE Proceedings Vol. 10915, Organic Photonic Materials and Devices XXI, 1091503
(February 27, 2019) - DOI: 10.1117/12.2512039
UV-visible (UV-Vis) transmittance spectra were measured by using either a Lambda 950 (Perkin Elmer) or a V-550
(Jasco) spectrophotometer. Photo-induced birefringence was investigated by a pump-probe experimental set-up, which is
typically used for the investigation of the optical Kerr effect.28 A laser probe beam (λprobe = 638 nm, linearly polarized) is
sent through the sample at normal incidence. The probe beam polarization is determined by a polarizer positioned
between the probe laser source and the sample, while the light transmitted by the sample is analyzed by a second
polarizer with axis perpendicular to the first. The intensity of the probe beam transmitted by the sample and by the
analyzer is measured by a Si photodiode. In order to induce optical anisotropy in the investigated sample, a pump laser
beam (λpump = 532 nm, linearly polarized) irradiates the same area probed by the red laser. The directions of the
polarization of the pump and probe laser beams form an angle of 45 degrees. This configuration is known to maximize
the photo-induced optical anisotropy.28 With the pump beam switched off, the sample does not feature any optical
birefringence and, as a consequence of the presence of two polarizers with crossed axis along the path of the probe beam,
no signal is detected by the photodiode. By switching the pump beam on, an optical anisotropy is induced in the area
irradiated by the pump, which leads to a rotation of the polarization of the probe beam passing through the sample, and
an increase of the signal transmitted at the second polarizer according to the relation:
(1)
where d is the thickness of the photoactive layer, Δn is photo-induced birefringence, and I0 and I are the incident and
transmitted probe intensities, respectively. By measuring the temporal evolution of this signal, one is able to measure the
characteristic times of the molecular photo-alignment occurring in the printed samples. In order to characterize the
possibility to modulate the probe beam by switching on and off the pump, a mechanical chopper with variable
modulation frequency in the range 1-800 Hz was positioned along the path of the pump beam.
3. RESULTS AND DISCUSSION
Figure 1 summarizes the measured UV-Vis optical properties of squared optical windows made of PC and printed by
FDM. An example of the printed sample is shown in the inset of Figure 1b, which evidences the good optical
transparency, allowing for easily seeing through the printed component. All the investigated samples show a low optical
transmittance in the range 200-300 nm (T=IT/Iin<4%, where IT and Iin are the intensities of the transmitted and incident
beam, respectively). This is typical of PC, which is, indeed, used as plastic blocking material for UV light.31 We have
investigated optical properties of the printed components upon varying two process parameters, i.e. the thickness of
single printed layer (Figure 1a,b) and the printing speed (Figure 1c,d). The intensity of the light transmitted by the
samples is found to drop by almost an order of magnitude upon increasing the layer thickness (Figure 1b). This trend is
found for wavelengths across the whole visible and near-infrared (NIR) spectral range (400-800 nm), with highest values
of the transmittance (T≈50 %) obtained for a layer thickness of 50 µm. In addition, the intensity of transmitted light
increases by an order of magnitude upon increasing the printing speed (Figure 1d). Other experimental reports evidenced
that in samples printed by FDM, more rough surfaces are associated to higher thicknesses of the single layers and to
lower values of the printing speed.32 The increase of the roughness of samples is expected to increase the intensity of
light that is diffused at angles out of the initial propagation direction, and, consequently to decrease the overall optical
transmittance. Controlling sintering of adjacent printed filaments7 and reducing the roughness of the printed structures
turns out to be especially critical for improving optical transmittance of layers printed by FDM.
Optical windows printed by STL share some of these features (Figure 2a,b). Here, the intensity of the light transmitted
by as-printed samples is found to be in the range 40-50 % in the visible and NIR spectral range. However, post-printing
surface finishing is highly effective in improving the transmittance of these samples, leading to T values larger than 80%.
The deposition of a thin resin film on printed samples decreases the surface roughness, and the associated diffusion of
incident light.18 Overall, the results here shown evidence that samples with acceptable optical transmittance can be
printed by both FDM and STL. In order to achieve values of T comparable to plastic and glass optical windows, a proper
surface finishing method has to be applied.
probednIIsin20
Published in SPIE Proceedings Vol. 10915, Organic Photonic Materials and Devices XXI, 1091503
(February 27, 2019) - DOI: 10.1117/12.2512039
Figure 1. (a) Spectra of the light intensity transmitted (T) by samples made of PS and printed by FDM, for various
thicknesses of the single layers. The values of T at fixed wavelength vs. the thickness of single printed layers is shown in
(b). The dashed line in (b) is a guide for the eye. Inset: photograph of a PC sample printed by FDM, highlighting achieved
transparency. Scale bar: 1 cm. (c) Spectral dependence of T on the printing speed. The corresponding values of T at fixed
wavelength as a function of the printing speed is shown in (d). The dashed line in (d) is a guide for the eye. The samples
used for measurements shown in (a)-(d) have area of 1×1 cm2 and thickness of 1 mm.
Figure 2. (a) Photographs of the optical windows printed by STL before (left image) and after (right image) the surface
finishing process. Scale bar: 1 cm. (b) UV-Vis spectra of T for samples printed by STL before (blue dashed line) and after
(red continuous line) the surface treatment. The samples used for the measurements are those shown in (a).
To explore the possibility of expanding the range of optical materials to photoactive systems, which can be shaped by 3D
printing, we investigated the properties of UV-polymerized layers containing photorefractive compounds. These layers
are made by spin-coating the E600/DR1 mixture on glass substrates, and curing them under UV light (10 mW/cm2) in a
(a)(b)(c)(d)(a)(b)
Published in SPIE Proceedings Vol. 10915, Organic Photonic Materials and Devices XXI, 1091503
(February 27, 2019) - DOI: 10.1117/12.2512039
nitrogen environment. In such systems, the embedded photo-isomerizable molecules undergo a series of trans-cis-trans
cycles, which leads to the formation of an anisotropic distribution of the molecule orientation, driven by the angular
redistribution of molecules due to the trans-cis conversion and rotational diffusion coming from thermal agitation.27,33
These effects are responsible of the build-up of optical anisotropy (dichroism and birefringence) in layers containing
azobenzene derivatives upon optical excitation with wavelength resonant with their absorption band. Figure 3a shows the
time evolution of the intensity of the probe beam transmitted by the sample and by crossed polarizers. The initial zero
signal increases in presence of the pump beam, reaching a saturation on timescales of hundreds of seconds. The all-
optical control of the probe is completely reversible, as shown in Figure 3a, highlighting a decay of the probe beam
intensity by turning off the pump. The signal build-up and decay is well described by a double exponential function, with
time constants of about 2 s for the fast component and 30-50 s for the slow one.
The optical anisotropy occurring in photo-responsive, nonlinear chromophores in polymer matrices has been investigated
in various host-guest systems, and most of these studies emphasized a complex dynamics characterizing both the build-
up and the decay regimes.27,34-38 A bi-exponential function is often used to account for the temporal evolution of the
observed rise and decay of optical nonlinearities.27,36 More specifically, for the relaxation process occurring when light is
turned off, the fast component is attributed to thermodynamically-activated cis-to-trans conformational changes, while
the slow one originates from angular re-distribution of the trans isomers, and the mobility and relaxation of the polymer
chains of the host. In other works,37,38 a stretched exponential function was found to describe the dynamics of the photo-
induced nonlinear properties, and a tight relationship between the mechanical/rheological properties of the polymeric
hosts and the temporal evolution of the photo-induced alignment of the guest molecules was also evidenced. As 3D
printed structures are going to be developed by embedding photoactive dopants in thermoplastic and photo-curable
polymers, the interplay between the matrix and the guest molecule will play a fundamental role for either enhancing or
depleting the ultimate nonlinear optical properties. This requires further experimental and theoretical investigation of
such effects, for rationalizing occurring phenomena and for engineering 3D printing processes, which can produce
complex structures with nonlinear optical properties, hopefully outperforming those observed in planar films.
Finally, the possibility of modulating light by light, namely the modulation of the probe beam by turning on and off the
pump by a mechanical chopper is shown in Figure 3b. Data show that the probe beam can be modulated at frequencies of
the order of hundreds of Hz, as observed typically for azobenzene derivatives in polymer and biopolymer matrices. 35 The
temporal profile of the modulated intensity of the probe signal reflects the complex dynamics of the photo-induced
alignment of the active molecules. Indeed, while the pump beam is switched on and off with ms characteristic times, the
probe signal follows an exponential trend with much longer characteristic times, as expected from the results shown in
Figure 3a. Recently, pyrazoline derivatives have been introduced which showing much faster photoswitching could be
used for enhancing the modulation frequencies of printed structures.28 Overall, the photoactive and photo-curable resins
here studied show optical properties suitable for all-optical control of light beams, that is promising for future
exploitation in 3D printing processes by STL.
Figure 3. (a) Temporal evolution of the intensity of the light transmitted by a layer of ES600 doped with DR1, upon
switching the pump laser beam with emission at 532 nm. The interval with pump laser 'on' is marked by vertical dashed
lines. The continuous lines are fit to the data by bi-exponential functions. (b) Examples of time evolution of a light beam
passing through a layer of ES600/DR1 and modulated at 25, 50 and 100 Hz by switching the pump beam (dashed line).
Published in SPIE Proceedings Vol. 10915, Organic Photonic Materials and Devices XXI, 1091503
(February 27, 2019) - DOI: 10.1117/12.2512039
4. CONCLUSIONS
In summary, optical windows have been printed by FDM and STL, and studied in their transparency. The investigation
of the dependence of the optical properties on printing parameters evidence a tight relationship. In particular, high
printing speeds and low layer thicknesses are suitable to increase the optical transmittance of structures printed by FDM,
while surface finishing by deposition of a film of uncured resin allows highly transparent sheets to be obtained by STL.
In addition, photo-curable sheets with light-responsive optical properties have been demonstrated, which allows the
intensity of a light beam to be modulated by an external beam at frequencies of hundreds of Hz. Such results are relevant
for the development of a novel class of 3D printed optical devices, whose properties can be varied in real-time by control
light beams.
Acknowledgments. The research leading to these results has received funding from the European Research Council
(ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement No. 682157,
"xPRINT").
REFERENCES
[1] Camposeo, A., Persano, L., Farsari, M. and Pisignano, D., "Additive manufacturing: applications and directions
in photonics and optoelectronics," Adv. Optical Mater. 7, 1800419 (2019).
[2] Shen, Y., Harris, N. C., Skirlo, S., Prabhu, M., Baehr-Jones, T., Hochberg, M., Sun, X., Zhao, S., Larochelle,
H., Englund, D. and Soljačić, M., "Deep learning with coherent nanophotonic circuits," Nat. Photon. 11, 441-
446 (2017).
[3] Kong, Y. L., Tamargo, I. A., Kim, H., Johnson, B. N., Gupta, M. K., Koh, T.-W., Chin, H.-A., Steingart, D. A.,
Rand, B. P. and McAlpine, M. C., "3D printed quantum dot light-emitting diodes," Nano Lett. 14, 7017 (2014).
[4] Park, S. H., Su, R., Jeong, J., Guo, S.-Z., Qiu, K., Joung, D., Meng, F. and McAlpine, M. C., "3D printed
polymer photodetectors," Adv. Mater. 30, 1803980 (2018).
[5] ISO/ASTM Standard 52900. Standard terminology for additive manufacturing-general principles. Part 1:
Terminology. ISO/ASTM International, Switzerland, 2015. https://www.iso.org/standard/69669.html
[6] Bourell, D. L., "Perspectives on additive manufacturing," Annu. Rev. Mater. Res. 46, 1-18 (2016).
[7] Turner, B. N., Strong, R. and Gold, S. A., "A review of melt extrusion additive manufacturing processes: I.
Process design and modeling," Rapid Prototyping J. 20, 192 (2014).
[8] Vak, D., Hwang, K., Faulks, A., Jung, Y.-S., Clark, N., Kim, D.-Y., Wilson, G. J. and Watkins, S. E., "3D
printer based slot-die coater as a lab-to-fab translation tool for solution-processed solar cells," Adv. Energy
Mater. 5, 1401539 (2015).
[9] Klippstein, H., Diaz De Cerio Sanchez, A., Hassanin, H., Zweiri, Y. and Seneviratne, L., "Fused deposition
modeling for unmanned aerial vehicles (UAVs): a review," Adv. Eng. Mater. 20, 1700552 (2018).
[10] Busch, S. F., Weidenbach, M., Fey, M, Schäfer, M., Probst, T. and Koch, M. "Optical properties of 3D printable
plastics in the THz regime and their application for 3D printed THz optics," J. Infrared Millimeter Terahertz
Waves 35, 993 (2014).
[11] Waheed, S., Cabot, J. M., Macdonald, N. P., Lewis, T., Guijt, R. M., Paull, B. and Breadmore, M. C., "3D
printed microfluidic devices: enablers and barriers," Lab Chip 16, 1993 (2016).
[12] Ota, H., Emaminejad, S., Gao, Y., Zhao, A., Wu, E., Challa, S., Chen, K., Fahad, H. M., Jha, A. K., Kiriya, D.,
Gao, W., Shiraki, H., Morioka, K., Ferguson, A. R., Healy, K. E., Davis, R. W. and Javey, A., "Application of
3D printing for smart objects with embedded electronic sensors and systems," Adv. Mater. Technol. 1, 1600013
(2016).
[13] Rimington, R. P., Capel, A. J., Christie, S. D. R. and Lewis, M. P." Biocompatible 3D printed polymers via
fused deposition modelling direct C2C12 cellular phenotype in vitro," Lab Chip 17, 2982-2993 (2017).
[14] Miller, A. T., Safranski, D. L., Smith, K. E., Sycks, D. G., Guldberg, R. E. and Gall, K.," Fatigue of injection
molded and 3D printed polycarbonate urethane in solution," Polymer 108, 121-134 (2017).
[15] Hull, C. W., (UVP, Inc.), U. S. patent 4,575,330, 1986.
[16] Kawata, S., Sun, H.-B., Tanaka, T. and Takada, K., "Finer features for functional microdevices," Nature 412,
697 (2001).
Published in SPIE Proceedings Vol. 10915, Organic Photonic Materials and Devices XXI, 1091503
(February 27, 2019) - DOI: 10.1117/12.2512039
[17] Chen, X., Liu, W., Dong, B., Lee, J., Ware, H. O. T., Zhang, H. F. and Sun, C., "High‐Speed 3D printing of
millimeter‐size customized aspheric imaging lenses with sub 7 nm surface roughness," Adv. Mater. 30,
1705683 (2018).
[18] Vaidya, N. and Solgaard, O., "3D printed optics with nanometer scale surface roughness," Microsyst. Nanoeng.
4, 18 (2018).
[19] Heinrich, A., Rank, M., Maillard, P., Suckow, A., Bauckhage, Y., Rössler, P., Lang, J., Shariff, F. and Pekrul, S.
"Additive manufacturing of optical components," Adv. Opt. Technol. 54, 293-301 (2016).
[20] Wu, S. L., Xia, H. B., Xu, J. H., Sun, X. Q. and Liu, X. G., "Manipulating luminescence of light emitters by
photonic crystals," Adv. Mater. 30, 1803362 (2018).
[21] Lenzini, F., Janousek, J., Thearle, O., Villa, M., Haylock, B., Kasture, S., Cui, L., Phan, H.-P., Dao, D. V.,
Yonezawa, H., Lam, P. K., Huntington, E. H. and Lobino, M., "Integrated photonic platform for quantum
information with continuous variables," Sci. Adv. 4, eaat9331 (2018).
[22] Wang, Q., Rogers, E. T. F., Gholipour, B., Wang, C.-M., Yuan, G., Teng, J. and Zheludev, N. I., "Optically
reconfigurable metasurfaces and photonic devices based on phase change materials," Nat. Photon. 10, 60-66
(2016).
[23] Eggleton, B. J., Luther-Davies, B. and Richardson, K., "Chalcogenide photonics," Nat. Photon. 5, 141-148
(2011).
[24] Bertrand, O. and Gohy, J.-F., "Photo-responsive polymers: synthesis and applications," Polym. Chem. 8, 52-73
(2017).
[25] Ubukata, T., Isoshima, T. and Hara, M., "Wavelength-programmable organic distributed-feedback laser based
on a photoassisted polymer-migration system," Adv. Mater. 17, 1630-1633 (2005).
[26] Goldenberg, L. M., Lisinetskii, V., Gritsai, Y., Stumpe, J. and Schrader, S., "Single step optical fabrication of a
DFB laser device in fluorescent azobenzene-containing materials," Adv. Mater. 24, 3339-3343 (2012).
[27] Marino, I. G., Bersani, D. and Lottici, P. P., "Photo-induced birefringence in DR1-doped sol-gel silica and
ORMOSILs thin flms," Opt. Mater. 15, 175-180 (2000).
[28] Szukalski, A., Moffa, M., Camposeo, A., Pisignano D. and Mysliwiec, J., "All-optical switching in dye-doped
DNA nanofibers," J. Mater. Chem. C 7, 170 (2019).
[29] Wu, L. Y. L., Zhao, Q., Huang, H. and Lim, R. J., "Sol-gel based photochromic coating for solar responsive
smart window," Surf. Coat. Technol. 320, 601-607 (2017).
[30] Sol, C., Schläfer, J., Parkin, I. P. and Papakonstantinou, I., "Mitigation of hysteresis due to a pseudo-
photochromic effect in thermochromic smart window coatings," Sci. Rep. 8, 13249 (2018).
[31] Jankovic, J., Ogle, B. R., Zontek, T. L., Biegalski, M. D., Hollenbeck, S. M. and Wells, T. M., "Suitability of
polycarbonate safety glasses for UV laser eye protection," J. Chem. Health Saf. 23, 29-33 (2016).
[32] Anitha, R., Arunachalam, S. and Radhakrishnan, P., "Critical parameters influencing the quality of prototypes
in fused deposition modeling," J. Mater. Process Technol. 118, 385-388 (2001).
[33] Cantatore, V., Granucci, G. and M. Persico, M., "The photo-orientation of azobenzene in viscous solutions,
simulated by a stochastic model," Phys. Chem. Chem. Phys. 16, 25081 (2014).
[34] Virkki, A., Kauranen, M. and Priimagi, A., "Different chromophore concentration dependence of photoinduced
birefringence and second-order susceptibility in all-optical poling," Appl. Phys. Lett. 99, 183309 (2011).
[35] Miniewicz, A., Kochalska, A., Mysliwiec, J., Samoc, A., Samoc, M. and Grote, J. G., "Deoxyribonucleic acid-
based photochromic material for fast dynamic holography," Appl. Phys. Lett. 91, 041118 (2007).
[36] Shimizu, F. M., Volpati, D., Giacometti, J. A., Sworakowski, J., Janus, K. and Luboch, E., "Kinetics of
photoinduced birefringence in the guest -- host system of poly(methyl methacrylate) doped with azobenzene-
containing crown ethers," J. Appl. Polym. Sci. 105, 130-136 (2007).
[37] Ribierre, J.-C., Mager, L., Fort, A. and Méry, S., "Effects of viscoelastic properties on the dielectric and
electrooptic responses of low-Tg guest-host polymers," Macromolecules 36, 2516-2525 (2003).
[38] Dhinojwala, A., Wong, G. K. and Torkelson, J. M., "Rotational reorientation dynamics of nonlinear optical
chromophores in rubbery and glassy polymers: α-relaxation dynamics probed by second harmonic generation
and dielectric relaxation," Macromolecules 26, 5943-5953 (1993).
|
1811.00510 | 1 | 1811 | 2018-11-01T17:29:59 | High speed imaging of solid needle and liquid micro-jet injections | [
"physics.app-ph"
] | High speed imaging was used to capture the fast dynamics of two injection methods. The first one and perhaps the oldest known, is based on solid needles and used for dermal pigmentation, or tattooing. The second, is a novel needle-free micro-jet injector based on thermocavitation. We performed injections in agarose gel skin surrogates, and studied both methods using ink formulations with different fluidic properties to understand better the end-point injection. Both methods were used to inject water and a glycerin-water mixture. Commercial inks were used with the tattoo machine and compared with the other liquids injected. The agarose gel was kept stationary or in motion at a constant speed, along a plane perpendicular to the needle. The agarose deformation process due to the solid needle injection was also studied. The advantages and limitations of both methods are discussed, and we conclude that micro-jet injection has better performance than solid injection when comparing several quantities for three different liquids, such as the energy and volumetric delivery efficiencies per injection, depth and width of penetrations. A newly defined dimensionless quantity, the penetration strength, is used to indicate potential excessive damage to skin surrogates. Needle-free methods, such as the micro-jet injector here presented, could reduce the environmental impact of used needles, and benefit the health of millions of people that use needles on a daily basis for medical and cosmetic use. | physics.app-ph | physics |
High speed imaging of solid needle and liquid micro-jet injections
Loreto Oyarte G´alvez1,∗ Maria Bri´o P´erez1, and David Fern´andez Rivas1
1Mesoscale Chemical Systems Group, MESA+ Institute and Faculty of Science and Technology,
University of Twente, P.O. Box 217, 7500 AE Enschede, Netherlands
High speed imaging was used to capture the fast dynamics of two injection methods. The first one
and perhaps the oldest known, is based on solid needles and used for dermal pigmentation, or tattoo-
ing. The second, is a novel needle-free micro-jet injector based on thermocavitation. We performed
injections in agarose gel skin surrogates, and studied both methods using ink formulations with
different fluidic properties to understand better the end-point injection. Both methods were used
to inject water and a glycerin-water mixture. Commercial inks were used with the tattoo machine
and compared with the other liquids injected. The agarose gel was kept stationary or in motion
at a constant speed, along a plane perpendicular to the needle. The agarose deformation process
due to the solid needle injection was also studied. The advantages and limitations of both methods
are discussed, and we conclude that micro-jet injection has better performance than solid injection
when comparing several quantities for three different liquids, such as the energy and volumetric
delivery efficiencies per injection, depth and width of penetrations. A newly defined dimensionless
quantity, the penetration strength, is used to indicate potential excessive damage to skin surrogates.
Needle-free methods, such as the micro-jet injector here presented, could reduce the environmental
impact of used needles, and benefit the health of millions of people that use needles on a daily basis
for medical and cosmetic use.
I.
INTRODUCTION
Tattooing, also known as dermal pigmentation, is done
by inserting exogenous substances such as pigments into
the dermis and leaving a permanent mark [1 -- 3]. The
earliest evidence of tattooing procedures traces back to
the fourth millennium BCE [4]. As evidenced by mum-
mified skin, ancient art, and the archaeological records,
tattoos have served two basic functions: medicinal or
cosmetic [5]. Two distinctive types of cosmetic tattoos
exist; the conventional or purely decorative, and those
intended to alleviate existing conditions and are defined
as permanent make-up, e.g. scar camouflaging, alopecia
or post-mastectomy pigmentation of a nipple on cancer
patients.
The societal acceptance of tattoos has varied over the
years, however, there is a recent worldwide increase in its
numbers and the social groups having tattoos or perma-
nent make-up. According to a recent report, 12% of the
European population has one or more tattoos [6]. This
corresponds to more than 44 million tattooed europeans,
while the figures in USA and other countries is supposed
to be similar or higher. This means that our society
as a whole will be posed with scientific and technologi-
cal challenges to reduce health risks caused by tattooing,
and palliate its economic consequences.
Despite the advances made in electronics and improved
hygienic conditions, the basic injection process has not
changed much. According to experts, depending on the
skin type and individual, a tattoo can be painful, cause
skin related allergies, while 20-50% of the ink is not in-
jected [7]. The method used for tattooing and permanent
make-up is in principle the same: the repeated injection
∗ [email protected]
of one or several needles into the skin delivers ink droplets
through the open wounds. The ink that adheres to the
needle surface is transported into the dermis of the skin,
as a function of the angle of the needle with respect to
the skin, and the pressure applied by the tattoo artist
or the cosmetic technician. Solid needles with single or
multiple tips are typically sold as single-use consumables.
The formulation of inks is kept as a secret by commercial
brands, but their ingredients can be roughly categorised
according to its function.
Inks are composed of differ-
ent pigments suspended in a carrier solution, together
with binders and additives[6]. Each ink formulation has
different ingredient proportions, conferring tailored fluid
dynamic properties, such as viscosity, surface tension and
density [7, 8]. Pigments are inorganic particles responsi-
ble for the ink colour tone, with a particle size range of
0.1 µm - 50 µm. The larger the particle, the least they
can be processed and removed by the immunological sys-
tem of the human body, which in turn guarantees the
permanence of the tattoo [9].
Less known is the fact that tattoo machines have been
adapted for intradermal injections to evenly inject into
a large area of the skin, dividing effectively the dose in
smaller portions [10]. During the last decades, alternative
cutaneous delivery methods have been developed, includ-
ing transdermal delivery injections at high pressures [11]
and needle-assisted jet injectors [12]. A liquid jet injector
is a needle-free medical device that pressurises thin liq-
uid filaments to penetrate the skin, with clear advantages
over conventional injections with needles [10, 13 -- 15]. Jet
injectors have helped in smallpox eradication, preventing
rabies, influenza, malaria, hepatitis A and B, injecting in-
sulin and analgesics, etc [10, 14 -- 18]. Some proven advan-
tages are higher immunological response, drug dose spar-
ing, reduction in pain with improved patient compliance,
and reduction of accidental needle-stick incidents [19].
To the best of our knowledge, a rigorous description
of the tattooing process -- one of the oldest transdermal
injection methods -- is presented for the first time. A com-
parison with a novel micro-jet injector device -- a needle-
free injection method [20] -- is performed on the basis of
quantification of different observables.
II. EXPERIMENTAL SETUP AND
PROCEDURE
A. Solid needle injector
A solid needle injector was vertically fixed in order to
inject liquid into a skin surrogate made of agarose gel,
as shown in figure 1 (a). The solid needle injector is a
conventional pigmentation instrument (PL-1000 Mobil,
Permanent Line GmbH). It is composed of an electric
control unit, a hand piece, and a consumable hygiene
needle module. The hand piece works with an encased
motor that moves the attached needle up and down in a
smooth, cyclical pattern [21]. The electric control unit
enables an adjustable injection frequency, with nominal
frequency values in the fn = [50 − 150] Hz range. The
needle nozzle consists of a disposable single-use module
attachable to the hand piece, which contains a sterilised
solid needle, made from stainless-steel. Needles with a
diameter of Dneedle=4 mm were used for this study.
FIG. 1.
(a) Schematics of the solid needle injector setup:
the hand-piece of the tattoo machine is vertically fixed with
a solid needle holder attached to it. The agarose gel skin
surrogates were located below and almost touching the needle
tip at rest. The agarose gel is kept stationary (vmotor = 0) or
in motion (vmotor = 2 mm/s), along a plane perpendicular to
the needle. A high-speed camera records the injection process
at 1000 fps. (b) Example of the images obtained with the
high-speed camera: the injection length L and diameter D
are determined from the front view, whereas the bottom view
shows the ink spread around the needle.
2
The agarose gel samples were placed inside a custom-
made holder, and the holder was located under the hand
piece with the surface of the gel almost touching the nee-
dle tip, as shown in figure 1 (a) and (b). The gel is
confined by glass transparent walls providing a depth (y-
axis), width (x-axis) and length (z-axis) of ∼3 mm, 3 mm
and 20 mm, respectively. A mirror with a 45◦ orientation
is placed below the gel. This system allows the simulta-
neous frontal and bottom visualisation of the injection
processes into the gels. The holder is attached to a one-
axis motorized translation stage (MT1/M-Z8, Thorlabs)
which allows to keep the gel stationary or in motion at a
constant speed vmotor = 2 mm/s, along a plane perpen-
dicular to the needle.
Front view images were obtained using a color high-
speed camera (Fastcam SA2, Photron) capturing 1000
frames per second; an example of the images acquired
is shown in figure 1 (b).
In addition, the agarose gel
deformation was measured with dry needles -- without
ink -- , using a monochromatic high-speed camera (Fast-
cam SA-X2, Photron) recording at 10000 frames per
second, and with a high-spatial resolution of 1409 pixels
per millimeter.
(i) Characterization of the needle displacement:
In order to characterize the needle vertical displace-
ment, we used the high-speed camera varying the nom-
inal frequency fn from 50 to 150 Hz. The videos were
recorded at 10000 frames per second, with a spatial res-
olution of 300 pixels per millimeter.
The tip position is obtained and plotted with respect
to time, as shown in figure 2 (a). During the first ∼ 0.15
seconds, the needle vertical displacement yneedle is the
same for every nominal frequency. A slow oscillation
is followed by a cyclical displacement with constant
frequency f0 = 122 ± 1 Hz and measured amplitude
am = 1.018 ± 0.006 mm. After this time, t (cid:38) 0.15 s, the
measured frequency fm reach an stable value directly
correlated to fn.
In figure 2 (b), we plot the power
spectral density of yneedle at t > 0.15 s for the 3 cases
shown in figure 2 (a), the position of the first peak cor-
responds to the measured frequency fm. In figure 2 (c),
we plot fm versus fn for all the experiments giving a
linear relation fm = 0.5fn + 42.
(ii) Injection force measurement:
In order to measure the force exerted by the solid nee-
dle injector into the agarose, we performed a separate
experiment where the injector and skin surrogate setup
were placed on a precision balance (Denver Instrument,
APX-1000, ∆m=0.1 mg), as shown in figure 3 (a). The
injector is switched on applying a normal force Fneedle
on the agarose which is recorded by the balance, indicat-
ing the measured effective mass meff. The force and the
effective mass are related by the equation
(1)
where (cid:126)g is the magnitude of the gravitational accelera-
Fneedle = meff(cid:126)g,
SkinsurrogateMirrorHandpieceNeedleInjectionprint(front)Injectionprint(bottom)NeedleholderBottom viewFront view 3
FIG. 3. (a) Schematic of the force measurement experimental
setup. The skin surrogate is placed on the precision balance
and the solid needle injector above it. (b) An initial peak mass
mrupture is measured, corresponding to the rupture of the gel
due to the needle penetration. After that, the balance mea-
sures an stable mass m∞, corresponding to the average mass
applied by the oscillatory acceleration of the needle yneedle.
(fn = 50 Hz), which corresponds to an applied normal
force: F rupture
F ∞
needle = 1.6 mN.
needle = 3.3 mN ,
B. Needle-free micro-jet injector
A continuous wave CW laser diode was focused at the
bottom surface of a glass microfluidic device, which was
partially filled with aqueous solutions of a molecular dye
matching the laser wavelength. The liquid heats up above
its boiling point in a few microseconds with an explosive
phase transition resulting in a fast growing vapour bub-
ble inside the microdevice; this phenomenon is known as
thermocavitation [22, 23]. The bubble pushes the liq-
uid forming a jet which in turn can penetrate into an
agarose gel located in front, as shown in figure 4. The
microfluidic devices used were similar to those described
elsewhere [20, 24].
FIG. 4. Schematics of the needle-free micro-jet injector setup:
A laser is focused at the bottom of a microfluidic device using
a microscope objective. As a result, the bubble and jet are
formed and are recorded using an ultrahigh-speed camera.
The liquid jet penetrates the skin surrogate located in front
of it. The image and the zoom-in insets show the agarose
gel holder and the injection print, respectively, using a red
colored glycerol-water mixture at 10%wt.
Microfluidic devices were designed and fabricated in
glass substrates under cleanroom conditions. Each device
has a fluidic chamber in which bubbles are created, and is
connected to a tapered channel with a 120 µm diameter
nozzle. For further details on this setup, the reader is
referred to Berrospe et al. [20]
FIG. 2. (a) Vertical position of the needle tip versus time,
for nominal frequencies fn =50 Hz, 70 Hz and 90 Hz. Dur-
ing the first ∼ 0.15 s, the needle oscillations are the same
in all cases. A slow oscillation is followed by a frequency
f0 (cid:39)122 Hz, after which a stable measured frequency fm is
reached in correspondence with the different nominal frequen-
cies. (b) The power spectral density of the needle y−position
for t (cid:38) 0.15 s is plotted, for nominal frequencies fn =50 Hz,
100 Hz and 150 Hz. The first peak corresponds to the mea-
sured frequency fm. (c) The measured frequency fm versus
the nominal frequency fn is also plotted. The fitted curve is
represented by the dashed-line and shows a linear dependence
fm ∝ 1/2fm.
tion.
When the solid needle starts to move, an initial peak
mass mrupture is measured. This mass corresponds to the
moment when the agarose gel is ruptured, i.e. when the
needle penetrates into the gel. After that, the measured
mass reaches a stable value m∞, corresponding to the av-
erage effective mass due to the the oscillatory movement
of the needle yneedle.
We have measured an effective mass mrupture = 0.34±
0.04 gr and m∞ = 0.16 ± 0.02 gr, for an agarose gel of
1%wt and needle displacement frequency fm = 70 Hz
-202-20200.050.10.150.20.250.3-2025010050100125751257510110210310-810-610-410-2Hand piece& NeedleSkin surrogatePrecision balanceLaserdiodeMicroscopeobjectiveMicrodeviceAgarose gelskin surrogateNozzle The laser diode, with a wavelength λ = 450 nm is
focused at the bottom of the device with a 10× micro-
scope objective. The spot has an elliptical shape, with
beam diameters rx = 33 µm and ry = 6 µm and variable
power P = 400-600 mW. The transparent glass walls of a
custom-made agarose holder permits the visualization of
injection processes, as shown in the inset of figure 4. The
agarose depth (y-axis), width (x-axis) and length (z-axis)
are fixed at ∼5 mm, 3 mm and 24 mm, respectively.
The bubble growth, the liquid jet formation and the
penetration into agarose slabs were recorded at ∼ 400000
frames per second using a ultrahigh-speed camera (Phan-
tom v2640). The camera sensor is protected from the
laser light using a colored glass filter centered at λ =
450 nm.
C. Liquid inks description
Three different liquid inks were used on the ex-
periments: commercially available Permanent make-up
(PMU) inks, red dyed water and red dyed glycerin-water
mixture at 10%wt. By adding glycerin, the injection pro-
cess, i.e. the adhesion to the needle tip and the diffusion
in the agarose, is modified with respect to pure water.
In order to maximize the absorbed energy by the liquid
from the focused laser, in the needle-free micro-jet injec-
tor, the aqueous solutions are coloured using a red dye
(Direct Red 81, CAS No. 2610-11-9) diluted at 0.5 %wt.
The PMU inks are colloidal dispersions with flow-
dependent viscoelastic properties. They are composed
mainly by water and glycerol, and extra additives such
as surfactants, solvents, binders and fillers [25, 26]. Pig-
ments provide colour and due to insolubility in water,
guarantee the permanent character of the injected ink.
Additives are used in order to avoid pigment sedimen-
tation while storage and help the re-dispersion of the
fluid after it is used. Microbiological contamination is
common in tattoos due to the high content of water an
organic substances present on inks.
In order to avoid
the contamination, preservatives are added to the mix-
ture. Other impurities that can be found on tattoos are
primary aromatic amines (PAA) and polycyclic aromatic
hydrocarbons (PAH) [6, 7]. In this study, we use two or-
ganic PMU inks: PMU-black (Amiea, Organic line, Deep
Black, MT. Derm) and PMU-red (Amiea, Organic line,
Cranberry, MT. Derm).
We performed rheology measurements of the liquid
inks. Specifically, we have measured the viscosity η vary-
ing the shear rate γ using a rheometer (Anton Paar
MCR502) with a cone-plate geometry (with diameter
d = 50 mm and cone angle α = 1◦ ). As shown in fig-
ure 5, the red colored water and the glycerin-water mix-
ture behave as Newtonian fluids with measured constant
viscosity ηwater = 0.9 mPa·s and ηglyc10% = 1.2 mPa·s.
On the contrary, the PMU inks show a shear thinning
behaviour, i.e. the viscosity of the fluid decreases when
the applied shear rate increases. The measured viscos-
γ (cid:38) 100 s−1, for the PMU-
ity at high shear rates,
black and PMU-red inks are ηPMU-black ∼ 300 mPa·s and
ηPMU-red ∼ 2000 mPa·s, respectively.
4
FIG. 5. The viscosity η versus shear rate γ of the liquid inks:
red dyed water (blue circles), red dyed glycerin-water 10%wt
(red squares), PMU-black ink (yellow triangles) and PMU-red
ink (purple diamonds).
The larger viscosity value of PMU inks is provided by
the high concentration of pigment particles, and other
additives not present in the colored water, and water-
glycerin solutions. PMU inks were not used on the
needle-free micro-jet injector due to visualisation limi-
tations caused by the light absorption of the pigments
that impeded the observation of thermocavitation.
D. Agarose gel skin surrogate preparation
Among the skin surrogates widely used to simulate the
mechanical properties of soft tissues, agarose is one of
the most used due to its transparency which allows the
optical quantification of injections [20, 27 -- 31]. The sur-
rogates were prepared by diluting agarose powder (Om-
niPur agarose, CAS No. 9012-36-6.), in deionised water,
with an agarose concentration of 1%wt. The solution was
heated up 45 seconds in microwave at full power. Once
the phantoms were prepared, they were cooled down at
room temperature for 5 minutes and stored at 4◦C.
III. SOLID NEEDLE INJECTION METHOD
In this section, we analyze the solid needle injector
method in three stages: micro-indentation, stationary
injection and moving injection. The micro-indentation
process is one where the needle pushes the skin surrogate
down without causing the rupture of the surrogate. The
stationary injection is when the needle start the injec-
tion of the ink into the surrogate, immediately after the
rupture occurs, keeping the gel fixed with respect to the
injector hand piece. Finally, the moving injection pro-
cess corresponds to a scenario closer to real life injection
101102100101102103104 conditions, in which the needle is moved perpendicular
to the skin surface.
A. Dynamic micro-indentation hardness test
A micro-indentation hardness test is used to describe
the hardness of a material to deformation with low ap-
plied loads. Recently, the interest to develop micro-
indentation testing to characterize skin surrogate hydro-
gels has grown [32 -- 37]. A specific indentation testing
configuration is the conical indenter, in which the inden-
ter is impressed into the surface of the agarose gel using
a known applied force, as shown in figure 6 (a).
FIG. 6. (a) A schematic representation of the conical inden-
tation test, where δ corresponds to the indenter tip position
respect to the gel surface and r is the radius of the impressed
cone volume. (b) Force relaxation in time during the indenta-
tion test. An initial instantaneous force F0 is exerted by the
gel opposing the indenting force, which directly relates to the
shear modulus G as F0 = 4Grδ.
During indentation early stages (t ≤ 1 ms), the water
in the agarose gel does not have time to flow away and
the gel behaves as an incompressible elastic solid. The
instantaneous force F0 correlates with the shear modulus
G of the gel [38] as
F0 = 4Grδ,
(2)
where δ corresponds to the indenter tip position with
respect to the gel surface and r is the radius of the im-
mersed indenter volume, as shown in figure 6 (a). After
this time, the solvent can flow away and the agarose gel
starts to relax, the force reaches a threshold value F∞,
and the gel behaves as a compressible elastic solid. The
qualitative force behaviour is plotted in figure 6 (b).
In our experiment, the needle solid injector can be
considered as a dynamic micro-indentation measuring in-
strument with a conical indenter. The solid needle trav-
els rapidly and does not allow the relaxation of the gel,
and as consequence, the measured force is the instan-
taneous force F0 in relation to the needle displacement.
The micro-indentation measurement will be valid before
the rupture of the surface agarose gel occurs, typically
before a millisecond. Furthermore, the tip of the solid
needle has a right circular conical geometry, as shown in
figure 7, with height and base radius htip = 0.174 mm
and rtip = 0.068 mm, respectively. The needle tip po-
sition δ, and the gel surface deformation are obtained
image sequences of the experiment.
5
FIG. 7. Image sequence of the needle solid injection process
before the skin surrogate rupture, the first needle-agarose con-
tact occurs at t=0 s. The needle tip has a right circular conical
surface, represented by the white-dashed line, with height and
base radius htip = 0.174 mm and rtip = 0.068 mm, respec-
tively (Multimedia view).
The gel deformation during indentation process is plot-
ted in figure 8 (a), where the zero position is defined in
the initial agarose gel surface plane, i.e. the surface be-
fore injection. We observe three stages in the surface
deformation: capillarity, indentation and injection. For
a time < 2 ms, when the needle is approaching, the first
contact is with a water film in the surface of the gel,
formed due to the environmental humidity and evapo-
ration. At the contact point, a liquid bridge wetting
the needle is created due to the capillary action, and
the visualized deformation is negative δ < 0. The liq-
uid bridge formation occurs too quickly to be captured
in greater detail by our high-speed camera, however this
phenomenon has been reported to play an important role
in, for example, nano- and micro-indentation and AFM
microscopy [39, 40].
After that, when yneedle ≥ 0 and t ≥ 0.2 ms, the nee-
dle pushes the agarose gel down and the dynamic micro-
indentation process starts. Finally, the force applied by
the needle against the agarose is high enough to induce
gel rupture, and the ink delivery, effectively starting the
injection process.
The maximum deformation δ caused by the needle tip
displacement is plotted in figure 8 (b), blue dots. The
expected needle position without the agarose gel yneedle
is taken from the calibration process described in sec-
tion II A, green line. As we expected, δ and yneedle are
in perfect agreement, which means the force exerted by
the agarose gel is negligible compared to that exerted by
the injector, meaning that the needle displacement is not
affected by the gel.
In the micro-indentation stage, the maximum deforma-
tion measured is δmax = 0.1062 mm which corresponds
to the applied force Frupture = 3.3 mN. We compare this
force with F0 as, Frupture = 3.3 mN = 4Gδmaxrmax = F0,
where rmax = rtipδmax/htip = 0.042 mm. We can cal-
culate the shear modulus G of the skin surrogate, which
represents the hardness, rigidity or stiffness of the ma-
6
a diameter ∼ 0.1 mm remains after one injection. After
the second injection, a darker region inside the hole was
observed, corresponding to a small water drop that came
out of the agarose gel.
FIG. 8. (a) Skin surrogate surface deformation during solid
needle micro-indentation. The maximum deformation just
before rupture δmax is shown. (b) Maximum deformation δ
and expected needle tip position yneedle versus time. The
three stages of the surface deformation (capillarity, micro-
indentation and rupture) are sketched and delimited.
terial, obtaining G = 185 mN/mm2 = 185 kPa. This
value matches the shear modulus range, ∼[30-3000] kPa,
reported in the literature for the agarose concentrations
used [41, 42].
B. Stationary injection
Stationary injections were performed to understand
the delivery process without the influence of specific fac-
tors, such as the needle injection angle, and translational
speed with respect to the skin surrogate. The injection
process in the case of a new needle holder loaded with
PMU-black ink, is shown in figure 9 (a). We observed
that it takes over 50 injections for the ink adhered to the
needle surface to slide down, and initiate the delivery
into the agarose; this instant is considered t = 0. After
that, another 50 injections are needed to make a spot-
width equivalent to the needle diameter 0.4 mm. Finally,
around 100 injections later, the injection width reaches
a threshold value of ∼ 0.8 mm. This plot shows that a
unique injection is not enough to deliver a dose equiv-
alent to the needle volume. Figure 9 (b) shows a high-
resolution image sequence of the agarose gel after four in-
jections without ink. The deformation of the agarose gel
surface after only four injections is clearly visible. Also
the gel is internally damaged, and a longitudinal hole of
FIG. 9. (a) Injected-ink width D(t) versus injection number.
A clean needle holder is loaded with PMU-black ink deliver-
ing ink into the agarose after ∼60 injections; this moment is
considered t = 0 (dashed line). Around 100 injections later,
the injection width reaches a threshold value ∼ 0.8 mm. Ev-
ery image shows the injection at the time corresponding to
the red point (Multimedia view). (b) High-resolution images
show the skin surrogate agarose gel after 4 injections. The
continuous deformation of the agarose gel surface is observed
after each frame. A longitudinal hole of a diameter ∼ 0.1 mm
remains in the agarose gel after every injection.
The injection process, from t=0, is compared for three
of the inks: PMU-black ink, glycerin-water mixture and
water. We calculate the ratio between the injection depth
and the immersed needle length L(t)/Lneedle and between
the injection width and the immersed needle diameter
D(t)/Dneedle, as shown in figure 10 (a) and (b). The to-
tal needle length is around 2 mm, but the total immersed
needle into the agarose can vary in dependence of the ini-
tial distance between the needle tip and the agarose sur-
face, mainly due to the agarose surface unevenness. We
observe that L(t)/Lneedle increases slower proportionally
to the ink viscosity, but after ∼ 40 injections they all
reach a threshold value L/Lneedle = 1.
In the case of D(t)/Dneedle, the spreading behaviour
is the opposite, for the PMU ink the change from zero
to one is almost instantaneous and continues to increase
up to D/Dneedle = 2 in around 100 injections. For the
aqueous solutions, D(t)/DNeedle saturates very quickly
-0.1-0.0500.05-0.15-0.1-0.0500.050.10.1510.80.60.40.2000.20.5010.1-0.15010015020000.20.40.60.8100-0.4-0.2 7
FIG. 10. (a) Injection depth and the immersed needle length ratio L(t)/Lneedle, (b) Injection width and the inmersed needle
diameter ratio D(t)/Dneedle and (c) Delivered volume V (t), versus time for three of the inks: PMU-black ink, glycerin-water
mixture and water. The image insets show the delivered dose at the same time ∼ 1 ms and/or the same injection number
∼ 75. The efficiency in volume delivery of PMU ink corresponds to the specific tailored properties for injection that the other
formulations do not have.
to 1, and the water and glycerin-water curves are almost
indistinguishable (yellow diamonds and green squares in
figure 10 (b)).
Assuming a circular conical injection shape, we esti-
mate the injected volume as
(cid:18) D(t)
(cid:19)2
V (t) =
1
3
π
2
L(t),
(3)
and plotted in figure 10 (c). The image insets show the
delivered volume at the same time ∼ 1 ms or the same
injection number ∼ 75. The plot shows that V (t) for
the PMU ink increases at least twice faster than for the
aqueous solutions, reaching volumes of ∼ 80 nl and ∼
20 nl respectively, as expected from the L(t)/Lneedle and
D(t)/Dneedle plots.
The large differences in behaviour are highly correlated
with the differences in viscosity. For the solid needle
used, with diameter Dneedle = 0.4 mm and an injection
frequency and amplitude fm = 74 Hz, am = 1 mm, the
shear rate of the ink film wetting the needle is
γ =
vneedle
Dneedle
=
am · 2πfm
Dneedle
= 1000
1
s
.
(4)
This means that the viscosities between the aqueous
and the PMU inks are two order of magnitude differ-
ent, from 1 mPa·s to 300 mPa·s. The PMU ink has a
high adherence to the needle, making necessary to clean
the needle in between experiments to prevent agglomer-
ation. Moreover, the images show that the residual -- not
injected -- ink accumulates in the surface, making the solid
needle injection method highly inefficient for low viscous
inks. Furthermore, there is remarkable difference of the
liquid adhesion onto the needle when the viscosity in-
creases from ηwater = 0.9 mPa·s to ηglyc10% = 1.2 mPa·s.
The images in figure 10 (c) show this difference, where
the amount of ink remaining in the agarose surface is
considerably higher for pure water than glycerin-water
mixture.
C. Moving injection
A qualitative characterisation of conventional tattoo
injection processes is performed by moving the transla-
tion stage at 2 mm/s orthogonal to the hand-piece. A to-
tal number of 50 injections, corresponding to t ∼ 700 ms,
are studied. Figure 11 shows a characteristic image se-
quence of the injection processes with different inks: (a)
PMU-black ink, (b) glycerin-water mixture and (c) wa-
ter. We observe that PMU ink has a much smaller
spreading into the agarose gel, which we attribute to
the several ingredients providing cohesion not present in
aqueous solutions, i.e. glycerin-water mixture and water.
In the case of the inks the surrogate saturation is faster
and the ink oozes out on top of the agarose surface. This
behaviour is expected because the agarose gel is mainly
composed of water, therefore a maximum spreading is
expected for the aqueous inks. Additionally, a residual
volume of PMU ink remains adhered to the needle dur-
ing the whole injection process. Pigment particles are
the main responsible of the higher viscosity and density
of PMU inks, giving such higher adhesion to the needle,
0.20.40.60.811.20500100015000.511.522.50002040608005001000150002550751000255075100 8
FIG. 11. Image sequences of the moving injection process for
three inks: (a) PMU-black ink, (b) glycerin-water mixture
and (c) water. The solid needle injector is initiated and the
agarose gel is moving at 2 mm/s, from right to left. The
spreading of the ink into the agarose is faster and wider for
aqueous solutions than the PMU ink, which has a viscosity
200 times higher (Multimedia view).
and also limit the spreading into the agarose. In prac-
tice, the tailored composition and resulting properties of
PMU ink, facilitate the drying process, and the needle
needs to be cleaned constantly to prevent agglomeration
as described in section II C.
Images of the injected skin surrogate were taken im-
mediately after the moving injection experiment. The
agarose gel was cleaned before the imaging with a
professional cleansing tonic (LaBina Aloe Vera tonic,
PERMANENT-Line),
in order to observe the post-
injection result at the front and bottom, as shown in
figure 12 for (a) PMU-red ink, (b) PMU-black ink, (c)
glycerin-water mixture and (d) water. For the PMU
inks, a well formed path line is visible in the front an
bottom views. For the glycerin-dyed water solutions,
the needle path is no longer visible due to its fast dif-
fusion. However, we observed that the glycerin-water
mixture presents less spread than pure water (see fig-
ure 11), in agreement with our observations described in
section III B. The bottom view shows that this spread
distributes asymmetrically around the needle, rather the
liquid spreads depending on the micro-characteristics of
the agarose gel, i.e. its porosity and inhomogeneities.
Those observations are quantified in figure 12 (e), the
intensity of the injection path observed in the mirrow
was calculated and averaged in the x direction. A zoom-
in of the mirrows for each ink is presented on the right.
Two well defined peaks for the PMU inks are observed in
the plot. The PMU-red peak is wider than the PMU-
FIG. 12. Side and top view of post-injection skin surrogate
for (a) PMU-red, (b) PMU-black, (c) glycerin-water mixture
and (d) water inks. (e) Average intensity of the injected path
observed in the mirrow.
black, because the red ink is one order of magnitude
more viscous, hence, the ink attaches better to the needle.
The aqueous inks are completely spread in the agarose,
showing an homogeneous average intensity. However, the
glycerin-water mixture has a higher intensity, indicating
that more ink has been delivered into the agarose.
IV. NEEDLE-FREE MICRO-JET INJECTION
METHOD
The needle-free micro-jet injector exhibited different
penetration characteristics from what we observed with
the solid needle injector. The needle-free injector creates
liquid micro-jets with a tip diameter Djet ∼ 50 µm and
a total ejected volume of ca. 50 nl, with a 100% of the
liquid delivered into the agarose for jet speeds larger than
∼ 40 m/s, which indicates no splash-back of liquid due
to the sufficient kinetic energy of the jets.
Figure 13 (a) shows image sequences of two successive
needle-free injections of water jets entering the agarose at
the same point. The penetration depths versus time for
both injections are plotted in figure 13 (b). Both micro-
jet speeds were 40 m/s before impacting the agarose. In
the first injection, the jet speed drops to 16 m/s, which
means a decrease of 85% in the micro-jet kinetic energy
and is capable to penetrate 1 mm into the agarose. After
the first jet opens a hole into the skin surrogate, each
subsequent jet follows a micrometric longitudinal orifice
into the gel. Therefore, the speed of the second micro-jet
drops less than the previous one, up to 20 m/s, with its
Avergage Intensity012345Distance mm-1-0.50.501 9
FIG. 13. (a) Two consecutive injections for pure water, with
jet speed 40 m/s. The red-dashed line corresponds to the
agarose gel surface. The cumulated ink from the previous
injection is visible into the gel (Multimedia view). (b) Pen-
etration depth versus time for the two continuous injections.
Before the micro-jet reaches the agarose surface both jets have
a speed of 40 m/s, with the impact, the first injection drops
its speed to 16 m/s while the second injection to 20 m/s,
penetrating around 40% deeper into the agarose.
kinetic energy decreasing 75% and allowing a penetration
into the agarose 40% deeper.
Experiments with glycerin-water mixture were per-
formed in order to compare the injection process with
the solid needle injection method. The shear rate experi-
enced by the liquid jets is at least two order of magnitude
higher than the shear rates provided by the solid needle
injector. An image sequence of a single injection with
micro-jet speed vjet ∼ 25 m/s, γ = 5·105 1/s, for pure wa-
ter and glycerin-water mixture is shown in figure 14 (a).
The differences observed in the agarose color and texture
is an optical effect due to small illumination differences.
It can be observed that, despite having the same initial
jet velocities, water jets penetrate deeper than glycerin
jets. The corresponding complete penetration events are
plotted in figure 14 (b).
We observed that inertial effects are more relevant in
the early stages of jetting and penetration, while viscous
FIG. 14.
(a) Image sequence of a single injection for pure
water and glycerin-water mixture. In both cases, the micro-
jet speed is vjet ∼ 25m/s and the skin surrogate is agarose gel
1%wt of concentration. (b) Penetration depth versus time for
the two injections. The pure water ink is able to penetrate
120% deeper than the glycerin-water mixture.
dissipation gains prominence towards the end-point in-
jection. The penetration speed of both jets right after
entering the agarose drops to a comparable speed, 8 m/s
and 7 m/s respectively. Afterwards, during jet deceler-
ation, water jets penetrate much deeper than glycerin-
water jets (120% increase) due to the differences in
the liquid viscosities of 25% (ηwater = 0.9 mPa·s and
ηglyc10% = 1.2 mPa·s). Though not visible in these fig-
ures, water jets spread (laterally) faster into the agarose
gel, while the spread area of glycerin jets remains almost
the same after each injection.
In contrast with the solid needle injector, no evident
damage was observed on the agarose surface after mul-
tiple injections recorded with an image resolution of 150
pixels per millimeter. Moreover, the agarose shows a ca-
pacity of self-recovery as the path created by the micro-
jet closes, and encapsulates the injected ink, see fig-
ure 13 (a) first image, second injection.
5010015020000-400-800-1200-16004005010015020025030035000-400-800-120040010002003000-500-10000-500-1000400 TABLE I. Comparison between solid needle and needle-free micro-jet methods. S is the strength ratio between the injection
pressure p and the skin surrogate shear modulus G = 185 kPa obtained in section III A. In the case of solid needle, the
penetration depth L, width D and volume V are quantified for a single injection and for the end point.
#1 injection
End point
10
Method
Ink
p [kPa] S = p/G K [µJ] energy % V0 [nl] L [mm] D [mm]
vol % L [mm] D [mm]
PMU-black
Glycerin-water
2400
13
Solid
needle
Water
Needle-free
micro-jet
Glycerin-water 3.2·105
3.1 · 105
Water
1.73
4.32
0.042
0.033
0.027
2.85
16.16
4.2·10−5
3.3·10−5
2.7·10−5
5.7·10−2
3.2·10−1
424
338
283
8.9
0.047
0.154
0.197
0.129
0.021
0.019
0.450
0.192
20.2
1.2
0.205
0.005
0.004
0.004
88.1
75.3
50(cid:80)
i=1
2(cid:80)
0.8
0.4
0.4
-
= 1.5
-
=1.65
0.205
V. COMPARISON OF INJECTION METHODS
The fact that no solid object is needed to rupture the
surrogate with the needle-free micro-jet injector is a clear
advantage in practical terms. To make a fair comparison,
the agarose gel surface rupture and damage, the spread-
ing of the ink, among other observations are detailed in
this section. The values calculated for all inks and both
injection methods are presented in table I, using the re-
sults in figure 10, for the solid needle, and figure 13 and
14, for the needle-free micro-jet.
The electric energy supplied to the injectors is calcu-
lated as E = P·t. The solid needle injector input power is
P = 7 W, and the time of a single injection is t = 1/70 s,
with a consumed energy E = 100 mJ. The needle-free
micro-jet injector input power corresponds to the laser
power Plaser = 0.5 W, and the corresponding time needed
for the liquid to cavitate t = 1 ms, corresponding to
E = 5 mJ. The kinetic energy transferred from the liquid
to the skin is calculated as K = 1/2mv2, where the speed
for the solid needle is vSN = am · 2πfm and for the jet is
vjet. Therefore, the injection efficiency in terms of energy
per injection is calculated as energy = K/E × 100 %. We
have calculated that the needle-free micro-jet injector has
a energy three orders of magnitude higher than the solid
injector.
The liquid volume deposited by the solid needle injec-
tor in each injection was estimated as the thickness of
the liquid film around the needle tip:
(cid:18) ηv
(cid:19)1/2
δ = c
ρg
,
(5)
where η is the liquid viscosity, ρ the liquid density, v the
flow velocity and g the gravitational acceleration. This
equation assumes that the flow velocity corresponds to
the needle velocity, and that the surface tension role is
negligible [43, 44]. The constant c is taken as 0.8, which
is a standard value for most Newtonian fluids. The ink
volume around the needle is estimated as the volume of
a hollow cylinder with inner radius as the needle radius
rinner = Dneedle/2 and outer radius router = Dneedle/2+δ.
i=1
The injection efficiency in terms of volume injected per
injection event, we calculate it as the ratio between the
deposited (solid needle) or ejected (needle-free micro-jet
injector) volume V0, and the volume remaining inside the
agarose gel. For the former case, this is defined as the
moment in which of solid needle retracts, and the latter
is after the hole in the agarose closes. This efficiency is
represented as vol = Vinj/V0 × 100 %. For the solid nee-
dle, the efficiency increases with each new injection be-
cause of the remaining ink from prior injections. There
is, however, excess of liquid on the top of the agarose gel
that does not penetrate at all, and contributes of the well
known ink loss of 50% in tattooing and permanent make
up procedures (personal communication with several tat-
toing and PMU companies). The vol is much higher for
the needle-free micro-jet injector than the solid needle:
five orders of magnitude.
In order to quantify the skin rupture and penetration
characteristics, we compare the local stress induced by
the solid needle and the jet impact, with a material-
dependent critical local stress as proposed before [17].
We define a new quantity, the penetration strength S, as
the ratio between the injection pressure p and the skin
surrogate shear modulus G = 185 kPa calculated in sec-
tion III A. In the case of the solid needle, the injection
pressure was calculated as the ration between the rupture
force Frupture and the needle cross sectional area:
p#1
needle =
Frupture
Aneedle
=
Frupture
πr2
max
,
(6)
where rmax = 0.042 mm is the radius of the inmersed nee-
dle volume at the moment of the maximum deformation
of the skin surrogate, as explained in section III A. The
pressure exerted by the jet onto the skin is pjet = 1
jet,
where ρ is the density of the liquid, and vjet the jet
speed [17].
2 ρv2
The penetration depth L and diameter D in the latest
stages or end-point of injection -- after 50 injections for
the solid needle, and after two injections for the needle-
free injectors -- show two interesting results. First, that
with only two jets is possible to reach the same depth
as with solid needles. Second, that the injection reso-
lution in spreading value (D), which corresponds to the
practical level of detail in tattooing, is half for the jet
injections. The current experimental setup for jet injec-
tions did not allow us to have the same reproducibility
of conditions to reach larger number of injections. Con-
trary to the solid needle injections, each new jet carries a
slightly higher velocity because of the reduction in liquid
inside the microfluidic chamber. In future studies we will
attempt to design a microfluidic device that ensures all
jets in a sequence are ejected with the same velocity and
total volume delivered.
VI. CONCLUSIONS
We consider that this study will advance the knowl-
edge of injection processes into soft substrates based on
solid needles and needle-free micro-jet injectors. Our
novel needle-free micro-jet injector employing thermocav-
itation is still in early development phases, however, the
results indicate that the injection outcomes are superior
to solid needle injectors in several aspects. Further work
is required, particularly aimed at increasing the repeata-
bility, frequency of injections, and overcoming challenges
related to the use of commercial inks with unknown in-
gredients. Our results indicate that:
• The power consumption of needle-free jet injec-
tions, 0.5 W, is an order of magnitude lower than
the solid injector (7 W). This has practical rele-
vance in future scenarios, where portability of injec-
tor devices is limited by the need of incorporating
heavy batteries.
• The calculated penetration strength S, reflects the
proportional final damaged state of the skin sur-
rogate. Values of S slightly above unity (jet in-
jections) seem to correspond to an effective pene-
tration with minimal damage, in contrast with ir-
reversible skin surrogate deformations when S∼10
(solid needle).
• The damage to the gel during the moving injec-
tions indicate that needle-free micro-jet injectors
could have less negative effect than solid needles
when injecting into skin.
In real-life conditions,
the displacement velocity and pressure applied with
a hand-piece can fluctuate, damaging the skin be-
cause of the hardness needles.
• The injection spreading increase with decreasing
ink viscosity. However, the volume and energy effi-
11
ciencies seem to increase proportionally to the vis-
cosity of inks.
• The injection efficiencies, energy and vol, are higher
for the needle-free micro-jet injector, with compa-
rable end-point injections. The needle-free micro-
jet injector in a single injection reaches penetra-
tion depths and widths comparable to ∼50 injec-
tions with a solid needle. Even though multiple in-
jections were not possible with the current needle-
free micro-jet injector, the higher efficiency values
demonstrate the superiority over the older solid
needle method. For needle-free micro-jet injectors
to compete with established injection methods in
real-life scenarios, however, future designs should
give multiple and reproducible injections, e.g. with
multiple nozzles or microfluidic geometries ensuring
self-refill of ink after each jet.
We conclude that needle-free micro-jet injections hold
potential to mitigate health problems associated to
needle-based injections, and avoid technical limitations
of solid needle injections in medical treatments, e.g. vac-
cines, scar camouflaging, alopecia, etc. The reduction of
environmental contamination with needle-free injections
could be huge, while benefiting millions of people using
needles on a daily basis for medical and cosmetic uses.
With a long list of challenges ahead, we believe there will
be a time, not far from now, when needle-free micro-jet
injectors will be reliable, safer, and efficient liquid deliv-
ery platforms, and where needles will not be that much
needed.
ACKNOWLEDGEMENTS
We would like to thank Stefan Schlautmann and Frans
Segerink for their technical support during fabrication
and optical setup construction. To Edgerton's centre for
the use of the Phantom high-speed camera and illumi-
nation. The material support of PERMANENT-Line
GmbH & Co. KG. and MT-Derm is kindly acknowl-
edged, as well as the practical and theoretical training
offered by PERMANENT-Line GmbH & Co. KG. DFR
acknowledges the recognition from the Royal Dutch Soci-
ety of Sciences (KHMW) that granted the Pieter Langer-
huizen Lambertuszoon Fonds, 2016.
REFERENCES
[1] K. Sperry, "Tattoos and tattooing. part i: History
and methodology.," The American journal of forensic
medicine and pathology, vol. 12, no. 4, pp. 313 -- 319, 1991.
[2] N. Kluger and V. Koljonen, "Tattoos, inks, and cancer,"
The lancet oncology, vol. 13, no. 4, pp. e161 -- e168, 2012.
[3] P. S. Islam, C. Chang, C. Selmi, E. Generali, A. Hunt-
ley, S. S. Teuber, and M. E. Gershwin, "Medical com-
plications of tattoos: a comprehensive review," Clinical
reviews in allergy & immunology, vol. 50, no. 2, pp. 273 --
286, 2016.
[4] R. Friedman, D. Antoine, S. Talamo, P. J. Reimer, J. H.
Taylor, B. Wills, and M. A. Mannino, "Natural mummies
from predynastic egypt reveal the world's earliest figu-
ral tattoos," Journal of Archaeological Science, vol. 92,
pp. 116 -- 125, 2018.
[5] J. Serup, N. Kluger, and W. Baumler, Tattooed skin and
health. Curr. Probl. Dermatol. Karger Medical and Sci-
entific Publishers, 2015.
[6] P. Piccinini, S. Pakalin, L. Contor, I. Bianchi, and
C. Senaldi, "Safety of tattoos and permanent make-
up. final report," JRC Science for policy report EUR,
vol. 27947, 2016.
[7] H. Petersen and K. Roth, "To tattoo or not to tattoo,"
Chemie in unserer Zeit, vol. 50, no. 1, pp. 44 -- 66.
[8] D. Jang, D. Kim, and J. Moon, "Influence of fluid physi-
cal properties on ink-jet printability," Langmuir, vol. 25,
no. 5, pp. 2629 -- 2635, 2009.
[9] C. A. Grant, P. C. Twigg, R. Baker, and D. J. Tobin,
"Tattoo ink nanoparticles in skin tissue and fibroblasts,"
Beilstein journal of nanotechnology, vol. 6, p. 1183, 2015.
[10] Y.-C. Kim, Skin Vaccination Methods: Gene Gun, Jet
Injector, Tattoo Vaccine, and Microneedle, pp. 485 -- 499.
Berlin, Heidelberg: Springer Berlin Heidelberg, 2017.
[11] A. Taberner, N. C. Hogan, and I. W. Hunter, "Needle-
free jet injection using real-time controlled linear lorentz-
force actuators," Medical engineering & physics, vol. 34,
no. 9, pp. 1228 -- 1235, 2012.
[12] X. Li, B. Ruddy, and A. Taberner, "Characterization of
needle-assisted jet injections," Journal of Controlled Re-
lease, vol. 243, pp. 195 -- 203, 2016.
[13] S. Mitragotri, "Current status and future prospects of
needle-free liquid jet injectors," Nat. Rev. Drug Discov.,
vol. 68, p. 341, 2006.
[14] A. Arora, Liquid and Powder Jet Injectors in Drug De-
livery: Mechanisms, Designs, and Applications, pp. 221 --
230. Berlin, Heidelberg: Springer Berlin Heidelberg,
2017.
[15] S. Munch, J. Wohlrab, and R. Neubert, "Dermal and
transdermal delivery of pharmaceutically relevant macro-
molecules," Eur. J. Pharm. Biopharm., vol. 119, pp. 235 --
242, 2017.
[16] J. Baxter and S. Mitragotri, "Jet-induced skin puncture
and its impact on needle-free jet injections: Experimental
studies and a predictive model," J. Controlled Release,
vol. 106, no. 3, pp. 361 -- 373, 2005.
[17] J. Schramm and S. Mitragotri, "Transdermal drug deliv-
ery by jet injectors: energetics of jet formation and pen-
etration," Pharm. Res., vol. 19, no. 11, pp. 1673 -- 1679,
2002.
[18] G. Nelmes, "The peace gun and the eradication of small-
pox," Military Medicine, vol. 182, no. 3-4, pp. 1512 -- 1513,
2017.
[19] N. C. Hogan, A. J. Taberner, L. A. Jones, and I. W.
Hunter, "Needle-free delivery of macromolecules through
the skin using controllable jet injectors," Expert Opin.
Drug Deliv., vol. 12, no. 10, pp. 1637 -- 1648, 2015.
[20] C. B. Rodr´ıguez, C. W. Visser, S. Schlautmann, D. F.
Rivas, and R. Ramos-Garcia, "Toward jet injection by
continuous-wave laser cavitation," Journal of biomedical
optics, vol. 22, no. 10, p. 105003, 2017.
12
[21] J. Serup, N. Kluger, and W. Baumler, Tattooed skin and
health. Tattoo Machines, Needles and Utilities. Karger
Medical and Scientific Publishers, 2015.
[22] S. F. Rastopov and A. T. Sukhodolsky, "Sound genera-
tion by thermocavitation-induced cw laser in solutions,"
in Optical Radiation Interaction with Matter, vol. 1440,
pp. 127 -- 135, International Society for Optics and Pho-
tonics, 1991.
[23] J. P. Padilla-Martinez,
C. Berrospe-Rodriguez,
G. Aguilar, J. C. Ramirez-San-Juan, and R. Ramos-
Garcia, "Optic cavitation with CW lasers: A review,"
Phys. Fluids, vol. 26, p. 122007, Dec. 2014.
[24] C. Berrospe-Rodriguez, C. W. Visser, S. Schlautmann,
R. Ramos-Garcia, and D. Fernandez Rivas, "Continuous-
wave laser generated jets for needle free applications,"
Biomicrofluidics, vol. 10, no. 1, p. 014104, 2016.
[25] H. Wijshoff, "The dynamics of the piezo inkjet printhead
operation," Phys. Rep., vol. 491, no. 4, pp. 77 -- 177, 2010.
[26] W. J. Grande, "Direct capillary printing in medical
device manufacture," Medical Coatings and Deposition
Technologies, pp. 309 -- 372, 2016.
[27] M. A. Kendall, "The delivery of particulate vaccines and
drugs to human skin with a practical, hand-held shock
tube-based system," Shock Waves, vol. 12, no. 1, pp. 23 --
30, 2002.
[28] Y. Deng, G. Winter, and J. Myschik, "Preparation and
validation of a skin model for the evaluation of intra-
dermal powder injection devices," Eur. J. Pharm. Bio-
pharm., vol. 81, no. 2, pp. 360 -- 368, 2012.
[29] J. Schramm-Baxter, J. Katrencik, and S. Mitragotri, "Jet
injection into polyacrylamide gels: investigation of jet in-
jection mechanics," J. Biomech., vol. 37, no. 8, pp. 1181 --
1188, 2004.
[30] R. Williams, Jet Injection of Viscous Fluids. PhD thesis,
ResearchSpace@ Auckland, 2016.
[31] Y. Tagawa, A. E. Oudalov, N.and Ghalbzouri, C. Sun,
and D. Lohse, "Needle-free injection into skin and soft
matter with highly focused microjets," Lab Chip, vol. 13,
pp. 1357 -- 1363, 2013.
[32] M. Ahearne, Y. Yang, A. J. El Haj, K. Y. Then, and K.-
K. Liu, "Characterizing the viscoelastic properties of thin
hydrogel-based constructs for tissue engineering applica-
tions," Journal of The Royal Society Interface, vol. 2,
no. 5, pp. 455 -- 463, 2005.
[33] Y. Hu, X. Zhao, J. J. Vlassak, and Z. Suo, "Using inden-
tation to characterize the poroelasticity of gels," Applied
Physics Letters, vol. 96, no. 12, p. 121904, 2010.
[34] D. M. Ebenstein and L. A. Pruitt, "Nanoindentation of
soft hydrated materials for application to vascular tis-
sues," Journal of Biomedical Materials Research Part A,
vol. 69A, no. 2, pp. 222 -- 232.
[35] G. Constantinides, Z. I. Kalcioglu, M. McFarland, J. F.
Smith, and K. J. V. Vliet, "Probing mechanical proper-
ties of fully hydrated gels and biological tissues," Journal
of Biomechanics, vol. 41, no. 15, pp. 3285 -- 3289, 2008.
[36] C.-Y. Hui, Y. Y. Lin, F.-C. Chuang, K. R. Shull, and W.-
C. Lin, "A contact mechanics method for characterizing
the elastic properties and permeability of gels," Journal
of Polymer Science Part B: Polymer Physics, vol. 44,
no. 2, pp. 359 -- 370.
[37] M. Galli, K. S. Comley, T. A. Shean, and M. L. Oyen,
"Viscoelastic and poroelastic mechanical characteriza-
tion of hydrated gels," Journal of Materials Research,
vol. 24, no. 3, p. 973979, 2009.
13
[38] K. L. Johnson, Contact Mechanics. Cambridge Univer-
738, 2000.
sity Press, 1985.
[39] S. Chen and A. Soh, "The capillary force in micro- and
nano-indentation with different indenter shapes," Inter-
national Journal of Solids and Structures, vol. 45, no. 10,
pp. 3122 -- 3137, 2008.
[40] Y. Men, X. Zhang, and W. Wang, "Capillary liquid
bridges in atomic force microscopy: Formation, rup-
ture, and hysteresis," The Journal of Chemical Physics,
vol. 131, no. 18, p. 184702, 2009.
[41] V. Normand, D. L. Lootens, E. Amici, K. P. Plucknett,
and P. Aymard, "New insight into agarose gel mechanical
properties," Biomacromolecules, vol. 1, no. 4, pp. 730 --
[42] V. Nayar, J. Weiland, C. Nelson, and A. Hodge, "Elas-
tic and viscoelastic characterization of agar," Journal of
the Mechanical Behavior of Biomedical Materials, vol. 7,
pp. 60 -- 68, 2012. 7th TMS Symposium on Biological
Materials Science.
[43] L. Landau and B. Levich, "Dragging of a liquid by a
moving plate," in Dynamics of Curved Fronts, pp. 141 --
153, San Diego: Academic Press, 1988.
[44] E. Rio and F. Boulogne, "Withdrawing a solid from a
bath: How much liquid is coated?," Advances in Colloid
and Interface Science, vol. 247, pp. 100 -- 114, 2017.
|
1705.00778 | 1 | 1705 | 2017-05-02T02:51:57 | Wavelength conversion of data at gigabit rates via nonlinear optics in an integrated micro-ring resonator | [
"physics.app-ph",
"physics.optics"
] | We present the first system penalty measurements for all-optical wavelength conversion in an integrated ring resonator. We achieve wavelength conversion over a range of 27.7nm in the C-band at 2.5 Gb/s by exploiting four wave mixing in a CMOS compatible, high index glass ring resonator at ~22 dBm average pump power, obtaining < 0.3 dB system penalty. | physics.app-ph | physics | Wavelength conversion of data at gigabit
rates via nonlinear optics in an integrated micro-
ring resonator
Alessia Pasquazi1, Raja Ahmad2, Martin Rochette2, Michael Lamont3,
Brent E. Little4, Sai T. Chu4 , Roberto Morandotti1, and David J. Moss1,3
1Ultrafast Optical Processing, INRS-EMT, Université du Québec,
1650 Blv. L. Boulet, Varennes, Québec J3X 1S2 Canada
2 McGill University, Dept. of Electrical and Computer Engineering, Montréal (PQ), H3A 2A7, Canada
3 CUDOS, School of Physics, University of Sydney, New South Wales 2006, Australia
4 Infinera Corp. 9020 Junction drive Annapolis, Maryland, 94089. USA
Abstract We present the first system penalty measurements for all-optical wavelength
conversion in an integrated ring resonator. We achieve wavelength conversion over a range of
27.7nm in the C-band at 2.5 Gb/s by exploiting four wave mixing in a CMOS compatible,
high index glass ring resonator at ~22 dBm average pump power, obtaining < 0.3 dB system
penalty.
1. Introduction
All-optical signal processing is recognized [1, 2] as being fundamental to meet the
exponentially growing global bandwidth demand and low energy requirements of ultra-high
bit rate communications systems. The possibility of exploiting ultra-fast optical nonlinearities
for the realization of critical signal processing functions has been widely explored in the last
two decades. Among the nonlinear phenomena of interest, of particular importance is four
wave mixing (FWM) via the Kerr nonlinearity (n2), as it can be used to perform frequency
conversion, signal reshaping, optical regeneration and other important all-optical functions [2-
6]. However, an optimal efficiency of this, and other, nonlinear optical phenomena still
requires significant improvement to ultimately achieve commercial deployment. Ring
resonators are seen as a key approach to enhancing the nonlinear efficiency [7] by effectively
recycling the optical pump power within the resonant cavity. The use of high Q-factor optical
resonators has enabled the demonstration of very low power continuous-wave (CW) nonlinear
optics based on pure silica microtoroids and microspheres [8-10], photonic crystal
nanocavities [11] or integrated micro-rings in high index glass [12,13] and silicon [14].
Similar benefits are expected [7] in terms of operation on optical signals containing high
bandwidth data, although the challenge in this case is that the bandwidth of the resonator must
be large enough to accommodate all of the spectral components of the optical signal. It is only
very recently [15] that the first demonstration of optical signal processing based on nonlinear
optics in a resonant cavity has been reported. In particular, wavelength conversion at 10 Gb/s
in silicon cascaded ring-resonator optical waveguide (CROW) devices has been [15] achieved.
However, full system penalty measurements have not yet been reported in these structures -
only in the context of FWM in straight waveguides and nanowires in silicon [16,17] and in
chalcogenide glass [18].
In this paper, we present the first full system penalty, or bit error ratio (BER),
measurements for wavelength conversion in an integrated ring resonator. We achieve near
error-free all optical wavelength conversion at 2.5 Gb/s in the C-band via four wave mixing
(FWM) in a high index doped silica glass ring resonator with a Q-factor of 65,000, a free-
spectral range (FSR) of 575 GHz and a full-width at half-maximum (FWHM) of 3 GHz. The
device is based on a CMOS compatible high index doped silica glass [12,13] that exhibits an
effective waveguide nonlinearity ( = n2 / c Aeff , where Aeff is the effective mode area, c is
the speed of light, and is the pump frequency) 200 times larger than that of standard single
mode fibers [7,8]. We achieve < 0.3 dB power penalty at 10-9 BER for converting a 2.5 Gb/s
signal from 1562 nm to 1535nm (a 27.7nm range), using 165 mW of CW pump power.
Further, because our waveguides exhibit negligible nonlinear saturation, or absorption, up to
extremely high intensities (25 GW/cm2) [19], we observe no saturation in the device
performance.
Fig.1 The four-port microring resonator used in the experiments. The trade-name for the
high index doped silica glass waveguide core material is Hydex®.
2. Experiment
The device under test consists of a four-port micro-ring resonator, 48 m in radius.
The guiding core is high index doped silica glass [12, 13] with a refractive index of 1.7 at
1550 nm, embedded in silica glass and integrated on a silicon wafer. Chemical vapor
deposition was used to deposit the glass layers, and the ring resonator was defined with high
resolution optical lithography followed by reactive ion etching. The dimensions (and
composition) of both the ring and bus are the same, with a rectangular cross section of
1.45 μm x 1.50 μm, yielding a tight modal field confinement due to a large refractive index
contrast. We use a vertical coupling scheme between the bus and ring in order to achieve
greater control over the gap. Details of the fabrication process are given in [12] and references
therein. Chromatic dispersion of this glass waveguide is relatively small with β2 -9 ps2/km at
1550 nm [20]. The ring resonator is coupled to two bus waveguides, as sketched in Fig.1,
Each of the four device ports (input, through, add, and drop) is pigtailed, with on-chip mode
converters to reduce coupling losses to 1.65 dB. The waveguides support a couple of very
weakly bound higher order modes with different symmetry, but they have no impact on the
device performance. The coupling scheme ensures that these modes are not excited and we
observe no resonances associated with them [12].
BPF=band-pass
filter,
Fig 2 a) Experimental set up for wavelength conversion of a 2.5Gb/s signal (BPF = bandpass
filter). DFB= distributed feedback laser, VA=variable attenuator, EDFA=erbium doped fiber
amplifier,
PD=photodiode,
MOD=modulator, ECL=edge couple laser. Note, we used an EDFA preamplifier rather than an
electronic preamplifier on the photodetector. The variable attenuator was used for the BER
measurements. b) Eye diagram of the pseudorandom signal at the input (left), before coupling
at the INPUT port, and as collected at the DROP port (right).
PC=polarization
controller,
Figure 2 shows the experimental setup. The pump was a CW distributed feedback (DFB)
tunable laser at = 1548.9 nm, amplified by an erbium doped fiber (EDFA). The signal was
obtained by modulating a CW tunable external cavity laser (ECL) at =1562.9 nm, also
amplified with an EDFA, at 2.5 Gb/s, using a pseudorandom bit sequence (PRBS, 215-1) in a
non return to zero (NRZ) format. We used a bit sequence length of 215-1 rather than 231-1 in
order to improve the stability of the BER measurements, given the relatively low bit rate of
2.5Gb/s. We have verified in other work at higher bit rates that this does not have a significant
impact on the results. The pump and signal were filtered using 0.9 nm and 0.8 nm bandwidth
filters (BPF), to remove the background amplified spontaneous noise (ASE), and were then
coupled into the ADD and INPUT ports, respectively. Two different resonances, offset by
three free spectral ranges (FSR) or 1.725 THz (13.8 nm), were excited by the pump and the
signal. The effective pump power inside the ring is estimated to be ~10W because of the
resonant enhancement of the cavity.
3. Results and discussion
The low power signal spectrum had a full width at half maximum of 10pm. The eye
diagrams of the signal as coupled into the INPUT port, and collected at the DROP port (output
of the sample) are shown in Fig. 2b Figure 2 shows that the signal eye diagram undergoes
only moderate degradation after passing through the ring, and so experiences very little high
frequency cut-off after passing through the ring resonance.
The spectra acquired in the presence of a 165 mW pump (coupled power at the ADD bus)
for several signal powers are visible in Fig. 3 (top left), measured at the DROP port of the ring
resonator. The linear growth in intensity of the generated idler at 1535.2 nm, versus signal
power (Fig. 3, bottom left) shows no saturation, and an internal average conversion efficiency
of -28.5 dB. The weak peaks visible in the spectrum of the pump at the DROP port are side-
modes of the DFB pump laser. The spectrum at the DROP port is not that of the incident
pump, but the residual spectrum left after the pump is coupled into the ring, and so the side-
modes lying outside the resonance are significantly enhanced relative to the central pump
wavelength.
Fig.3 Experimental results obtained with a 165mW pump at increasing signal powers (colors
from blue to red ). Top left: OSA spectra collected at the DROP port. Bottom left: idler power
versus signal power. Right: Eye diagrams corresponding to the idler in the right panel, acquired
with 1350 samples
We did not observe any intensity dependent saturation because of the negligible nonlinear
losses in the sample [19]. Note that comparable silicon-on-insulator (SOI) ring resonators
typically show saturation for efficiencies ~ -39 dB [14]. A thermo-optical nonlinearity
appeared at high pump powers level, however. We observed a slow red shift of the cavity
resonances of about 1 nm, consistent with an increase in the refractive index of 10-3. Thermo-
optical shifts can be conveniently employed to finely tune the resonance wavelengths,
obtaining a stable working point at the desired wavelength [13, 21-22].
To test the wavelength conversion performance of the ring resonator, the idler was
collected at the DROP output, filtered and amplified to measure the eye diagrams with a 65
GHz bandwidth sampling oscilloscope. The BER testing curves and eye diagrams are
presented in Fig. 4 for the converted idler and for the signal at the output of the device, for
fixed signal and power levels of 10mW and 165mW, respectively. Note that we used an
optical preamplifier (EDFA) in place of an electronic preamplifier on the photodiode, which is
why the received power levels are comparatively high. Both approaches yield equivalent
system penalties. The BER results associated with the idler signal show a remarkably low
system power penalty at 10-9 BER of < 0.3 dB. The idler eye diagram is smoothed slightly
with respect to the coupled signal (Fig.2b), indicating that the generated wavelength is not
exactly centered at the resonance. This accounts for part of the observed 0.3 dB penalty.
Figure 4 indicates that there may be an onset of a slight noise floor just above 10-9 BER,
although this is difficult to conclude definitively since the effect is comparable to the
experimental scatter in the data. We do not believe this is a fundamental limitation. The
quoted penalty of 0.3dB was obtained by comparing linear fits to the data. The lowest
experimental BER we achieve is within a factor of 2 of "error free". Figure 3 shows that the
quality of the idler eye diagrams improves significantly with increasing signal power. We
would therefore expect the BER penalty to be improved by using higher signal powers, which
was limited experimentally in our case to 10mW. This will be reported in future work.
This work demonstrates a practical performance benchmark for ring resonators and a
basis for nonlinear all-optical signal processing based on these devices. Following the
improvements in spectral efficiency expected from more advanced modulation formats and
from the use of lower Q resonators, as well as from the realization of higher order filter
designs to improve the bandwidth, we anticipate that the next generation of high-index
microring resonators will be capable of operation at bit rates of 40Gb/s and higher. This
platform has also demonstrated high efficiency parametric gain [23], supercontinuum
generation [24], and many other functions.
Fig.4 Left: BER measurement for the signal (black line) and the idler (red line) generated by a
10mW signal and a 165mW pump: the power penalty of the two curves is 0.3dBm Right: eye
diagram for the signal (top) and the idler (bottom)
4. Conclusions
We present the first system penalty measurements of all-optical wavelength conversion in
an integrated ring resonator. We achieve wavelength conversion in the C-band at 2.5Gb/s via
four wave mixing in a CMOS compatible high index silica glass ring resonator with a Q factor
of 65,000. The resulting system penalty of < 0.3dB paves the way for ring resonators to enable
low power, low bit error ratio operation for future network telecommunication systems.
Acknowledgements
We acknowledge financial support of the Natural Sciences and Engineering Research
Council of Canada (NSERC), of the FQRNT (Fonds Québécois de la Recherche sur la Nature
et les Technologies) as well as of the Australian Research Council (ARC).
References
1.
2.
3.
Nature Photonics Workshop on the Future of Optical Communications; Tokyo, Japan, Oct. 2007.
www.nature.com/nphoton/supplements/techconference2007
B.J.Eggleton, .D.J. Moss,., and S.Radic, Nonlinear Optics in Communications: From Crippling
Impairment to Ultrafast Tools Ch. 20 (Academic Press, Oxford, 2008).
R. Salem, M. A. Foster, A. C. Turner, D. F. Geraghty, M. Lipson, A. L. Gaeta "Signal regeneration using
low-power four-wave mixing on silicon chip". Nature Photonics 2, 35-38 (2008).
4. M. Pelusi, F. Luan, T. D. Vo, M. R. E. Lamont, S. J. Madden, D. A. Bulla, D.-Y. Choi, B. Luther-Davies,
5.
and B.J. Eggleton, et al, "Photonic-chip-based radio-frequency spectrum analyzer with terahertz
bandwidth", Nature Photonics, (2009). 3(3): p. 139.
V.G. Ta'eed, L.Fu, M.Rochette, I.C. M. Littler, D.J. Moss, B.J. Eggleton, "Error-Free All-Optical
Wavelength Conversion in Highly Nonlinear As2Se3 Chalcogenide Glass Fiber", Optics Express 14 10371
(2006).
6. M. A. Foster, R. Salem, D. F Geraghty, A, C Turner-Foster, M. Lipson and A. L Gaeta "Silicon-chip-
7.
8.
9.
based ultrafast optical oscilloscope". Nature, 456 7218, 2008.
A. Melloni, F. Morichetti, and M.Martinelli, "Four-wave mixing and wavelength conversion in coupled-
resonator optical waveguides," J. Opt. Soc. Am. B 25, C87-C97 (2008)
T.Carmon, and K.J.Vahala, "Visible continuous emission from a silica microphotonic device by third-
harmonic generation". Nature Physics 3, 430-435, (2007).
S.M.Spillane, T.J., Kippenberg, and K.J. Vahala, "Ultralow-threshold Raman laser using a spherical
dielectric microcavity". Nature 415, 621-623, (2002).
10. T. J. Kippenberg, S. M. Spillane, and K. J. Vahala "Kerr-Nonlinearity Optical Parametric Oscillation in an
Ultrahigh-Q Toroid Microcavity", Phys. Rev. Lett., 93, 8, (2004).
11. T. Tanabe, M. Notomi, S. Mitsugi, A. Shinya, and E.i Kuramochi, et al., "All-optical switches on a silicon
chip realized using photonic crystal nanocavities", Appl. Phys. Lett. 87, 151112, (2005).
12. M. Ferrera, L. Razzari, D. Duchesne, R. Morandotti, Z. Yang, M. Liscidini, J. E. Sipe, S. Chu, B. E. Little
and D. J. Moss, "Low-power continuous-wave nonlinear optics in doped silica glass integrated
waveguide structures", Nature Photonics 2, 737, (2008).
13. L.Razzari, D. Duchesne, M. Ferrera, R. Morandotti, S. Chu, B. E. Little and D. J. Moss, "CMOS-
14.
15.
compatible integrated optical hyper-parametric oscillator", Nature Photonics 4, doi:10.1038 /
nphoton.2009.236 (2010).
A. C. Turner, M. A. Foster, Al. L. Gaeta, and Michal Lipson, "Ultra-low power parametric frequency
conversion in a silicon microring resonator," Opt. Express 16, 4881-4887 (2008)
F. Morichetti A. Melloni, A. Canciamilla, C. Ferrari, M. Torregiani., "Phase preserving wavelength
conversion over 6 Thz in a silicon coupled resonator optical waveguide", Optical Fiber Communications
(OFC) Postdeadline Paper PDPA6, San Diego, March (2009).
16. W. Mathlouthi, H.Rong, and M. Paniccia, "Characterization of efficient wavelength conversion by four -
wave mixing in sub-micron silicon waveguides,"Optics Express 16 16735 (2008).
17. B.G. Lee et al., B. G. Lee, A. Biberman, A. C. Turner-Foster, M. A. Foster, M. Lipson, A. L. Gaeta, and K.
Bergman,"Demonstration of broadband wavelength conversion at 40 Gb/s in silicon waveguides,"Photon.
Technol. Lett. 21, 182-184 (2009).
18. M. D. Pelusi, V. G. Ta'eed, M. R. E. Lamont, S. Madden, D. Y. Choi, B. Luther-Davies, and B. J.
Eggleton, "Ultra-high Nonlinear As2S3 planar waveguide for 160-Gb/s optical time-division emultiplexing
by four-wave mixing," IEEE Photonics Technol. Lett. 19, 1496-1498 (2007).
19. D. Duchesne, M.Ferrera, L.Razzari, R.Morandotti, S.Chu, B.Little, and D. J. Moss, "Efficient self-phase
modulation in low loss, high index doped silica glass integrated waveguides", Optics Express 17 1865
(2009).
20. M.Ferrera, D.Duchesne, L.Razzari, M.Peccianti, R.Morandotti, P.Cheben, S.Janz, D.Xu, B.E Little, S.Chu
and D.J Moss, "Low Power CW Parametric Mixing in a Low Dispersion High Index Doped Silica Glass
Micro-Ring Resonator with Q-factor > 1 Million", Optics Express 17 pp. 14098–14103 (2009).
P. L. Hansen and P. Buchhave, ''Thermal self-frequency locking of a doubly resonant optical parametric
oscillator,'' Opt. Lett. 22, 1074–1076 (1997).
21.
22. A. Douillet, J.-J. Zondy, A. Yelisseyev, S. Lobanov, and L. Isaenko, ''Stability and frequency tuning of
thermally loaded continuous-wave AgGaS2 optical parametric oscillators,'' J. Opt. Soc. Am. B 16, 1481–
1495 (1999).
23. A.Pasquazi, M.Peccianti, M.Lamont, R.Morandotti, B.E Little, S.Chu and D.J Moss, "Efficient wa velength
conversion and net parametric gain via Four Wave Mixing in a high index doped silica waveguide", Optics
Express 18, (8) 7634-7641 (2010). DOI: 10.1364/OE.18.007634.
24. D.Duchesne, M.Peccianti, M.R.E.Lamont, M.Ferrera, L.Razzari, R.Morandotti, B.E Little, S.Chu and D.J
Moss, "Super-continuum generation in 45cm long spiral high index glass waveguide", Optics Express 18,
923-930 (2010). DOI: 10.1364/OE.18.000923.
|
1810.11830 | 1 | 1810 | 2018-10-28T16:32:44 | The cosine law of field enhancement factor variation: generic emitter shapes | [
"physics.app-ph",
"cond-mat.mes-hall",
"physics.acc-ph",
"physics.plasm-ph"
] | The cosine law of field enhancement factor variation was recently derived for a hemi-ellipsoidal emitter and numerically established for other smooth emitter shapes (Biswas et al, Ultramicroscopy, 185, 1 (2018)). An analytical derivation is provided here for general smooth vertical emitter shapes aligned in the direction of the asymptotic electrostatic field. The law is found to hold in the neighbourhood of the emitter apex from where field emisson pre-dominantly occurs. | physics.app-ph | physics |
The cosine law of field enhancement factor variation: generic emitter shapes
Debabrata Biswas,1, 2 Gaurav Singh,1, 2 and Rajasree Ramachandran1
1)Bhabha Atomic Research Centre, Mumbai 400 085, INDIA
2)Homi Bhabha National Institute, Mumbai 400 094, INDIA
The cosine law of field enhancement factor variation was recently derived for a hemi-ellipsoidal emitter and
numerically established for other smooth emitter shapes (Biswas et al, Ultramicroscopy, 185, 1 (2018)). An
analytical derivation is provided here for general smooth vertical emitter shapes aligned in the direction of
the asymptotic electrostatic field. The law is found to hold in the neighbourhood of the emitter apex from
where field emission pre-dominantly occurs.
I.
INTRODUCTION
Field emitters finds application in various vacuum na-
noelectronics devices where a cold, bright source of elec-
trons is required. They generally involve ultrasharp emit-
ting tips with apex radius of curvature in the nanometer
regime. This leads to enhancement of the local field on
the emitter surface so that electric fields of the order of
V/nm are generated even at moderate applied voltages.
A measure of local field enhancement is the apex field
enhancement factor. It refers to the ratio of the magni-
tude of the local electric field at an emitter apex to the
asymptotic electric field away from the cathode plane1 -- 7.
While this is an important quantity in field emission
theory8 -- 13 and a topic of considerable research, a calcula-
tion of the net emission current also requires knowledge
about the variation of the enhancement factor close to
the emitter apex3,4,14 -- 16. In Ref. [15], it was shown that
the field enhancement factor γ at any point close to the
apex of an axially symmetric emitter, is related to the
apex field enhancement factor γa by
where
γ = γa cos θ
(cid:112)(z/h)2 + (ρ/Ra)2
z/h
,
cos θ =
(1)
(2)
h is the height of the emitter, Ra its apex radius of cur-
vature and (ρ, z) is a point close to the emitter apex
(ρ = 0, z = h). The underlying assumption is that the
emitter is aligned along the asymptotic electrostatic field
E0 z. Eq. 1 was established analytically15 for a hemi-
ellipsoid emitter and numerically found to be true for
other emitter shapes including a cone and a cylindrical
post with a parabolic cap.
In the following, we shall establish Eq. 1 analytically
for general emitter shapes starting from the line charge
model5,17,18.
electric field −E0 z (see Fig. 1). Alternately, consider
a parallel-plate geometry with the two plates separated
by a distance D, the emitter mounted on a grounded
cathode and the anode at a positive potential V . If D is
large compared to the height h of emitter, the anode has
negligible effect on the emitter apex and the asymptotic
field E0 (cid:39) V /D.
The termination of the field lines on the cathode sur-
face gives rise to a surface charge density σ(ρ, z), which in
turn can be projected on the emitter axis as a line charge
density, Λ(z). For the hemi-ellipsoid in an asymptotic
field E0, it is known that Λ(z) = λz while in general, the
surface and line charge densities are related as18
Λ(z) = 2πρ(z)(cid:112)1 + (dρ/dz)2σ(z)
(3)
FIG. 1. An emitter mounted on a metallic cathode plate in
the presence of an asymptotic field −E0 z. The emitter is
equivalently modeled as a line charge distribution (bold line).
II. A MODEL FOR GENERAL EMITTER SHAPES
Consider an emitter mounted on an infinite metallic
cathode plate, aligned in the direction of the asymptotic
where ρ = ρ(z) defines the surface of the axially sym-
metric emitter.
In general, Λ(z) is expected to be a
smooth nonlinear function of z with the nonlinear terms
depending on the excursion from the hemi-ellipsoidal
shape. Note that close to the emitter base, the field lines
RaρE0z(ρ,z)hzL 2
A. Linear line charge density
We shall first deal with linear line charge density
Λ(z) = λz where λ is a constant which can be evaluated
using V (0, h) = 0. Thus,
λ = −
h ln
4π0E0h
(cid:16) h+L
h−L
(cid:17) − 2L
(8)
where L is related to the height h and apex radius of cur-
vature Ra through the relation (h2 − L2)/h = Ra which
holds for linear as well as nonlinear charge distributions6.
The field components can be calculated by differentiat-
ing Eq. 4 and the integrals are easy to evaluate when the
line charge density is linear. The methods employed are
however sufficiently general to be of use in the nonlinear
case as well6.
The ρ component of the electric field, -∂V /∂ρ can be
evaluated to yield
(cid:34)
Eρ =
λ
4π0
1
ρ
(cid:112)ρ2 + (z + L)2
ρ2 + z(z + L)
(cid:112)ρ2 + (z − L)2
− ρ2 + z(z − L)
(cid:35)
(9)
while the z component of the field is
λ
4π0
ln
L
+
(cid:34)
(cid:112)ρ2 + (z + L)2
(cid:110)(cid:112)ρ2 + (z + L)2 − (z + L)
(cid:112)ρ2 + (z − L)2 − (z − L)
(cid:34)(cid:0)z +
(cid:0)z +
(cid:1)(cid:0)1 −
(cid:1)(cid:0)1 −
z + L
1
ρ
ρ2
ρ2
z − L
Eρ (cid:39) λ
4π0
For small values of ρ near the tip,
which, on further simplification yields
L
(cid:112)ρ2 + (z − L)2
(cid:111)(cid:35)
− E0.
(cid:1) −
(cid:1)(cid:35)
ρ2
2(z + L)2
ρ2
2(z − L)2
(cid:34)
Eρ (cid:39) λρ
4π0
2z2L
(z2 − L2)2 −
2L
(z2 − L2)
+
(10)
(11)
(12)
(cid:35)
.
are expected to terminate at the cathode plate so that
Λ(z) → 0 for z → 0. A convenient form for the line
charge density of such a vertically aligned emitter is thus
Λ(z) = zf (z) where f (z) depends on the emitter shape
and is otherwise unknown. We refer to this as the line
charge model for a generic emitter shape.
The potential at any point (ρ, z) due to a vertical line
charge placed on a grounded conducting plane can be
expressed as
Λ(s)
(cid:104)(cid:90) L
(cid:2)ρ2 + (z + s)2(cid:3)1/2
(cid:2)ρ2 + (z − s)2(cid:3)1/2
(cid:105)
Λ(s)
ds
0
ds −
(4)
+ E0z
V (ρ, z) =
1
4π0
(cid:90) L
0
where L is the extent of the line charge distribution. Note
that the second integral arises from the image of the line
charge distribution. The zero-potential contour corre-
sponds to the surface of the desired emitter shape so that
the parameters defining the line charge distribution in-
cluding its extent L, can, in principle be calculated by
imposing the requirement that the potential should van-
ish on the surface of the emitter.
Starting with Eq. 4, we shall investigate the cosine law
of Eq. 1 for a general sharp emitter with h >> Ra.
The surface of a vertically aligned emitter z = z(ρ) can
be expressed in the neighbourhood of the apex by Taylor
expanding z at ρ = 0. Since h = z(0) and dz/dρ = 0 for
a vertically aligned emitter,
z (cid:39) h − ρ2
2Ra
(5)
in the immediate vicinity of the apex from where field
emission occurs. The electric field lines are normal to
this surface and thus in the direction
1(cid:112)1 + (ρ/Ra)2
n =
(
ρ
Ra
ρ, z).
(6)
It is thus necessary to establish that (cid:126)E = (cid:126)E.n = Ea cos θ
in order that the cosine law is valid for a general Λ(z)
where
(cid:126)E = Eρ ρ + Ez z = −(cid:16) ∂V
(cid:17)
ρ +
∂V
∂z
z
∂ρ
and Ea is the field at the apex. Alternately, it can be
shown that (cid:126)E =
z = Ea cos θ. We shall pursue
both approaches in deriving the cosine law for a sharp
emitter.
ρ + E2
E2
(cid:113)
III. THE ELECTRIC FIELD COMPONENTS ON THE
SURFACE AND THE COSINE LAW
Ez =
Since z ∼ h and z2 − L2 ∼ hRa, the first term dominates
for a sharp emitter. Thus,
(7)
Eρ (cid:39) λ
4π0
2zL
(z2 − L2)
zρ
(z2 − L2)
(13)
for small ρ.
(cid:113)
E2
ρ + E2
z
E =
(cid:39) λ
4π0
2zL
(z2 − L2)
(cid:34)
(14)
−
ρ2
(z2 − L2)2 (z2 + 3L2)
3
(21)
(22)
(23)
The z component of the field can be similarly simpli-
fied. Neglecting the logarithmic term and E0 for a sharp
emitter, Eq. 10 can be expressed as
(cid:34)
Ez (cid:39) λ
4π0
L
z − L
ρ2
2(z − L)2
L
L + z
ρ2
2(z + L)2
(cid:1)+
(cid:1)(cid:35)
(cid:0)1 −
(cid:0)1 −
(cid:34)
which can be further simplified as
(cid:35)
.
(15)
Ez (cid:39) λ
4π0
2zL
(z2 − L2)
1 − ρ2(z2 + 3L2)
2(z2 − L2)2
We are now in a position to establish the cosine law
using (cid:126)E.n or E =
z . Note that on the surface
of the emitter close to the apex, ρ and z are related by
z = h − ρ2/(2R2
(cid:113)
a). Also, L =(cid:112)h(h − Ra) (cid:39) h and
ρ + E2
E2
2zL
z2 − L2 (cid:39) 2h
Ra
(1 +
ρ2
R2
a
)
so that
Eρ (cid:39) λ
4π0
Ez (cid:39) λ
4π0
2h
Ra
2h
Ra
(1 +
(1 +
ρ2
R2
a
ρ2
R2
a
(cid:104)
ρ
Ra
1 − 2
)
)
ρ2
R2
a
(cid:105)
(16)
(17)
Thus, close to the apex, using the relation between z
and ρ on the emitter surface,
(cid:126)E.n =
λ
4π0
2h
Ra
(cid:113)
)
(1 + ρ2
R2
a
(1 + ρ2
R2
a
(cid:34)
)
ρ2
R2
a
+ 1 − 2ρ2
R2
a
(cid:35)
.
(18)
Since, we are interested in the variation close to the apex
(ρ small),
(1 − ρ4/R2
a)
(cid:112)1 + ρ2/R2
a
(cid:126)E.n = γaE0
(cid:112)1 + ρ2/R2
γaE0
a
(cid:39)
(cid:39) γaE0 cos θ
(19)
(20)
where we have used (λ/4π0)(2h/Ra) = Ea = γaE0. Al-
ternately,
(z2 − L2)2 + 1
z2ρ2
(cid:35)1/2
(cid:34)
(cid:35)1/2
(cid:17)
)
ρ2
R2
a
=
λ
4π0
2zL
(z2 − L2)
1 − 3L2ρ2
(z2 − L2)2
which approximates to
(cid:16)
(cid:16)
E (cid:39) λ
4π0
(cid:39) λ
4π0
2h
Ra
2h
Ra
1 +
1 +
ρ2
R2
a
ρ2
R2
a
(cid:17)(cid:16)
(cid:17)(cid:16)
1 − 3ρ2
2R2
a
1 − 3ρ2
2R2
a
(1 + 2
(cid:17)
.
This can be further simplified to yield
(cid:17)
(cid:16)
(cid:112)1 + ρ2/R2
2h
Ra
γaE0
a
E (cid:39) λ
4π0
(cid:39)
ρ2
R2
a
1 − 1
2
(cid:39) γaE0 cos θ.
Thus, the cosine law of field enhancement factor has been
shown to hold close to the apex from where emission
predominantly occurs.
B. Nonlinear Line Charge Density
As discussed earlier, the line charge density may be
assumed to be of the form Λ(z) = zf (z) without any loss
of generality. As in the linear case, we shall first derive
expressions for Eρ and Ez close to the apex (ρ small).
The Eρ component can be expressed as
= − ρ
4π0
(cid:90) L
[ρ2 + (z + s)2]3/2
0
sf (s)
(cid:35)
sf (s)
[ρ2 + (z − s)2]3/2
ds
0
Eρ = − ∂V
∂ρ
For ρ small,
(cid:34)(cid:90) L
0
(cid:104)
1 −
Eρ = − ρ
4π0
(cid:90) L
0
sf (s)
(z + s)3
ds
(cid:104)
ds
sf (s)
(z − s)3
1 −
3ρ2
2(z − s)2
3ρ2
2(z + s)2
(cid:105)(cid:35)
(25)
.
ds−
(24)
(cid:105)−
(cid:34)(cid:90) L
Note that as before z (cid:39) h. The expansion in the second
integral is justified since z2 − L2 = hRa implies z − L (cid:39)
Ra/2. Since z − s is bounded from below by z − L, the
expansion holds so long as ρ < Ra/2.
We shall first deal with the leading terms in the inte-
(cid:35)
(cid:35)L
(cid:34)
0
ds
+
(cid:90) L
(cid:90) L
0
sf (s)
(z + s)3 − sf (s)
(cid:34)
(z − s)3
= −
f (s)
2s3
(z2 − s2)2
= − f (L)
2L3
(z2 − L2)2
1 −
ds
0
f(cid:48)(s)
2s3
(z2 − s2)2 ds
f(cid:48)(s)
f (L)
s3/(z2 − s2)2
L3/(z2 − L2)2
(26)
(cid:35)
(cid:34)
gral:
(cid:90) L
0
Thus,
where
Eρ = f (L)
2L2
(z2 − L2)
Lρ
z2 − L2
(cid:104)
1 − C0
(cid:105)
(27)
(cid:90) L
0
C0 =
f(cid:48)(s)
f (L)
s3/(z2 − s2)2
L3/(z2 − L2)2 .
ds
(28)
where
Since the charge distribution is well behaved for a smooth
emitter and can be expressed as a polynomial function
of degree n (for cases of interest here, n ≤ 5), f (s) obeys
Bernstein's inequality19
f(cid:48)(x) ≤
n
(cid:107)f(cid:107)
(1 − x2)1/2
(29)
where x ∈ [−1, 1] and (cid:107)f(cid:107) denotes the maximum value
of f in this interval. With x = s/h and applying the
inequality, it can be shown that C0 ∼ (z2 − L2)1/2 is
vanishingly small for sharp-tipped emitters since z (cid:39) h.
The dominant contribution to Eρ is
Eρ (cid:39) f (L)
4π0
2zL
(z2 − L2)
zρ
(z2 − L2)
(30)
Note that the ρ2 correction terms in the integrand of
Eq. 25 can be neglected for small ρ. Thus Eq. 30 is the
final expression for Eρ close to the apex.
The z component of the electric field can be similarly
evaluated. Thus,
(cid:34)
(cid:90) L
0
−
ds
(z − s)sf (s)
[ρ2 + (z − s)2]3/2
(cid:35)
ds
(z + s)sf (s)
[ρ2 + (z + s)2]3/2
Ez = − ∂V
∂z
= − 1
4π0
(cid:90) L
+
0
For ρ small (ρ < Ra/2 as mentioned earlier),
(cid:34)(cid:90) L
0
(cid:110)
(cid:110)
ds sf (s)
1
(z + s)2 −
(z − s)2
ds sf (s)
1
(z + s)4 −
1
(z − s)4
Ez = − 1
4π0
(cid:90) L
0
3
2
ρ2
4
(cid:111)−
− E0
(32)
1
(cid:111)(cid:35)
We shall first consider the integral
(cid:111)
(cid:35)L
0
f (s)
1
(cid:111)
(cid:90) L
1
(z + s)2 −
(z − s)2
(cid:90) L
0
=
ds sf (s)
(cid:110)
(cid:34)(cid:110) − 2zs
(cid:90) L
(z2 − s2)
dsf(cid:48)(s)
+
0
= − 2zL
(cid:39) − 2zL
z2 − L2 f (L)
z2 − L2 f (L)
2zs
+ ln
z + s
z − s
(cid:104)
(cid:105)
z2 − s2 −
(cid:105)
(cid:104)
1 − C1
1 − C1
f(cid:48)(s) ln
0
+ ln
z + s
z − s
(cid:104)
(33)
z + s
z − s
1 − C2
(cid:105)
(cid:90) L
(cid:90) L
0
C1 =
C2 =
0
ds
f(cid:48)(s)
f (L)
f(cid:48)(s)
f (L)
s/(z2 − s2)
L/(z2 − L2)
ln
(cid:16) z+s
(cid:16) z+L
(cid:17)
(cid:17) ds.
z−s
ln
z−L
(34)
(35)
(36)
(37)
emitter (O(cid:0)z2−L2)1/2(cid:1) but C2 cannot be neglected due to
Using Bernstein's inequality, C1 is negligible for a sharp
z2−L2 >> ln z+L
the logarithmic dependence. However,
z−L
for a sharp emitter so that the last line of Eq. 33 follows.
2zL
Consider now the integral
(cid:90) L
0
ds
(cid:110) sf (s)
(z + s)4 − sf (s)
(z − s)4
(cid:111)
.
Using the methods adopted earlier, this reduces to
(cid:104) − 8
3
(cid:105)L
+
0
s3z
f (s)
(cid:90) L
s3z
(z2 − s2)3
f(cid:48)(s)
8
3
0
= − 8
3
f (L)
(z2 − s2)3 ds
L3z
(cid:104)
(z2 − L2)3
1 − C0
(cid:105)
+ E0
It follows from Bernstein's inequality that C0 ∼ (z2 −
L2)1/2 which is vanishingly small. Thus,
(31)
(cid:34)
1 − ρ2
2
4L2
(z2 − L2)2
(cid:35)
.
(38)
Ez (cid:39) f (L)
4π0
2zL
z2 − L2
Since z (cid:39) L close to the emitter apex, both Eρ and Ez
(Eqns. 30 and 38 respectively) have the same form as in
the linear case derived earlier (Eqns. 13 and 15 respec-
tively) with f (L) replacing λ. Note also that the linearly
varying line charge density is a special case of the non-
linear density with f(cid:48)(s) = 0 so that the correction terms
C0,C1,C2 are identically equal to zero instead of being
merely small. Furthermore, as shown in Ref. [6],
5
Isolated emitters, while of immense interest in mi-
croscopy, are not the norm in cathodes. Bright electron
sources have large area field emitters (LAFE) with sev-
eral emitters in an array or randomly placed. It is not
too difficult to establish that the cosine law is valid here
as well, at least for emitters that are well separated and
contribute significantly to the LAFE current. Thus, the
cosine law is applicable in general situations and may be
used to evaluate the field emitter current.
V. ACKNOWLEDGMENT
The authors acknowledge useful discussions with Dr.
Raghwendra Kumar and Rashbihari Rudra.
f (L)
4π0
2zL
z2 − L2 = E0γa = Ea
VI. REFERENCES
(39)
so that the cosine law follows as in the linear case.
IV. DISCUSSION AND CONCLUSIONS
The local field at the emitter apex and its neighbour-
hood is a very important factor in field emission calcu-
lations. Using the line-charge formalism applicable for
a vertically aligned axially symmetric emitter in a diode
configuration, we have established the cosine law of field
enhancement factor variation close to the emitter apex.
The general result for a nonlinear line charge is derived
in the same spirit as for a linear line charge distribution
where it is known that the cosine law is exact5,15 and in
fact valid further away from the emitter apex.
The results presented here are valid for a single emitter
when the anode is sufficiently far away for the line charge
and its image to have negligible effect on the anode. This
is generally true when the anode-cathode plate distance
is more than than three times the height of the emitter.
Since this condition may hold in practical devices, the
anode proximity effect is not a major impediment to the
cosine law.
1C.J. Edgcombe and U. Valdr`e, Phil. Mag. B 82 (2002) 987.
2R. G. Forbes, C.J. Edgcombe and U. Valdr`e, Ultramicroscopy
95, 65 (2003).
3F. H. Read and N. J. .Bowring, Microelectronic Engineering, 73-
74, 679 (2004).
4S. Podenok, M. Sveningsson, K. Hansen and E.E.B. Campbell,
Nano 1, 87 (2006).
5E.G. Pogorelov, A.I. Zhbanov, Y.-C. Chang, Ultramicroscopy 109
(2009) 373.
6D. Biswas, Phys. Plasmas 25, 043113 (2018).
7D. Biswas and R. Rudra, Physics of Plasmas 25, 083105 (2018).
8R. H. Fowler and L. Nordheim, Proc. R. Soc. A 119, 173 (1928).
9L. Nordheim, Proc. R. Soc. A 121, 626 (1928).
10E. L. Murphy and R. H. Good, Phys. Rev. 102, 1464 (1956).
11R. G. Forbes, App. Phys. Lett. 89, 113122 (2006).
12R. G. Forbes and J. H. B. Deane, Proc. Roy. Soc. A 463, 2907
(2007).
13K. L. Jensen, Field emission - fundamental theory to usage, Wi-
ley Encycl. Electr. Electron. Eng. (2014).
14W. P. Dyke and W. W. Dolan, Advances in Electronics and Elec-
tron Phys. 8 (1956).
15D. Biswas, G. Singh, S. G. Sarkar and R. Kumar, Ultrami-
croscopy 185, 1 (2018).
16D. Biswas, Phys. Plasmas, 25, 043105 (2018).
17J. R. Harris, K. L. Jensen, D. A. Shiffler and J. J. Petillo, Appl.
Phys. Lettrs. 106, 201603 (2015).
18D. Biswas, G. Singh and R. Kumar, J. App. Phys. 120, 124307
(2016).
19See for example, R. Whitley, J. Mathematical Analysis and Ap-
plications, 105, 502 (1985).
|
1812.03816 | 2 | 1812 | 2019-01-11T00:57:37 | Designing Silicon Photonic Devices using Artificial Neural Networks | [
"physics.app-ph",
"physics.optics"
] | We develop and experimentally validate a novel neural network design framework for silicon photonics devices that is both practical and intuitive. The framework is applicable to nearly all known integrated photonics devices, but as case studies we consider simple waveguides and chirped Bragg Gratings. By using artificial neural networks, we decrease the computational cost relative to traditional design methodologies by more than 4 orders of magnitude. We also demonstrate the abstraction of the device models to a few simple input and output parameters relevant to designers. We then apply the results to various design problems and experimentally compare fabricated devices to the neural network's predictions. | physics.app-ph | physics | Designing Silicon Photonic Devices using Artificial Neural Networks
Alec M. Hammond1 and Ryan M. Camacho1,*
1Department of Electrical & Computer Engineering, Brigham Young University, Provo,
*Corresponding author: [email protected]
Utah
Abstract
We develop and experimentally validate a novel artificial neural network (ANN) design framework for
silicon photonics devices that is both practical and intuitive. As case studies, we train ANNs to model
both strip waveguides and chirped Bragg gratings using a small number of simple input and output
parameters relevant to designers. Once trained, the ANNs decrease the computational cost relative to
traditional design methodologies by more than 4 orders of magnitude. To illustrate the power of our
new design paradigm, we develop and demonstrate both forward and inverse design tools enabled by the
ANN. We use these tools to design and fabricate several integrated Bragg grating devices. The ANN's
predictions match very well with the experimental measurements and do not require any post-fabrication
training adjustments.
1
Introduction
Silicon photonics has become a viable technology for integrating a large number of optical components in
a chip-scale format. Driven primarily by telecommunications applications, a growing number of CMOS
fabrication facilities dedicated to silicon photonics are now in operation and available for researchers and
engineers to submit photonic integrated circuit (PIC) designs to be fabricated [1].
Designing the silicon photonics components and circuits, however, remains a major bottleneck. Cur-
rent design flows are complicated by computational tractability and the need for designers with extensive
experience [2]. Unlike their electronic counterparts, photonic integrated circuits require computationally
expensive simulation routines to accurately predict their optical response functions. The typical time to
design integrated photonic devices now often exceeds the time to manufacture and test them.
To address this challenge, we propose and experimentally validate a new design paradigm for silicon
photonics that leverages artificial neural networks (ANN) in an intuitive way and is at least four orders of
magnitude faster than traditional simulation methods. Our design paradigm only requires a small number
of input and output neurons corresponding to descriptive parameters relevant to the designer.
This new approach provides benefits such as rapid prototyping, inverse design, and efficient optimization,
but requires a more sophisticated ANN than previous approaches. We demonstrate practical confidence
in our method's accuracy by fabricating and measuring devices that experimentally validate the ANN's
predictions. As illustrative examples, we design, fabricate, and test chirped integrated Bragg gratings. Their
large parameter spaces and nonlinear responses are typical for devices that are computationally prohibitive
using other techniques. The experimental results show remarkable agreement with the ANN's predictions,
and to the authors' knowledge represent the first experimental validation of photonic devices designed using
ANN's.
This work builds on recent theoretical results showing that it is possible to model nanophotonic structures
using ANNs. For example, Ferreira et al. [3] and Tahersima et al. [4] demonstrated that ANNs could assist
with the numerical optimization of waveguide couplers and integrated photonic splitters respectively.
In
both cases the input parameter space was the entire 2D array of grid points, showing the power of ANNs in
blind "black box" approach, though limiting the designer's ability to intuitively adjust input parameters. A
1
9
1
0
2
n
a
J
1
1
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
2
v
6
1
8
3
0
.
2
1
8
1
:
v
i
X
r
a
related approach has also been applied to periodic photonic structures, which are often difficult to efficiently
model [5, 6].
Two recent theoretical papers by Zhang et al [7], and Peurifoy et al.
[8] go beyond simply optimizing
over a large parameter space, and used ANNs to calculate complicated spectra using a smaller number of
intuitive, smoothly varying input parameters. Even though in both cases each wavelength point in the
calculated spectra required its own ANN output neuron, the usefulness of using ANNs to model systems
with intuitive input parameters was demonstrated.
To illustrate the power of our new design paradigm, we demonstrate both forward and inverse design tools
that use a chirped Bragg grating ANN as a computational backend. The forward design tool is interactive,
and was used to design our fabricated circuits. The inverse design tool quickly constructs a temporal pulse
compressing chirped Bragg grating within specified design constraints -- a task typically too computationally
expensive for traditional methods.
2 Results
Overview
To motivate our approach, we first describe a neural network that models the effective index of a sili-
con photonic strip waveguide with various widths and thicknesses. While waveguide simulation is already
straightforward from a designer's perspective, the model illustrates the advantages of our approach and is
a key building block for more advanced ANN models described below which are less straightforward using
existing techniques. These advantages include the ANN's computational speedup of over 4 orders of mag-
nitude, and the simplification and speedup of other complicated simulation routines that rely on effective
index calculations. A more advanced example that is computationally intractable via traditional methods
is then given, in which we demonstrate an ANN that models the complex relationship between a chirped
silicon photonic Bragg grating's design parameters and its corresponding spectral response. Many design-
ers leverage silicon photonic chirped Bragg gratings to equalize optical amplifier gain [9], compensate for
semiconductor laser dispersion [10, 11], and enable nonlinear temporal pulse compression [12, 13].
Figure 1 illustrates the new design methodology. First, we iterated between generating an appropriate
dataset and training the ANN until the model adequately characterized the device. Next, we used the ANN
to simulate circuits and solve inverse design problems. Finally, we fabricated devices to validate the results.
Waveguide Neural Network
We first report on a simple waveguide neural network capable of estimating the effective index of any arbitrary
silicon photonic waveguide geometry for a variety of modes. Specifically, we modeled the relationship between
the waveguide's width, thickness, and operating wavelength and the effective index for the first two TE and
TM modes. We note that including wavelength as an input parameter is a unique and enabling strategy
not previously adopted (see Discussion section below). Figure 2 (f) compares the ANN's predicted effective
index to a its corresponding simulation. The first TE and TM mode for any silicon photonic waveguide
with a width between 350 nm and 1000 nm and a thickness between 150 nm and 350 nm are demonstrated.
The network estimates a smooth response for both modes simultaneously, even for data points outside of its
training set. The ANN's smooth output also produces smooth analytic derivatives, which are essential for
calculating group index profiles and for gradient-based optimization routines.
We implemented various tests to validate the network's accuracy. First, we split the initial dataset into
a training set and a validation set. While the network evaluated both sets after each epoch (i.e. training
iteration) only the training set's results were used to update the network's weights. We monitored the
validation set's results to assess overfitting. To better understand the network's performance after each
iteration, we recorded each epoch's mean-square-error (MSE) and coefficient of determination (R2). Figure
2 (a) and (b) illustrate the MSE and R2 respectively after each epoch. To prevent overfitting, we stopped
training at 100 epochs, where the MSE and R2 appear to converge. At this point, the network demonstrated
a MSE of 1.323× 10−4 for the training set and 7.490× 10−5 for the validation set. The final R2 values for the
training data and validation data were 0.9996 and 0.9997 respectively. The MSE and R2 evolution for both
the training set and validation set converge well, indicating little to no overfitting. Figure 2 (c) illustrates
2
Figure 1: The process overview describing the new design methodology. First, datasets are generated using
traditional numerical methods (described in Methods). From this dataset, a neural network is trained to
characterize the device under consideration. Figures 2 & 3 illustrate this process for a strip waveguide and
a chirped grating respectively. Often, the designer iterates between these two steps until an appropriate
model is developed. Once the model is ready, several design applications, like circuit simulations and inverse
design solutions, are available. The designs are then fabricated to validate the model's results. From here,
the model can be shared and extended.
3
the relative error for both the training and validation sets after the final epoch. Both the training set errors
and validation set errors are similarly distributed and tightly bounded between −1% and 1%, once again
indicating little to no overfitting.
With confidence in the waveguide neural network's prediction accuracy, we benchmarked its speed and
found that a single neural network evaluation was 104 times faster on average than the corresponding finite
difference eigenmode simulation. Figure 2 (d) compares the computation speed for the ANN to the eigenmode
solver, Meep Photonic Bands (MPB). This significant speedup enables many simulation techniques, like
the layered dielectric media transfer matrix method (LDMTMM) [14] or the eigenmode expansion method
(EMM) [15], where photonic components are discretized into individual waveguides. Using the ANN, a
transfer matrix for each waveguide can be quickly generated and cascaded to formulate a fairly accurate
response for the device.
In addition, modeling fabrication variations is now much quicker since existing
Monte Carlo sampling routines can leverage the ANN's speed.
Bragg Grating Neural Network
Modeling the relationship between a Bragg grating's physical design parameters and its corresponding re-
sponses is difficult since no one-to-one mapping exists. Consequently, many designers resort to black-box
optimization routines that strategically search the parameter space for viable design options. As a result,
inverse design problems -- where a simulation needs to run each iteration -- become intractable for even
modest size gratings. If a full 3D FDTD simulation is performed, for example, each optimization iteration
can take between 8-12 hours on typical desktop computing systems. In addition, the optimization routines
tend to inefficiently simulate redundant test scenarios for different design problems. We train and demon-
strate a Bragg ANN, however, that can predict a grating's response on the order of milliseconds on the same
system, enabling much faster solutions to more complex design problems. We fabricate various test devices
and validate our neural network's predictions.
Using the waveguide neural network, we generated a dataset to train our Bragg grating neural network to
predict the reflection spectrum and group delay response of a silicon photonic, sidewall-corrugated, linearly
chirped Bragg grating, as illustrated in Figure 3 (d). We note that generating the training dataset was
approximately 2 orders of magnitude faster using the waveguide ANN reported above rather than traditional
methods. To smooth apodization dependent ringing, we pre-processed the training data. More information
regarding this step is provided in the Supplementary Material. We parameterized the gratings by length of
the first grating period (a0), length of the last grating period (a1), number of grating periods N G, and grating
corrugation width difference( ∆w = w1 − w0). We designed the network to receive these four parameters
along with a single wavelength point as inputs. The network has two outputs: reflected optical power and
group delay.
Similar to the waveguide network, we divided the dataset into a training set and validation set. We
tracked both the MSE and the R2 metrics after each epoch. The Bragg training set was much larger than
the waveguide training set, owing to the larger parameter space. Consequently, the MSE converged within
the first few epochs and we stopped training after just five epochs to prevent overfitting. The final MSE for
the training and validaton sets was 1.845 × 10−4 and 1.677 × 10−4 respectively. The final R2 was 0.9975
and 0.9977 respectively. Once again, the MSE and R2 evolution for both the training set and validation
set converge well, indicating little to no overfitting. Figure 3 (a-b) illustrates the network's MSE and R2
evolution. Figure 3 (c) illustrates the absolute error for both the training sets and validation sets. We
calculated the absolute error because several training samples were at or near zero and skewed the relative
error.
We note that calculating Bragg grating response with the ANN is much more computationally efficient
than previously demonstrated methods. This is because the Bragg ANN linearly increases in computation
complexity with added grating parameters, while LDMTMM and all other methods known to these authors
increase at least quadratically.
To validate the Bragg ANN, we fabricated and measured several silicon photonic Bragg gratings with
different chirping patterns and compared their transmission, reflection, and group delay spectra to the neural
network's predictions. The gratings were arranged in one of two configurations: (1) a simple circuit that only
measured the Bragg grating's transmission spectra and (2) a more complicated interrogation circuit capable
of measuring the reflection, transmission, and group delay profiles from the same device simultaneously.
4
Figure 2: Waveguide artificial neural network training results demonstrated by the training convergence
with reference to the mean square error (a), the coefficient of determination (b), and the residual errors after
training (c). Panel (d) compares the computational cost for the ANN and the eigenmode solver thas is used
to simulate the mode profiles (e). Panel (f) exhibits the effective index profiles as a function of a waveguide
geometry at 1550 nm for the first TE and TM modes.
5
Figure 3: Bragg grating artificial neural network training results demonstrated by the training convergence
with reference to the mean square error (a), the coefficient of determination (b), and the absolute error after
training (c). (d) illustrates the different adjustable grating parameters and (e) illustrates the interrogation
circuit used to extract the reflection, transmission, and group delay profiles simultaneously from the chirped
Bragg grating. A grating coupler (GC) feeds light into various Y-branches (YB) and directional couplers
(DC) such that the transmission and reflection spectra can both be extracted from the chirped Bragg grating
(BG). Half of the reflected signal is sent through a Mach-Zehner Interferometer (MZI). The output of which
is used to extract the group delay. .
6
Figure 3 illustrates the interrogation circuit used to measure all three responses simultaneously. In both
configurations, grating couplers were used to route light on and off the chip. While the simpler circuit required
less de-embedding, the full interrogator circuit allowed for a more comprehensive device characterization.
The transmission-only gratings were designed with various grating period bandwidths from 5 nm to 20
nm, each with 600 periods and a corrugation width of 50 nm. The initial design parameters produced ANN
predictions that match the measured data extremely well. Small discrepancies in the grating responses are
largely attributed to the grating's apodization profile and detector noise. Figure 4 illustrates the comparison
between the ANN's predictions and the measured data.
We designed the remaining gratings using a much smaller chirp bandwidth of 3 nm with 750 grating
periods and a 30 nm corrugation width. We mirrored the orientation of half the gratings in order to
measure both positive and negative sloped group delay profiles. Once measured, we normalized the data
by de-embedding the responses from the various Y-branches, directional couplers, and grating couplers that
complicate the measurement data. The process is explained in the Supplementary Material. Even with the
rather complex transfer function, the transmission, reflection, and group delay profiles match the ANN's
corresponding predictions well except for occasional resonant features caused by fabrication defects. These
defects are expected since the narrow bandwidth devices have a grating pitch with a fine discretization
that approaches the e-beam raster grid resolution. Small changes in grating pitch that don't align with the
raster grid occasionally produce weak Fabry-Perot resonance conditions visible in the data. These raster-
induced defects also account for a small lateral shift (1 nm) in the responses. Even with these fabrication
challenges, it is notable that the ANN successfuly predicts the transmission, reflection, and group delay
profiles simultaneously.
In fact, the ability to do so in noisy fabrication environments is one of the key
advantages of the ANN and may allow for efficient parameter extraction where other methods fail.
Forward design
The neural network's speed and flexibility enable forward design exploration. For example, Figure 5 illus-
trates a graphical user interface (GUI) built with slider bars to adjust the Bragg grating's design parameters
(i.e. corrugation widths, grating length, chirp pattern, etc). The plots dynamically update, calling the neural
network every time the user modifies the input, and display the corresponding reflection and group delay
profiles. Because wavelength is included as an input to the ANN rather than an output, arbitrary wave-
length sampling within the domain is allowed. Computing these responses in real time is not possible using
traditional techniques. This capability is valuable and allows even novice designers the ability to rapidly
gain device intuition without necessarily understanding the underlying numerical techniques.
Inverse design
This new approach also enables an entirely new set of inverse design problems. For example, we used the
neural network in conjunction with a truncated Newtonian optimization algorithm to design a temporal
pulse compressor. Designers often rely on dispersive Bragg gratings to generate short, optical pulses for
high-capacity communications [12]. In this particular case study, we assumed an arbitrary source generates
a 20 ps wide chirped pulse with 4 nm of bandwidth. Figure 6 illustrates the optimization routine's evolution,
the resulting grating response, and the pulse shape before and after the Bragg grating. Such optimization
algorithms run much quicker than previously known methods, owing to the accelerated cost function. The
agnostic nature of the neural network interface works well with typical optimization routines, especially since
any arbitrary wavelength sampling is allowed. Depending on the cost function formulation, gradient-based
methods could directly evaluate the Jacobian and Hessian tensors from the ANN without any extra sampling
or discretization.
3 Discussion
Our method demonstrates a new, viable platform for silicon photonic circuit design. With a single global pa-
rameter fit, we successfully modeled silicon photonic waveguides and silicon photonic chirped Bragg gratings
with arbitrary bandwidths, chirping patterns, lengths, and corrugation widths and allowed arbitrary wave-
length sampling. Future work could explore new network architectures (e.g. different activation functions,
7
Figure 4: Fabrication data compared to corresponding ANN predictions. (a1-a4) Measured transmission
responses for gratings with a period chirp of 5 nm (a1), 10 nm (a2), 15 nm (a3), and 20 nm (a4). (b1-b2)
Transmission and reflection responses for two different Bragg gratings. Both gratings share the same design
parameters, and have an identical but opposite linear chirp. The result of the mirrored chirping is seen in
both the normalized MZI interference patterns (c1 , c2) and the extracted group delay responses (d1 , c2).
8
Figure 5: Graphical user interface used to explore the design space of a chirped Bragg grating. The slider
bars on the left control physical parameters like grating length (NG), grating corrugation (dw), and the
grating chirp (a1 and a2). Any time the user adjusts these parameters, the program calls the ANN and
reproduces the expected reflection and group delay profiles for that particular grating. Due to the ANN's
speed, the program is extremely responsive.
9
Figure 6: ANN-assisted design of a monolithic temporal pulse compressor using a silicon photonic chirped
Bragg grating. A truncated Newton algorithm was tasked with constructing a grating that compressed an
arbitrary chirped pulse by a factor of 2. After 340 grating simulations, the optimizer sufficiently minimized
a cost function (right) that compared the new pulse's width to the old pulse. The resulting grating is
demonstrated below and the input, output, and desired pulses for iterations 1, 140, and 340 are demonstrated
on the left.
10
layer connections, etc.) and training algorithms. Several other devices, like ring resonators [16], can also be
modeled. One could subsequently cascade several ANNs that model the scattering parameters of different
devices, opening the door to large-scale optimization problems.
An important feature of this work is the choice to model the wavelength as a continuous input parameter
rather than fix each output neuron at a specific wavelength point as done in all previous work known to
the authors. The waveguide ANN, for example, outputs effective index values and the Bragg ANN outputs
reflection and group delay values across the entire input spectrum. This approach, while more difficult to
train, is more convenient for the designer. For example, an optimization routine tasked with designing a
Bragg filter can focus more on parameters like the bandwidth and shape, rather than an arbitrarily sampled
wavelength profile. Furthermore, this method doesn't require the training spectra to have the same sampling.
Training sets for structures like ring resonators, whose features may require finer wavelength resolution than
other devices, can now be strategically simulated to highlight these features. Assuming the network is trained
correctly across a suitable domain, the ANN will seamlessly interpolate between both design parameters and
wavelength points without any additional routines.
Unlike traditional simulation methods, training an arbitrary device ANN requires large datasets that
are too time-intensive for most typical computers. With the growing availability of vast cloud-based com-
putational resources, however, several million training simulations can now be run in hours or days.
[17].
Once trained, a neural network can reliably interpolate between training data, is compact and easily shared
with the community, and can even continue to learn on new datasets via transfer learning [18]. Thus, the
computational complexity inherent in designing integrated photonic devices can be moved to the front end of
the design process, allowing individual designers to work with abstracted components whose optical response
can be rapidly calculated.
As with all deep learning applications, the network's utility is limited by biases introduced in the training
set, the network architecture, or even the training process itself [19]. Fortunately, we can anticipate these
biases by extracting the model's prediction uncertainty without modifying our network architecture. Dropout
inference techniques leverage models that rely on dropout layers to mitigate over-fitting (a form of network
bias) [20]. Even pre-trained networks can use dropout inference to extract prediction uncertainties without
any modifications to the network. This particular network design methodology opens the door to many more
applications, like training on fabricated device data. Foundry's that develop process design kits (PDKs), for
example, can use this technique to model their fabrication processes while preserving their trade secrets.
4 Methods
Training data generation and preprocessing
We generated our waveguide neural network's training set on a high performance computing cluster (HPC)
using MEEP Photonic Bands (MPB) [21], a finite difference eigenmode solver. The solver simulated 31
different waveguide widths from 350 nm to 1500 nm and 31 waveguide thicknesses ranging from 150 nm to
400 nm resulting in 961 different geometries. The solver simulated 200 distinct wavelength points in the
range of 1400 nm to 1700 nm. The total number of training samples fed into the neural network was 192,200.
70% of the data set was used as training samples and the remaining 30 % was used as validation samples.
Each sample had three inputs (width, thickness, and wavelength) and four corresponding outputs (effective
indices for the first two TE and TM modes). No postprocessing was performed on the waveguide training
data.
On the same HPC, we generated our Bragg training set by simulating 104,131 different gratings with
the layered dielectric media transfer matrix method (LDMTMM) [14]. The LDMTMM method models each
individual section of the Bragg grating as an ideal waveguide and cascades each sections's corresponding
transfer matrix to estimate the grating's response for each wavelength point of interest. We calculated each
individual waveguide's effective index using the waveguide neural network. Our simulations swept through
10 different corrugation widths from 10 nm to 100 nm, 11 different grating lengths from 100 periods to 2000
periods, and 32 different chirping patterns.
Once the grating spectra were generated, we fit the results to a generalized skewed Gaussian (see Sup-
plementary Material) to reduce ringing and to generalize the grating's response to arbitrary apodization
11
profiles. We found that without fitting, the resulting oscillations significantly complicate the training pro-
cess and restrict the network's domain to a single apodization. We fit both the reflection spectrum and group
delay responses to generalized Gaussians and resampled the results with 250 wavelength points from 1.45
µm to 1.65 µm. Since the nonlinear fitting routine occasionally failed, not all of the simulated gratings were
appropriate for testing. After filtering through the results, we generated a database of 26,032,750 training
samples.
Neural network design and training
Both neural networks were trained on the same HPC cluster using Keras [22] and Tensorflow [23]. Several
hundred different architectures were tested. To gauge the effectiveness of each architecture, the mean-
squared-error and coefficient of determination (R2) metrics were used. The waveguide neural network that
worked best had 4 hidden layers with 128 neurons, 64, neurons, 32 neurons, and 16 neurons. Each neuron
used a hyperbolic tangent activation function. The Bragg grating neural network was designed with 10
hidden layers and 128 neurons each. RELU activation functions were used. Both networks were trained with
16 sample batch sizes. While the waveguide neural network was trained with 100 epochs, the Bragg grating
neural network only needed about 5 epochs to reach sufficient results, primarily due to the large training
set. The Bragg training set was normalized to improve the network's expressive capabilities.
Simulation benchmarks
We performed all benchmarks using a quad-core Intel(R) i5-2400 CPU clocked at 3.10 GHz with 12 GB of
RAM. To evaluate the waveguide ANN's speed, we simulated various waveguide parameters in serial using
both the ANN and MPB. To evaluate the BG ANN's speed, we simulated various grating's in serial using
both the ANN and the LDMTMM. We linearly fit each method's results and compared the slopes to examine
the speedup.
Device fabrication
The silicon photonic Bragg gratings were fabricated by Applied Nanotools Inc (Edmonton, Canada) using
a direct-write 100 keV electron beam lithographic process. Silicon-on-Insulator wafers with 220 nm device
thickness and 2 µm thick insulator layer were used. The devices were patterned with a raster step of 5 µm
and etched with a ICP-RIE etch process. A 2.2 µm oxide cladding was deposited using a plasma-enhanced
chemical vapour deposition (PECVD) process.
Device measurement
Each device was measured using an automated process at the University of British Colombia (UBC). An
Agilent 81600B tunable laser was used as the input source and Agilent 81635A optical power sensors as
the output detectors. The wavelength was swept from 1500 to 1600 nm in 10 pm steps. A polarization
maintaining (PM) fibre was used to maintain the polarization state of the light, to couple the TE polarization
in and out of the grating couplers. Several dembedded test structures were used to normalize out the coupler
profiles.
Data Availability
The data that support the plots within this paper and other findings of this study are available from the
corresponding authors upon reasonable request.
References
[1] L. Chrostowski and M. Hochberg, Silicon Photonics Design: From Devices to Systems, 1st ed. Cam-
bridge ; New York: Cambridge University Press, May 2015.
12
[2] W. Bogaerts and L. Chrostowski, "Silicon Photonics Circuit Design: Methods, Tools and Challenges,"
Laser & Photonics Reviews, vol. 12, no. 4, p. 1700237, Apr. 2018.
[3] A. da Silva Ferreira, C. H. da Silva Santos, M. S. Gon¸calves, and H. E. Hern´andez Figueroa, "Towards an
integrated evolutionary strategy and artificial neural network computational tool for designing photonic
coupler devices," Applied Soft Computing, vol. 65, pp. 1 -- 11, Apr. 2018.
[4] M. H. Tahersima, K. Kojima, T. Koike-Akino, D. Jha, B. Wang, C. Lin, and K. Parsons, "Deep Neural
Network Inverse Design of Integrated Nanophotonic Devices," arXiv:1809.03555 [physics.app-ph], Sep.
2018.
[5] S. Inampudi and H. Mosallaei, "Neural network based design of metagratings," Applied Physics Letters,
vol. 112, no. 24, p. 241102, Jun. 2018.
[6] A. D. Silva Ferreira, G. N. Malheiros-Silveira, and H. E. Hernandez-Figueroa, "Computing Optical Prop-
erties of Photonic Crystals by Using Multilayer Perceptron and Extreme Learning Machine," Journal
of Lightwave Technology, vol. 36, no. 18, pp. 4066 -- 4073, Sep. 2018.
[7] T. Zhang, Q. Liu, J. Dai, X. Han, J. Li, Y. Zhou, and K. Xu, "Spectrum prediction and inverse design
for plasmonic waveguide system based on artificial neural networks," arXiv:1805.06410 [physics], May
2018.
[8] J. Peurifoy, Y. Shen, L. Jing, Y. Yang, F. Cano-Renteria, B. G. DeLacy, J. D. Joannopoulos,
M. Tegmark, and M. Soljacic, "Nanophotonic particle simulation and inverse design using artificial
neural networks," Science Advances, vol. 4, no. 6, p. eaar4206, Jun. 2018.
[9] M. Rochette, M. Guy, S. LaRochelle, J. Lauzon, and F. Trepanier, "Gain equalization of EDFA's with
Bragg gratings," IEEE Photonics Technology Letters, vol. 11, no. 5, pp. 536 -- 538, May 1999.
[10] D. T. H. Tan, K. Ikeda, R. E. Saperstein, B. Slutsky, and Y. Fainman, "Chip-scale dispersion engineering
using chirped vertical gratings," Optics Letters, vol. 33, no. 24, pp. 3013 -- 3015, Dec. 2008.
[11] M. J. Strain and M. Sorel, "Design and Fabrication of Integrated Chirped Bragg Gratings for On-Chip
Dispersion Control," IEEE Journal of Quantum Electronics, vol. 46, no. 5, pp. 774 -- 782, May 2010.
[12] D. T. H. Tan, P. C. Sun, and Y. Fainman, "Monolithic nonlinear pulse compressor on a silicon chip,"
Nature Communications, vol. 1, p. 116, Nov. 2010.
[13] N. M. L. B. J. Eggleton, G. Lenz, "Optical Pulse Compression Schemes That Use Nonlinear Bragg
Gratings," Fiber and Integrated Optics, vol. 19, no. 4, pp. 383 -- 421, Oct. 2000. [Online]. Available:
https://doi.org/10.1080/014680300300001725
[14] R. Helan, "Comparison of methods for fiber Bragg gratings simulation," ser. International Spring Sem-
inar on Electronics Technology ISSE.
IEEE, 2006, pp. 175+.
[15] D. F. G. Gallagher and T. P. Felici, "Eigenmode expansion methods for simulation of optical propagation
in photonics - Pros and cons," in Integrated Optics: Devices, Materials, and Technologies Vii, Y. S.
Sidorin and A. Tervonen, Eds. Bellingham: Spie-Int Soc Optical Engineering, 2003, vol. 4987, pp.
69 -- 82.
[16] W. Bogaerts, P. De Heyn, T. Van Vaerenbergh, K. De Vos, S. K. Selvaraja, T. Claes, P. Dumon,
P. Bienstman, D. Van Thourhout, and R. Baets, "Silicon microring resonators," Laser & Photonics
Reviews, vol. 6, no. 1, pp. 47 -- 73, Jan. 2012.
[17] C. Vecchiola, S. Pandey, and R. Buyya, High-Performance Cloud Computing: A View of Scientific
Applications. New York: IEEE, 2009.
[18] S. J. Pan and Q. Yang, "A Survey on Transfer Learning," IEEE Transactions on Knowledge and Data
Engineering, vol. 22, no. 10, pp. 1345 -- 1359, Oct. 2010.
13
[19] N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, "Dropout: A Simple Way
to Prevent Neural Networks from Overfitting," Journal of Machine Learning Research, vol. 15, pp.
1929 -- 1958, Jun. 2014.
[20] Y. Gal and Z. Ghahramani, "Dropout as a Bayesian Approximation: Representing Model Uncertainty
in Deep Learning," arXiv:1506.02142 [cs, stat], Jun. 2015.
[21] S. G. Johnson and J. D. Joannopoulos, "Block-iterative frequency-domain methods for Maxwell's equa-
tions in a planewave basis," Optics Express, vol. 8, no. 3, pp. 173 -- 190, Jan. 2001.
[22] "Keras." [Online]. Available: https://keras.io/
[23] TensorFlow:
Large-Scale Machine Learning on Heterogeneous Systems.
[Online]. Available:
http://tensorflow.org/
Acknowledgments
The authors would like to thank Lukas Chrostowski for useful discussions relating to the Bragg structures
and for facilitating the SiEPIC fabricating process, as well as David Buck for supplying additional fabrication
data.
Author Contributions
AMH & RMC conceived the idea, AMH designed and trained the ANN, AMH & RMC designed the devices,
AMH & RMC evaluated and interpreted the data, AMH & RMC wrote the manuscript.
Competing Interests
The authors declare no competing interests.
Materials & Correspondence
Please send all correspondence and data requests to [email protected]
14
|
1904.11731 | 1 | 1904 | 2019-04-26T09:23:11 | Effect of Ion Migration Induced Electrode Degradation on the Operational Stability of Perovskite Solar Cells | [
"physics.app-ph"
] | Perovskite-based solar cells are promising due to their rapidly improving efficiencies, but suffer from instability issues. Recently it has been claimed that one of the key contributors to the instability of perovskite solar cells is ion migration induced electrode degradation, which can be avoided by incorporating inorganic hole blocking layers (HBL) in the device architecture. In this work, we investigate the operational environmental stability of methylammonium lead iodide (MAPbI3) perovskite solar cells that contain either an inorganic or organic HBL, with only the former effectively blocking ions from migrating to the metal electrode. This is confirmed by X-ray photoemission spectroscopy measured on electrodes of degraded devices, where only electrodes of devices with an organic HBL show a significant iodine signal. Despite this, we show that when these devices are degraded under realistic operational conditions (i.e. constant illumination in a variety of atmospheric conditions), both types of devices exhibit nearly identical degradation behavior. These results demonstrate that contrary to prior suggestions, ion-induced electrode degradation is not the dominant factor in perovskite environmental instability under operational conditions. | physics.app-ph | physics | Effect of Ion Migration Induced Electrode
Degradation on the Operational Stability of
Perovskite Solar Cells
Boris Rivkin,1,2 Paul Fassl,1,2 Qing Sun,1,2 Alexander D. Taylor,1,2 Zhuoying Chen3 and Yana
Vaynzof1,2*
1 Kirchhoff Institute for Physics, Im Neuenheimer Feld 227, 69120, Germany
2 Centre for Advanced Materials, Im Neuenheimer Feld 225, 69120, Germany
3 Laboratoire de Physique et d'Etude des Matériaux (LPEM), ESPCI Paris, PSL Research
University, CNRS, Sorbonne Université, 10 Rue Vauquelin, 75005 Paris, France
*Correspondence to: [email protected]
ABSTRACT
Perovskite-based solar cells are promising due to their rapidly improving efficiencies, but suffer
from instability issues. Recently it has been claimed that one of the key contributors to the
instability of perovskite solar cells is ion migration induced electrode degradation, which can be
avoided by incorporating inorganic hole blocking layers (HBL) in the device architecture. In this
1
work, we investigate the operational environmental stability of methylammonium lead iodide
(MAPbI3) perovskite solar cells that contain either an inorganic or organic HBL, with only the
former effectively blocking ions from migrating to the metal electrode. This is confirmed by X-
ray photoemission spectroscopy measured on electrodes of degraded devices, where only
electrodes of devices with an organic HBL show a significant iodine signal. Despite this, we show
that when these devices are degraded under realistic operational conditions (i.e. constant
illumination in a variety of atmospheric conditions), both types of devices exhibit nearly identical
degradation behavior. These results demonstrate that contrary to prior suggestions, ion-induced
electrode degradation is not the dominant factor in perovskite environmental instability under
operational conditions.
INTRODUCTION
The favorable electronic and optical properties of hybrid lead-halide perovskites have enabled the
remarkable performance increase in solar cells based on such materials, which are currently
reaching an efficiency of 22.7%.1 Device stability, however, remains a major factor impeding their
commercialization, where lifetimes of more than 25 years are required.2 Previous studies have
investigated the degradation of device performance due to environmental factors such as oxygen,3 --
5 moisture,6 -- 9 and heat10,11 and explanatory models have been proposed for some scenarios.12,13 At
the same time, device degradation under inert (nitrogen atmosphere) operating conditions is a
widely observed but not conclusively explained phenomenon.14,15 In one popular model, it is
claimed that device degradation originates predominantly from metal electrode corrosion: mobile
halide ions, such as iodide, migrate through the active material and the adjacent electron extraction
layer towards the metal cathodes, such as silver.16 -- 20 This could then give rise to the formation of
2
insulating silver iodine,21,22 which would inhibit the extraction of charges, increase series
resistance, and enable the formation of an undesirable dipole interface layer.17,23 This hypothesis
is often accompanied by the claim that a buffer layer of a dense electron transporting material,
such as ZnO, TiOx or SnO2, could serve as an "ion-blocking layer" and prohibit the migration of
ions and thus improve the stability of perovskite solar cells.24 -- 27 While such studies have
demonstrated increased device shelf life (dark storage) stability, few studies featuring full device
degradation, under constant illumination and both inert and non-inert atmospheres, as well as
employing rigorous compositional analysis to conclusively prove either of these two claims have
been presented to date.27
In this work, we compare the environmental stability of methylammonium lead iodide (MAPbI3)
perovskite solar cells that contain inorganic ZnO nanoparticle-based HBL against reference
devices that employ the commonly used bathocuproine (BCP) as the HBL. While both types of
devices show a similar initial photovoltaic performance, the inorganic HBL effectively blocks
mobile iodide ions from reaching and reacting with the metal electrode. This is not the case for
devices with an organic HBL, in which a significant amount of iodine is detected by X-ray
photoemission spectroscopy. Characterizing the environmental stability of the two types of devices
under illumination allows us to probe the role of ion induced electrode degradation on the stability
of the devices in various atmospheres. We find that both types of devices show very similar
degradation dynamics, revealing that suppressing ion induced electrode degradation does not
improve the operational stability of these cells.
3
EXPERIMENTAL METHODS
Device fabrication
If not stated otherwise, all materials were purchased from Sigma Aldrich. To fabricate devices,
pre-patterned ITO-coated glass substrates (PsiOTech Ltd., 15 Ωsq-1) were first cleaned
sequentially with acetone and 2-propanol, followed by 10 min oxygen plasma treatment.
PEDOT:PSS (Clevios Al 4083, Heraeus) was spin-coated onto the clean ITO substrates and then
annealed at 150 °C for 10 min in air. For the perovskite layer fabricated by the lead acetate
trihydrate recipe, CH3NH3I (GreatCell Solar) and Pb(Ac)2·3(H2O) (3:1, molar ratio) were
dissolved in anhydrous N,N-dimethylformamide (DMF) with a concentration of 40 wt% with the
addition of hypophosphorous acid solution (6 µL mL-1 DMF). The perovskite solution was spin-
coated at 2000 rpm for 60 s in a drybox (RH < 0.5 %). After drying for 5 min, the samples were
annealed at 100 °C for 5 min. Subsequently, the samples were transferred to a N2 filled glovebox.
PCBM (Solenne BV) in chlorobenzene (20 mg mL−1) was dynamically spin-coated at 2000 rpm
for 45 s and annealed at 100 °C for 10 min. BCP was fully dissolved in 2-propanol (0.5 mg mL−1)
and dynamically spin-coated at 4000 rpm for 30 s. ZnO nanoparticles were synthesized following
an adapted procedure of Pacholski, which is briefly described in Supplementary Note 1.28 The
resulting nanoparticles were further characterized by UV-visible absorbance and transmission
electron microscopy (TEM) (Figure S2 and Figure S3, Supporting Information). TEM
characterization was carried out by a JEOL 2010 TEM (200 kV) equipped with a Gatan camera.
Single or multiple layers of ZnO were cast from colloidal solution to obtain layers of different
thicknesses. To complete the device, 80 nm silver electrodes were deposited via thermal
evaporation under high vacuum.
4
Device characterization and degradation
To properly assess the degradation of the perovskite solar cells each device was identically
prepared, stored in a nitrogen filled glovebox, and transferred to a sealed environmental box
without exposure to ambient air. A constant flow of either nitrogen, nitrogen and oxygen (80:20,
v:v) or humidified nitrogen (30% relative humidity) was connected to an environmental box. The
oxygen percentage was controlled by adjusting the relative flow rate of O2 to N2, and monitored
by a zirconia sensor (Cambridge Sensotet, Rapidox 2100) continuously before being connected to
the environmental box. All of the devices were operated under simulated AM1.5 sunlight at 100
mW cm−2 irradiance (Abet Sun 3000 Class AAA solar simulator) for 10 h (in open circuit
condition), then "rested" in the dark for two hours before another two hour measurement period,
bringing the total experiment time to 14 h. This rest period was performed in response to Nie et
al., who demonstrated that short rest periods could "heal" degraded devices.14 The J -- V
measurements were performed with a Keithley 2450 Source Measure Unit. The cells were scanned
from forward bias to short circuit and back at a rate of 0.5 V s−1 after holding under illumination
at 1.2 V for 2 s. The light intensity was calibrated with a Si reference cell (NIST traceable, VLSI)
and corrected by measuring the spectral mismatch between the solar spectrum, the spectral
response of the perovskite solar cell, and the reference cell. The mismatch factor was calculated to
be approximately 11%.
X-ray photoemission spectroscopy and depth profiling
The perovskite devices investigated by photoemission spectroscopy (PES) measurements were
fabricated and degraded as described above. The Ag electrodes were peeled off and transferred
into an ultrahigh vacuum (UHV) chamber of the PES system (Thermo Scientific ESCALAB
5
250Xi) for XPS measurements. XPS measurements were performed using an XR6
monochromated Al Kα source (hν = 1486.6 eV) and a pass energy of 20 eV. Depth profiles were
performed on the remaining layers of the devices using a MAGSIC Ar etching source.
RESULTS AND DISCUSSION
To investigate the role of ion induced electrode degradation on the operational stability of
perovskite solar cells, we study the degradation of complete cells using methylammonium lead
iodide (MAPbI3) in the inverted ITO/PEDOT:PSS/MAPbI3/PCBM/HBL/Ag architecture, as
displayed in Figure 1. We fabricate two types of perovskite solar cells: one with the commonly
used organic BCP HBL and another with ZnO nanoparticles with all other layers fabricated in an
identical fashion. In order to provide the best possible comparison of the degradation behavior
between devices with the two HBLs, our first step was to optimize the thickness of the ZnO layer
in order to achieve a comparable initial photovoltaic performance of the two types of devices with
similar charge extraction efficiency and recombination. ZnO nanoparticles were deposited via
spin-coating at various spin speeds, and the resulting device's performance was measured under
AM 1.5G solar illumination and are displayed in Figure 2. As the spin speed decreases (and thus
the ZnO layer becomes thicker), device performance first increases up until approximately 8 nm
of ZnO thickness and then decreases with further increasing ZnO thickness. At this thickness, both
current and fill factor are maximized, while the voltage is largely independent of HBL thickness.
At this optimal thickness the ZnO devices possess comparable performance to the BCP reference
devices. Typical J-V curves for both device types are shown in Figure S1 (Supporting
Information).
6
As mentioned before, previous work has suggested that one of the primary degradation
mechanisms in perovskite-based devices is due to migration of ions towards and subsequent
reaction with the metal electrodes. In the case of the devices in this study, this would be the
migration of iodide to the silver electrode, which could then react with the silver to form the
insulator AgI.16,19,29 To examine whether the ZnO HBL did effectively block iodide migration,
both ZnO and BCP devices were degraded in an inert environment and under constant solar
illumination for 14 h, and afterwards the Ag electrodes were removed via tape and characterized
by XPS. This allowed us to look for the presence of iodine in the electrodes. We note that since
XPS experiments are performed in ultra-high vacuum, volatile iodine cannot be detected, so only
species of bound iodide, such as in metal-iodides would be detected. The I3d signals for both the
ZnO and BCP devices' electrodes are displayed in Figure 3a. While the BCP devices show a clear
and strong I3d signal, indicating that iodide has indeed reacted with the electrode, the ZnO devices,
in contrast, show no presence of iodine. This confirms that, on the timescale of the degradation
experiment, a ZnO HBL effectively blocks the migration of iodide ions into a metal electrode,
most likely due to the layer's higher density when compared to BCP. Figure 3b shows the excess
iodine signal measured by XPS depth profiling, of degraded devices, from which the electrodes
were peeled off. Similar to the results of Figure 3a, significant excess of iodine is detected in the
PCBM extraction layer of the BCP device, while there is a far smaller iodine excess for the ZnO
device. This result suggests that the inorganic HBL not only blocks iodide from reaching the metal
electrode, but also prevents Ag from penetrating into the PCBM layer during the thermal
evaporation of the electrode, where in turn it could also react with I ions. This mechanism has also
been reported to be detrimental to the device performance.30 These results confirm that the
7
incorporation of an inorganic hole-blocking layer can suppress ion migration induced electrode
degradation, while the organic HBL allowed for this degradation to take place.
Figure 4 displays the degradation behavior of the BCP devices under exposure to AM 1.5G solar
illumination in the three different atmospheres: N2, dry air, and humidified N2 (30% RH) with an
additional period at 10-12 hours where the cells were left to rest in the dark to investigate possible
performance recovery.14 We chose to compare three different degradation environments to
eliminate the possibility that the difference in degradation dynamics would be associated with the
different HBL, rather than the ion induced electrode degradation. Devices were held in the open
circuit condition while not being measured. During the fourteen-hour measurement period we
observe a moderate to severe decrease across all PV parameters, with VOC being the least affected
and JSC the strongest. For the inert and dry air environments, VOC retains above 90% of its initial
value after 14 hours. The devices exposed to water dipped below 90% of VOC after approximately
3 hours of measurement. JSC and FF plummet in all cases, falling to ~ 60% and 20% of their initial
values for N2/dry air and 30% RH respectively, and are the primary reasons for the steep loss in
PCE. Interestingly, there appears to be little difference between the degradation behavior in inert
and dry air environments, with devices exposed to humidity degrading markedly faster than both.
Additionally, exposure to humidity appears to affect the various pixels on each device differently,
as evidenced by the jagged trend lines and significantly wider error margins.
The degradation behavior for the experimental devices employing ZnO as HBL, under identical
degradation conditions, is displayed in Figure 5. Overall, the results are very similar to the BCP
devices, with a relatively stable VOC and strongly degraded JSC and FF. Just like the reference
devices, the effects under N2 and dry air atmospheres are identical, and exposure to water again
degrades the devices both more quickly as well as less uniformly. The overall amount of
8
degradation is slightly reduced for the ZnO devices, with the PCE falling to roughly 40% rather
than 30% of their initial values. This is, however, only a slight improvement and could very easily
be due to natural variation between devices.
Combining the degradation behavior with the XPS measurements of the electrodes yields two main
conclusions. First, degradation for devices under constant illumination with the architecture
ITO/PEDOT:PSS/CH3NH3PbI3/PCBM/[BCP or ZnO]/Ag is not driven by an oxygen-related
mechanism, as devices in both inert and dry air atmospheres displayed similar degradation
characteristics. Aristidou et al. proposed that oxygen related perovskite decomposition proceeds
by the generation of photoexcited electrons within the perovskite crystal.3 Since PCBM is known
to extract electrons from MAPbI3 on shorter timescales than chemical reactions can occur,31 the
presence of PCBM at the perovskite interface possibly prevents oxygen-related degradation by
withdrawing photoexcited electrons rapidly after their formation. The presence of moisture
accelerates the degradation, regardless of the HBL used. With several vulnerable components, the
root cause of the rapid decay in moisture is difficult to isolate. For example, while water is known
to have a significant impact on perovskite stability2,32 -- 34 and also might facilitate the diffusion of
volatile products inside the perovskite,26,35 PCBM has also been shown to undergo strong,
irreversible degradation in the presence of water.36
The second overall conclusion is that in contrast to the prevailing view, ZnO HBLs do not
significantly alter the degradation characteristics when compared to the more common HBL
BCP.21 XPS measurements of degraded device electrodes did confirm that ZnO hinders the
reaction of iodide ions with the electrode on the timescale of the experiment, however there was
no meaningful difference between the ZnO and BCP device performance deterioration. Han et al.37
found similar iodide infiltration into the silver electrode through the hole transport layer Spiro-
9
OMeTAD in their devices, accompanied with a dramatic loss in PCE. They suggested that
replacement of the silver contact with a more chemically inert electrode, such as Cu or C, could
improve device stability by preventing this modification. Our results suggest that this iodide
infiltration, and subsequent modification of the electrode into a more insulating species, cannot be
the dominant factor in the observed overall performance loss. Lee et al. previously showed that
the removal and re-evaporation of the silver electrode in degraded devices did partially restore the
device's performance,19 however a significant reduction in JSC remained. This suggests that an
additional irreversible degradation mechanism is present.
In a study related to our work, Back et al. demonstrated that a layer of titanium suboxide (TiOx)
placed between the PCBM and silver electrode could effectively increase device stability.27
Indeed, on similar time scales (~10h) to our experiments, devices with a TiOx layer maintained ~
80% of their initial PCE, compared to 50% with our ZnO HBL. This seemingly contradictory result
can be explained by two factors. First, their reference devices contained no HBL on top of the
PCBM. It is widely known that directly evaporating metallic contacts onto organic films can
damage and thereby harm the film's stability.38 Therefore, the TiOx layer will increase device
stability simply by protecting the PCBM during thermal evaporation. Secondly, Back et al. held
their devices at the maximum power point (MPP), which has been shown to slow degradation
significantly when compared to devices held at VOC.39 Once this difference is accounted for, the
stability for devices containing TiOx and ZnO nanoparticle blocking layers is extremely similar.
In another study using a similar device structure, Akbulatov et al. demonstrated strongly enhanced
stability when replacing PCBM with a perylene diimide derivative in encapsulated devices
degraded in nitrogen under illumination. They attributed the strong degradation to the
10
accumulation of volatile methylammonium iodide within the PCBM layer, subsequent reaction of
iodide with the silver electrode, and the resulting formation of PbI2 inside the perovskite layer.20
Taken all these results together, a plausible explanation for the observed deterioration of the device
performance is obtained by considering the effects of ion migration not on the electrode, but
instead on the perovskite active layer. While the ZnO layer prevents iodide infiltration into the
electrode, the diffusion of volatile components out of the perovskite layer into PCBM should still
be possible (and would remain undetectable by XPS) and has been previously shown to initiate
significant degradation, e.g. by way of passivating the perovskite at the crystal grain
boundaries.20,26,29,40 Passivation at the boundaries in turn isolates each individual grain, leading to
difficulties in conduction and charge extraction. This conclusion is supported by the specific
mechanism of degradation observed in our study, with the primary drivers of PCE deterioration
being a loss of JSC and FF, the two parameters associated with conductivity and charge extraction
efficiency. Additionally, recently Zhao et al showed that ion induced degradation of charge
extraction layers can also be a major cause for device instabilities.41 Our results indicate that even
without the iodide migration induced degradation of the electrode, the above mentioned
degradation pathways can still take place and strongly affect device stability.
CONCLUSIONS
To summarize, we investigated the role of ion induced electrode degradation in inverted perovskite
photovoltaic devices under realistic operational conditions. We find that contrary to what was
previously suggested, suppressing degradation by blocking ion migration to the electrode does not
necessarily improve device stability significantly under operational conditions. While it has been
11
shown that this degradation mechanism plays a significant role in determining long-term dark
storage stability of devices, its significance is diminished once more prominent degradation
processes are present under full illumination. We propose that it is not ion-migration induced
electrode corrosion, but rather the diffusion of volatile products out of the perovskite active layer
and into the PCBM layer, that causes strong and irreversible degradation under operational
conditions. Further research is required to elucidate the exact role that ion migration plays in
affecting device performance and stability, however this work shows that the scope of such
research must encompass a thorough examination, under realistic operating conditions, of both the
perovskite active layer as well as the other device components.
12
FIGURES
(a)
(b)
(c)
Zinc Acetate
PCBM
Bathocuproine (BCP)
Figure 1: (a) Device architecture, (b) energy diagram, and (c) ETL/HBL materials used.
13
Figure 2: Photovoltaic (PV) parameters as a function of ZnO nanoparticle layer thickness. One
sample with BCP as HBL was fabricated in the same batch for direct comparison. Error bars
represent the standard deviation. An optimal value of 8 nm was found for ZnO nanoparticle layer's
thickness.
14
(a)
(b)
Figure 3: (a) XPS measurement of the surface of the Ag electrodes of fully degraded devices,
showing the I3d (iodine) signal. (b) Excess Iodine in the Ag/HBL/PCBM layer, obtained by XPS
conducted after etching via argon beam. After approximately 60s of etching the interface with the
perovskite film is reached.
15
Figure 4: Evolution of PV parameters (open-circuit voltage, short-circuit current, efficiency, and
fill factor) for devices with BCP as HBL, degraded in N2, dry air, and humidified N2 (30% RH)
atmospheres over 14 hours. From hours 10-12 the devices were left in the dark, in order to test the
reversibility of the degradation. The shaded region represents the standard deviation for the
measurement.
16
Figure 5: Evolution of PV parameters (open-circuit voltage, short-circuit current, efficiency, and
fill factor) for devices with ZnO as HBL, degraded in N2, dry air, and humidified N2 (30% RH)
atmospheres over 14 hours. From hours 10-12 the devices were left in the dark, in order to test the
reversibility of the degradation. The shaded region represents the standard deviation for the
measurement.
17
ASSOCIATED CONTENT
Supporting Information. Synthesis and characterization of ZnO nanoparticles, J-V curves of
fresh photovoltaic devices
AUTHOR INFORMATION
Corresponding Author
*email: [email protected]
Kirchhoff Institute for Physics and Centre for Advanced Materials, Im Neuenheimer Feld
227/225, 69120, Germany
Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval
to the final version of the manuscript. This work has received funding from the European
Research Council (ERC) under the European Union's Horizon 2020 research and innovation
programme (ERC Grant Agreement n° 714067, ENERGYMAPS).
ACKNOWLEDGMENT
The authors would like to kindly thank Prof. Uwe Bunz for providing access to the device
fabrication facilities. P.F. thanks the HGSFP for scholarship.
18
REFERENCES
(1) Research Cell Efficiency Records: https://www.nrel.gov/pv/assets/images/efficiency-
chart.png (accessed: July 2018).
(2) Leijtens, T.; Eperon, G. E.; Noel, N. K.; Habisreutinger, S. N.; Petrozza, A.; Snaith, H. J.
Stability of Metal Halide Perovskite Solar Cells. Adv. Energy Mater. 2015, 5, 1500963.
(3) Aristidou, N.; Sanchez-Molina, I.; Chotchuangchutchaval, T.; Brown, M.; Martinez, L.;
Rath, T.; Haque, S. A. The Role of Oxygen in the Degradation of Methylammonium Lead
Trihalide Perovskite Photoactive Layers. Angew. Chemie - Int. Ed. 2015, 54, 8208 -- 8212.
(4) Sun, Q.; Fassl, P.; Becker-Koch, D.; Bausch, A.; Rivkin, B.; Bai, S.; Hopkinson, P. E.;
Snaith, H. J.; Vaynzof, Y. Role of Microstructure in Oxygen Induced Photodegradation of
Methylammonium Lead Triiodide Perovskite Films. Adv. Energy Mater. 2017, 7, 1700977.
(5) Bryant, D.; Aristidou, N.; Pont, S.; Sanchez-Molina, I.; Chotchunangatchaval, T.; Wheeler,
S.; Durrant, J. R.; Haque, S. A. Light and Oxygen Induced Degradation Limits the
Operational Stability of Methylammonium Lead Triiodide Perovskite Solar Cells. Energy
Environ. Sci. 2016, 9, 1655 -- 1660.
(6) Spina, M.; Karimi, A.; Andreoni, W.; Pignedoli, C. A.; Náfrádi, B.; Forró, L.; Horváth, E.
Mechanical Signatures of Degradation of the Photovoltaic Perovskite CH3NH3PbI3 upon
Water Vapor Exposure. Appl. Phys. Lett. 2017, 110, 121903.
(7) Wang, Q.; Chen, B.; Liu, Y.; Deng, Y.; Bai, Y.; Dong, Q.; Huang, J. Scaling Behavior of
Moisture-Induced Grain Degradation in Polycrystalline Hybrid Perovskite Thin Films.
Energy Environ. Sci. 2017, 10, 516 -- 522.
19
(8) Yang, J.; Siempelkamp, B. D.; Liu, D.; Kelly, T. L. Investigation of CH3NH3PbI3
Degradation Rates and Mechanisms in Controlled Humidity Environments Using in Situ
Techniques. ACS Nano 2015, 9, 1955 -- 1963.
(9) Dong, X.; Fang, X.; Lv, M.; Lin, B.; Zhang, S.; Ding, J.; Yuan, N. Improvement of the
Humidity Stability of Organic -- inorganic Perovskite Solar Cells Using Ultrathin Al 2 O 3
Layers Prepared by Atomic Layer Deposition. J. Mater. Chem. A 2015, 3, 5360 -- 5367.
(10) Conings, B.; Drijkoningen, J.; Gauquelin, N.; Babayigit, A.; D'Haen, J.; D'Olieslaeger, L.;
Ethirajan, A.; Verbeeck, J.; Manca, J.; Mosconi, E.; et al. Intrinsic Thermal Instability of
Methylammonium Lead Trihalide Perovskite. Adv. Energy Mater. 2015, 5, 1500477.
(11) Dualeh, A.; Gao, P.; Seok, S. Il; Nazeeruddin, M. K.; Grätzel, M. Thermal Behavior of
Methylammonium Lead-Trihalide Perovskite Photovoltaic Light Harvesters. Chem. Mater.
2014, 26, 6160 -- 6164.
(12) Li, B.; Li, Y.; Zheng, C.; Gao, D.; Huang, W. Advancements in the Stability of Perovskite
Solar Cells: Degradation Mechanisms and Improvement Approaches. RSC Adv. 2016, 6,
38079 -- 38091.
(13) Wang, D.; Wright, M.; Elumalai, N. K.; Uddin, A. Stability of Perovskite Solar Cells. Sol.
Energy Mater. Sol. Cells 2016, 147, 255 -- 275.
(14) Nie, W.; Blancon, J. C.; Neukirch, A. J.; Appavoo, K.; Tsai, H.; Chhowalla, M.; Alam, M.
A.; Sfeir, M. Y.; Katan, C.; Even, J.; et al. Light-Activated Photocurrent Degradation and
Self-Healing in Perovskite Solar Cells. Nat. Commun. 2016, 7, 11574.
(15) Xiao, Z.; Yuan, Y.; Shao, Y.; Wang, Q.; Dong, Q.; Bi, C.; Sharma, P.; Gruverman, A.;
20
Huang, J. Giant Switchable Photovoltaic Effect in Organometal Trihalide Perovskite
Devices. Nat. Mater. 2015, 14, 193 -- 197.
(16) Li, C.; Guerrero, A.; Zhong, Y.; Gräser, A.; Luna, C. A. M.; Köhler, J.; Bisquert, J.; Hildner,
R.; Huettner, S. Real-Time Observation of Iodide Ion Migration in Methylammonium Lead
Halide Perovskites. Small 2017, 13, 1701711.
(17) Guerrero, A.; You, J.; Aranda, C.; Kang, Y. S.; Garcia-Belmonte, G.; Zhou, H.; Bisquert,
J.; Yang, Y. Interfacial Degradation of Planar Lead Halide Perovskite Solar Cells. ACS
Nano 2016, 10, 218 -- 224.
(18) Tress, W.; Marinova, N.; Moehl, T.; Zakeeruddin, S. M.; Nazeeruddin, M. K.; Grätzel, M.
Understanding the Rate-Dependent J -- V Hysteresis, Slow Time Component, and Aging in
CH3NH3PbI3 Perovskite Solar Cells: The Role of a Compensated Electric Field. Energy
Environ. Sci. 2015, 8, 995 -- 1004.
(19) Lee, H.; Lee, C. Analysis of Ion-Diffusion-Induced Interface Degradation in Inverted
Perovskite Solar Cells via Restoration of the Ag Electrode. Adv. Energy Mater. 2018, 8,
1702197.
(20) Akbulatov, A. F.; Frolova, L. A.; Griffin, M. P.; Gearba, I. R.; Dolocan, A.; Vanden Bout,
D. A.; Tsarev, S.; Katz, E. A.; Shestakov, A. F.; Stevenson, K. J.; et al. Effect of Electron-
Transport Material on Light-Induced Degradation of Inverted Planar Junction Perovskite
Solar Cells. Adv. Energy Mater. 2017, 7, 1700476.
(21) Kato, Y.; Ono, L. K.; Lee, M. V.; Wang, S.; Raga, S. R.; Qi, Y. Silver Iodide Formation in
Methyl Ammonium Lead Iodide Perovskite Solar Cells with Silver Top Electrodes. Adv.
21
Mater. Interfaces 2015, 2, 1500195.
(22) Yuan, Y.; Wang, Q.; Shao, Y.; Lu, H.; Li, T.; Gruverman, A.; Huang, J. Electric-Field-
Driven Reversible Conversion between Methylammonium Lead Triiodide Perovskites and
Lead Iodide at Elevated Temperatures. Adv. Energy Mater. 2016, 6, 1501803.
(23) Yang, D.; Ming, W.; Shi, H.; Zhang, L.; Du, M. H. Fast Diffusion of Native Defects and
Impurities in Perovskite Solar Cell Material CH3NH3PbI3. Chem. Mater. 2016, 28, 4349 --
4357.
(24) You, J.; Meng, L.; Song, T. Bin; Guo, T. F.; Chang, W. H.; Hong, Z.; Chen, H.; Zhou, H.;
Chen, Q.; Liu, Y.; et al. Improved Air Stability of Perovskite Solar Cells via Solution-
Processed Metal Oxide Transport Layers. Nat. Nanotechnol. 2016, 11, 75 -- 81.
(25) Bai, S.; Wu, Z.; Wu, X.; Jin, Y.; Zhao, N.; Chen, Z.; Mei, Q.; Wang, X.; Ye, Z.; Song, T.;
et al. High-Performance Planar Heterojunction Perovskite Solar Cells: Preserving Long
Charge Carrier Diffusion Lengths and Interfacial Engineering. Nano Res. 2014, 7, 1749 --
1758.
(26) Brinkmann, K. O.; Zhao, J.; Pourdavoud, N.; Becker, T.; Hu, T.; Olthof, S.; Meerholz, K.;
Hoffmann, L.; Gahlmann, T.; Heiderhoff, R.; et al. Suppressed Decomposition of
Organometal Halide Perovskites by Impermeable Electron-Extraction Layers in Inverted
Solar Cells. Nat. Commun. 2017, 8, 13938.
(27) Back, H.; Kim, G.; Kim, J.; Kong, J.; Kim, T. K.; Kang, H.; Kim, H.; Lee, J.; Lee, S.; Lee,
K. Achieving Long-Term Stable Perovskite Solar Cells via Ion Neutralization. Energy
Environ. Sci. 2016, 9, 1258 -- 1263.
22
(28) Pacholski, C.; Kornowski, A.; Weller, H. Self-Assembly of ZnO: From Nanodots to
Nanorods. Angew. Chemie - Int. Ed. 2002, 41, 1188 -- 1191.
(29) Li, J.; Dong, Q.; Li, N.; Wang, L. Direct Evidence of Ion Diffusion for the Silver-Electrode-
Induced Thermal Degradation of Inverted Perovskite Solar Cells. Adv. Energy Mater. 2017,
7, 1602922.
(30) Zhang, T.; Meng, X.; Bai, Y.; xiao, S.; Hu, C.; Yang, Y.-L.; Chen, H.; Yang, S. Profiling
the Organic Cation-Dependent Degradation of Organolead Halide Perovskite Solar Cells.
J. Mater. Chem. A 2016, 5, 1103 -- 1111.
(31) Chueh, C.-C.; Li, C.-Z.; Jen, A. K.-Y. Recent Progress and Perspective in Solution-
Processed Interfacial Materials for Efficient and Stable Polymer and Organometal
Perovskite Solar Cells. Energy Environ. Sci. 2015, 8, 1160 -- 1189.
(32) Frost, J. M.; Butler, K. T.; Brivio, F.; Hendon, C. H.; Van Schilfgaarde, M.; Walsh, A.
Atomistic Origins of High-Performance in Hybrid Halide Perovskite Solar Cells. Nano Lett.
2014, 14, 2584 -- 2590.
(33) Kang, S. M.; Ahn, N.; Lee, J.-W.; Choi, M.; Park, N.-G. Water-Repellent Perovskite Solar
Cell. J. Mater. Chem. A 2014, 2, 20017 -- 20021.
(34) Huang, J.; Tan, S.; Lund, P.; Zhou, H. Impact of H2O on Organic-Inorganic Hybrid
Perovskite Solar Cells. Energy Environ. Sci. 2017, 10, 2284 -- 2311.
(35) Ma, C.; Shen, D.; Qing, J.; Thachoth Chandran, H.; Lo, M.-F.; Lee, C.-S. Effects of Small
Polar Molecules (MA + and H2O) on Degradation Processes of Perovskite Solar Cells. ACS
Appl. Mater. Interfaces 2017, 9, 14960 -- 14966.
23
(36) Bao, Q.; Liu, X.; Braun, S.; Fahlman, M. Oxygen- and Water-Based Degradation in [6,6]-
Phenyl-C61-Butyric Acid Methyl Ester (PCBM) Films. Adv. Energy Mater. 2014, 4,
1301272.
(37) Han, Y.; Meyer, S.; Dkhissi, Y.; Weber, K.; Pringle, J. M.; Bach, U.; Spiccia, L.; Cheng,
Y.-B. Degradation Observations of Encapsulated Planar CH3NH3PbI3 Perovskite Solar
Cells at High Temperatures and Humidity. J. Mater. Chem. A 2015, 3, 8139 -- 8147.
(38) Jørgensen, M.; Norrman, K.; Krebs, F. C. Stability/Degradation of Polymer Solar Cells. Sol.
Energy Mater. Sol. Cells 2008, 92, 686 -- 714.
(39) Domanski, K.; Alharbi, E. A.; Hagfeldt, A.; Tress, W. Systematic Investigation of the
Impact of Operation Conditions on the Degradation Behaviour of Perovskite Solar Cells.
Nat. Energy 2018, 3, 61 -- 67.
(40) Chen, Q.; Zhou, H.; Song, T.; Luo, S.; Hong, Z. Controllable Self-Induced Passivation of
Hybrid Lead Iodide Perovskites toward High Performance Solar Cells. Nano Lett. 2014, 14,
4158 -- 4163.
(41) Zhao, Y.; Zhou, W.; Tan, H.; Fu, R.; Li, Q.; Lin, F.; Yu, D.; Walters, G.; Sargent, E. H.;
Zhao, Q. Mobile-Ion-Induced Degradation of Organic Hole-Selective Layers in Perovskite
Solar Cells. J. Phys. Chem. C 2017, 121, 14517 -- 14523.
24
|
1708.00593 | 1 | 1708 | 2017-08-02T03:49:40 | Strongly exchange-coupled and surface-state-modulated magnetization dynamics in Bi2Se3/YIG heterostructures | [
"physics.app-ph"
] | We report strong interfacial exchange coupling in Bi2Se3/yttrium iron garnet (YIG) bilayers manifested as large in-plane interfacial magnetic anisotropy (IMA) and enhancement of damping probed by ferromagnetic resonance (FMR). The IMA and spin mixing conductance reached a maximum when Bi2Se3 was around 6 quintuple-layer (QL) thick. The unconventional Bi2Se3 thickness dependence of the IMA and spin mixing conductance are correlated with the evolution of surface band structure of Bi2Se3, indicating that topological surface states play an important role in the magnetization dynamics of YIG. Temperature-dependent FMR of Bi2Se3/YIG revealed signatures of magnetic proximity effect of $T_c$ as high as 180 K, and an effective field parallel to the YIG magnetization direction at low temperature. Our study sheds light on the effects of topological insulators on magnetization dynamics, essential for development of TI-based spintronic devices. | physics.app-ph | physics | Strongly exchange-coupled and surface-state-modulated magnetization
dynamics in Bi2Se3/YIG heterostructures
Y. T. Fanchiang1, K. H. M. Chen2, C. C. Tseng2, C. C. Chen2, C. K. Cheng1, C. N. Wu2, S. F.
Lee3*, M. Hong1*, and J. Kwo2*
1Department of Physics, National Taiwan University, Taipei 10617, Taiwan
2Department of Physics, National Tsing Hua University, Hsinchu 30013, Taiwan
3Institute of Physics, Academia Sinica, Taipei 11529, Taiwan
Abstract
We report strong interfacial exchange coupling in Bi2Se3/yttrium iron garnet (YIG)
bilayers manifested as large in-plane interfacial magnetic anisotropy (IMA) and enhancement
of damping probed by ferromagnetic resonance (FMR). The IMA and spin mixing
conductance reached a maximum when Bi2Se3 was around 6 quintuple-layer (QL) thick. The
unconventional Bi2Se3 thickness dependence of the IMA and spin mixing conductance are
correlated with the evolution of surface band structure of Bi2Se3, indicating that topological
surface states play an important role in the magnetization dynamics of YIG. Temperature-
dependent FMR of Bi2Se3/YIG revealed signatures of magnetic proximity effect of Tc as high
as 180 K, and an effective field parallel to the YIG magnetization direction at low temperature.
Our study sheds light on the effects of topological insulators on magnetization dynamics,
essential for development of TI-based spintronic devices.
Topological insulators (TIs) are emergent quantum materials hosting topologically
protected surface states, with dissipationless transport prohibiting backscattering [1,2]. Strong
spin-orbit coupling (SOC) along with time reversal symmetry (TRS) ensures that the electrons
in surface states have their direction of motion and spin "locked" to each other [1,3,4]. When
interfaced with a magnetic layer, the interfacial exchange coupling can induce magnetic order
in TIs and break TRS [5-8]. The resulting gap opening of the Dirac state is necessary to realize
novel phenomena such as topological magneto-electric effect [9] and quantum anomalous Hall
effect [10,11]. Another approach of studying a TI/ferromagnet (FM) system focuses on spin-
transfer characteristic at the interface and intends to exploit the helical spin texture of
topological surface states (TSSs). Attempts have been made to estimate the spin-charge
conversion (SCC) efficiency, either by using microwave-excited dynamical method [12-16]
(e.g. spin pumping and spin-torque FMR) or thermally induced spin injection [17]. Very large
values of SCC ratio have been reported [13,15,16]. Recently, TIs are shown to be excellent
sources of spin-orbit torques for efficient magnetization switching [18].
Since the magnetic proximity effect (MPE) and spin-transfer process rely on interfacial
exchange coupling of TI/FM, understanding the magnetism at the interface has attracted strong
interests in recent years. Several techniques have been adopted to investigate the interfacial
magnetic properties, including spin-polarized neutron reflectivity [7,19], second harmonic
generation [20], electrical transport [6,21], and magneto-optical Kerr effect [6]. All these
studies clearly indicate the existence of MPE resulting from exchange coupling and strong
spin-orbit interaction in TIs. For device application, however, it is equally important to
understand how the interfacial exchange coupling affects the magnetization dynamic of FM.
For example, TIs can introduce additional magnetic damping and greatly alter the dynamical
properties of FM layer, as commonly observed in FM/heavy metals systems [22-24]. The
enhanced damping is visualized as larger linewidth of FMR spectra [22-24]. In TI/FM, the
presence of TSS and, possibly, MPE complicate the system under study, and is still an open
question to answer.
In this work we systematically investigated FMR characteristic of the ferrimagnetic
insulator YIG under the influence of the prototypical three-dimensional TI Bi2Se3 [25]. We
choose YIG as the FM layer because of its technological importance, with high 𝑇𝑐 ~550 K
and extremely low damping coefficient α [26]. When YIG is interfaced with TIs, its good
thermal stability minimizes interdiffusion of materials. Through Bi2Se3 thickness dependence
study, we observed strong modulation of FMR properties, attributed to the TSS of Bi2Se3.
Temperature-dependent study unraveled an effective field parallel to the magnetization
direction existing in Bi2Se3/YIG. Such effective field built up as temperature decreased, which
was utilized to demonstrate the zero-applied-field FMR of YIG. Furthermore, we identified the
signature of MPE of 𝑇𝑐 as high as 180 K manifested as enhanced spin pumping in a fluctuating
spin system.
Bi2Se3 thin films were grown by molecular beam epitaxy (MBE) [27] on magnetron-
sputtered YIG films. The high quality of Bi2Se3/YIG samples in this work is verified by high
resolution X-ray diffraction (XRD) and transmission electron microscope (TEM) [28]. To
investigate the magnetic properties of Bi2Se3/YIG, room-temperature angle- and frequency-
dependent FMR measurements were performed independently using a cavity and co-planar
waveguide (CPW), respectively (Figure 1(a) and (b)). For the temperature-dependent FMR,
the CPW was mounted in a cryogenic probe station (Lake Shore, CPX-HF), which enable
samples to be cooled as low as 5 K. The external field is modulated for lock-in detection in all
of the measurements.
The FMR spectra in Figure 1(c) are compared for single layer YIG(12) and
Bi2Se3(25)/YIG(12) bilayer (digits denote thickness in nanometer), showing a large shift of
resonance field (𝐻𝑟𝑒𝑠) ~317 Oe after the Bi2Se3 growth plus a markedly broadened peak-to-
peak width ∆𝐻 for Bi2Se3/YIG. Figure 1(d) shows 𝐻𝑟𝑒𝑠 vs applied field angle with respect
to surface normal 𝜃𝐻 for YIG(12) and Bi2Se3(25)/YIG(12). The larger variation of 𝐻𝑟𝑒𝑠 with
𝜃𝐻 indicates stronger magnetic anisotropy in the bilayer sample. When the applied field was
directed in the film plane, clear negative 𝐻𝑟𝑒𝑠 shifts induced by Bi2Se3 were observed at all
microwave frequency f in Figure 1e. The data can be fitted in the scheme of magnetic thin films
having uniaxial perpendicular magnetic anisotropy [28]. Since the XRD results show that the
YIG films did not gain additional strain after growing Bi2Se3, the enhanced anisotropy cannot
be attributed to changes of magnetocrstalline anisotropy. Defining eff ective demagnetization
field 4𝜋𝑀𝑒𝑓𝑓 = 4𝜋𝑀𝑠 − 𝐻𝑎𝑛 − 𝐻𝑖𝑛𝑡, where 4𝜋𝑀𝑠, 𝐻𝑎𝑛 and 𝐻𝑖𝑛𝑡 are the demagnetization
field of bare YIG, the magnetocrystalline anisotropy field of YIG, and the anisotropy field
induced by Bi2Se3, we obtain 𝐻𝑖𝑛𝑡 = -926 and -1005 Oe from Figure 1(d) and 1(e),
respectively. The minus sign indicates the additional anisotropy points in the film plane.
The above observations suggested the presence of IMA in Bi2Se3/YIG. To verify this
assumption, we systematically varied the thickness of YIG 𝑑𝑌𝐼𝐺 while fixing the thickness of
Bi2Se3. Figure 2(a) presents the 𝑑𝑌𝐼𝐺 dependence of 4𝜋𝑀𝑒𝑓𝑓 for single and bilayer samples.
The 4𝜋𝑀𝑒𝑓𝑓 of single layer YIG was independent of 𝑑𝑌𝐼𝐺 varying from 12 to 30 nm. In
sharp contrast, 4𝜋𝑀𝑒𝑓𝑓 of Bi2Se3/YIG became significantly larger, especially at thinner YIG,
which is a feature of an interfacial effect. The f- and 𝜃𝐻-dependent FMR were performed
independently to doubly confirm the trends. The IMA can be further characterized by defining
the effective anisotropy constant 𝐾𝑒𝑓𝑓 = (1/2)4𝜋𝑀𝑒𝑓𝑓𝑀𝑠 = (1/2)(4𝜋𝑀𝑠 − 𝐻𝑎𝑛) + 𝐾𝑖/
𝑑𝑌𝐼𝐺, with the interfacial anisotropy constant 𝐾𝑖 = 𝑀𝑠𝐻𝑎𝑛𝑑𝑌𝐼𝐺/2. The 𝐾𝑒𝑓𝑓𝑑𝑌𝐼𝐺 vs 𝑑𝑌𝐼𝐺
data in Figure 2(b) is well fitted by a linear function, indicating that the 𝑑𝑌𝐼𝐺 dependence
presented in Figure 2(a) is suitably described by the current form of 𝐾𝑒𝑓𝑓 . The intercept
obtained by extrapolating the linear function corresponds to 𝐾𝑖 = −0.075 erg/cm2.
To further investigate the physical origin of the IMA, we next varied the thickness of
Bi2Se3 (𝑑𝐵𝑆) to see how 𝐾𝑖 evolved with 𝑑𝐵𝑆. Figure 2(c) shows the 𝑑𝐵𝑆 dependence of 𝐾𝑖.
Starting from 𝑑𝐵𝑆 = 40 nm sample, the magnitude of 𝐾𝑖 went up as 𝑑𝐵𝑆 decreased. An
extremum of 𝐾𝑖 − 0.12 ± 0.02 erg/cm-2 was reached at 𝑑𝐵𝑆 = 7 nm. An abrupt upturn of Ki
occurred in the region 3 nm < 𝑑𝐵𝑆 < 7 nm . The Ki magnitude dropped drastically and
exhibited a sign change in the interval. The Ki value of 0.014 erg/cm2 at 𝑑𝐵𝑆 = 3 nm
corresponds to weak interfacial perpendicular anisotropy. Based on previous investigation on
surface band structure of ultrathin Bi2Se3 [29], 𝑑𝐵𝑆 = 6 nm was identified as the 2D quantum
tunneling limit of Bi2Se3. When 𝑑𝐵𝑆 < 6 nm , the hybridization of top and bottom TSS
developed a gap of the surface states. Spin-resolved photoemission study later showed that the
TSS in this regime exhibited decreased in-plane spin polarization. The modulated spin texture
may lead to the weaker IMA than that in 3D regime [30,31]. We thus divide Figure 2(c) into
two regions and correlate the systematic magnetic properties with the surface state band
structure. The sharp change of 𝐾𝑖 around 𝑑𝐵𝑆 = 6 nm strongly suggests that the IMA in
Bi2Se3/YIG is of topological origins.
The ∆𝐻 broadening in FMR spectra after growing Bi2Se3 on YIG indicates additional
damping arising from spin-transfer from YIG to Bi2Se3, which is a signature of spin pumping
effect [24]. The spin pumping efficiency of an interface can be evaluated by spin mixing
conductance 𝑔↑↓, using the following relation [24],
𝑔↑↓ =
4𝜋𝑀𝑠𝑑𝑌𝐼𝐺
𝑔𝜇𝐵
(𝛼𝐵𝑆/𝑌𝐼𝐺 − 𝛼𝑌𝐼𝐺) (1)
, where 𝑔 , 𝜇𝐵 , 𝛼𝐵𝑆/𝑌𝐼𝐺 and 𝛼𝑌𝐼𝐺 are the Landé g factor, Bohr magneton, the damping
coefficient of Bi2Se3/YIG and YIG, respectively. Figure 2(d) displays the dBS dependence of
effective spin mixing conductance of Bi2Se3/YIG. Similar to Ki in Figure 2(c), 𝑔↑↓ had its
maximum at 𝑑𝐵𝑆 = 7 nm with a very large value ~2.2 × 1015 cm-2, about four times larger
than that of our Pt/YIG sample indicated by the red dashed line. The inset shows ∆𝐻 vs 𝑓
data for Bi2Se3 (7)/YIG(13) and YIG(13) fitted by linear functions. One can clearly see a
significant change of slope, from which we determined 𝛼𝐵𝑆/𝑌𝐼𝐺 − 𝛼𝑌𝐼𝐺 to be 0.014. The large
𝑔↑↓ of Bi2Se3/YIG implies an efficient spin pumping to an excellent spin sink of Bi2Se3. We
again divide Figure 2(d) into two regions according to the surface state band structure. Upon
crossing the 2D limit from 𝑑𝐵𝑆 = 7 nm, 𝑔↑↓ dropped remarkably to 1.7 × 1014 at 𝑑𝐵𝑆 =
3 nm. The trend in Figure 2(d) is distinct from that of normal metal (NM)/FM structures. In
NM/FM, the 𝑔↑↓ increases with increasing NM thickness as a result of vanishing spin
backflow in thicker NM [32]. It is worth noting that the conducting bulk of Bi2Se3 can dissipate
the spin-pumping-induced spin accumulation at the interface [12,33]. In this regard, the 𝑑𝐵𝑆 =
7 nm sample has the largest weight of surface state contribution to 𝑔↑↓. Such unconventional
𝑑𝐵𝑆 dependence of 𝑔↑↓ suggests that TSS play a dominant role in the damping enhancement.
Since the effects of TSS are expected to enhance at low temperature, we next performed
temperature-dependent FMR on Bi2Se3/YIG. Two bilayer samples Bi2Se3(25)/YIG(15) and
Bi2Se3(16)/YIG(17), and a single layer YIG(23) were measured for comparison. Figure 3(a)
and (b) show the 𝐻𝑟𝑒𝑠 vs 𝑓 data at various T for YIG(23) and Bi2Se3(25)/YIG(15). The 𝐻𝑟𝑒𝑠
of both samples show negative shifts at all f with decreasing T. The data of YIG(23) can be
reproduced by the Kittel equation with increasing 𝑀𝑠 of YIG at low T. In sharp contrast,
Bi2Se3(25)/YIG(15) exhibited negative intercepts at 𝐻𝑟𝑒𝑠 , and the intercepts gained its
magnitude when the sample was cooled down. This behavior of non-zero intercept is common
for all of our Bi2Se3/YIG samples. To account for the behavior, a phenomenological effective
field 𝐻𝑒𝑓𝑓 is added to the Kittel equation, i.e.
𝑓 =
𝛾
2𝜋
√(𝐻𝑟𝑒𝑠 + 𝐻𝑒𝑓𝑓)(𝐻𝑟𝑒𝑠 + 𝐻𝑒𝑓𝑓 + 4𝜋𝑀𝑒𝑓𝑓). (2)
The solid lines in Figure 3(b) generated by the modified Kittel equation fitted the experimental
data very well.
Figures 3(c) and (d) presents the T dependence of 𝐻𝑟𝑒𝑠 and ∆𝐻 for the YIG(23) and
two Bi2Se3/YIG samples. As we lowered T, all of the samples had decreasing 𝐻𝑟𝑒𝑠, which was
viewed as effects of the concurrent increasing 𝑀𝑒𝑓𝑓 and 𝐻𝑒𝑓𝑓 as seen in Figure 3(a) and (b).
On the other hand, ∆𝐻 built up with decreasing T. We first examine ∆𝐻 of the YIG(23)
single layer. The ΔH remained relatively unchanged with T decreasing from RT, and
dramatically increased below 100 K. The pronounced T dependence of ∆𝐻 or α has been
explored in various rare-earth iron garnet and was explained by the low T slow-relaxation
process via rare-earth elements or Fe2+ impurities [34]. For sputtered YIG films, specifically,
the increase of ∆𝐻 was less prominent in thicker YIG, indicating that the dominant impurities
located near the YIG surface [35]. Distinct from that of YIG(23), the ∆𝐻 progressively
increased for the bilayer samples. We were not able to detect FMR signals with ∆𝐻 beyond
100 Oe due to our instrumental limits. However, one can clearly see that, for
Bi2Se3(25)/YIG(15) and Bi2Se3(16)/YIG(17), the ∆𝐻 broadened due to increased spin
pumping at first. For Bi2Se3(25)/YIG(15), the ∆𝐻 curve gradually leveled off, and intersected
with that of YIG(23) at T ~40 K. The seemingly "anti-damping" by Bi2Se3 at low T may be
related to the modification of the YIG surface chemistry during the Bi2Se3 deposition.
Additional analyses are needed to verify the scenario, which is, however, beyond the scope of
this work.
In both 𝐻𝑟𝑒𝑠 and ∆𝐻 curves, bump-like features located at T = 140 and 180 K (indicated
by the arrows) were revealed for Bi2Se3(25)/YIG(15) and Bi2Se3(16)/YIG(17), respectively.
The bumps are reminiscent of spin pumping in the case of fluctuating magnets, where an
enhancement of spin pumping is expected as the spin sink is closed to its magnetic phase
transition point [36-38]. In our system, a possibly newly formed magnetic phase would be the
interfacial magnetization driven by the proximity effect, namely, Tc = 140 and 180 K for our
Bi2Se3(25)/YIG(15) and Bi2Se3(16)/YIG(17), respectively. In fact, the Tc values of our samples
are in good agreement with the reported ones of MPE in TI/YIG systems [6,21]. We did not
detect anomalous Hall effect in our samples, which might be obscured by the bulk conduction
of Bi2Se3 in the magneto-transport measurements.
Using Eq. (2), we further determine the T dependence of 4𝜋𝑀𝑒𝑓𝑓 and 𝐻𝑒𝑓𝑓 of YIG(23)
and the two bilayers samples, as shown in Figures 3(e) and (f) [28]. The 4𝜋𝑀𝑒𝑓𝑓 of YIG(23)
went larger monotonically as previously discussed, while the 4𝜋𝑀𝑒𝑓𝑓 of bilayer samples
increased before reaching a maximum of 4000 Oe at T around 150 K, and then decreased
slightly at low T. With 4𝜋𝑀𝑒𝑓𝑓
𝐵𝑆/𝑌𝐼𝐺 − 4𝜋𝑀𝑒𝑓𝑓
𝑌𝐼𝐺 ≈ −𝐻𝑖𝑛𝑡, such T dependence corresponds to
decreasing in-plane IMA below 150 K. Note that 150 K is closed to our assumed 𝑇𝑐 from
MPE at 140 and 180 K. Calculations of total electronic energy at an TI/FM interface show that
perpendicular anisotropy is in favor [39], which, in our case, may effectively weaken the in-
plane IMA. The decreasing in-plane IMA can also be viewed as a result of modified spin texture
by MPE. The analyses thus provide another clue supporting the existence of MPE in our
samples. The 𝐻𝑒𝑓𝑓 of bilayer samples, again, show different T evolution than that of the single
layer in Figure 3(f). 𝐻𝑒𝑓𝑓 built up with decreasing T in bilayers while the 𝐻𝑒𝑓𝑓 of the YIG
single layer was T-independent and closed to zero. The positive 𝐻𝑒𝑓𝑓 corresponds to an
effective field parallel to the magnetization vector M and resembles the exchange bias field.
However, we did not observe shifts of magnetization hysteresis loop characteristic of an
exchange bias effect [28,40]. The 𝐻𝑒𝑓𝑓 present in FMR measurement suggests it may come
from spin imbalance at the interface [41].
Finally, we demonstrated that the large IMA and 𝐻𝑒𝑓𝑓 in Bi2Se3/YIG are strong enough
to induce FMR without an 𝐻𝑒𝑥𝑡 , which we term zero-field FMR. Figure 4(a) displays T
evolution of FMR first derivative spectra of Bi2Se3(25)/YIG(15) at 𝑓 = 3.5 GHz. The spectral
shape started to deform when the 𝐻𝑟𝑒𝑠 was approaching zero. The sudden twists at 𝐻𝑒𝑥𝑡 ~
+30 (-30) for positive (negative) field sweep arose from magnetization switching of YIG, and
therefore led to hysteric spectra. The two spectra merged at 25 K and then separated again when
T was further decreased. Figure 4(b) shows the microwave absorption intensity I spectra with
positive field sweeps. We traced the position of I spectrum 𝐻𝑝𝑒𝑎𝑘 using the red dashed line,
and found it coincided with zero 𝐻𝑒𝑥𝑡 at the zero-field FMR temperature 𝑇0 ~25 K. Below
25 K, 𝐻𝑝𝑒𝑎𝑘 moved across the origin and one needed to reverse 𝐻𝑒𝑥𝑡 to counter the internal
effective field comprised of the demagnetization field 4𝜋𝑀𝑠, 𝐻𝑖𝑛𝑡 and 𝐻𝑒𝑓𝑓 (Figure 4(e)).
Note that the presence of 𝐻𝑖𝑛𝑡 alone would be inadequate to realize zero-field FMR. Only
when 𝐻𝑒𝑓𝑓 is finite would the system exhibit non-zero intercepts as we have seen in Figure
3(b). We further calculate 𝑇0 as a function of microwave excitation frequency f (Figure 4(f))
using Eq. (2) and the extracted 𝐻𝑒𝑓𝑓 of Figure 3(f). We obtain that, with finite 𝐻𝑒𝑓𝑓
persisting up to room temperature, zero-field FMR can be realized at high T provided f is
sufficiently low. However, we emphasize that it's advantageous for YIG to be microwave-
excited above 3 GHz. When 𝑓 < 3 GHz , parasitic effects such as three-magnon splitting
[42,43] take place and significantly decrease the microwave absorption in YIG. Here, we
demonstrate that the strong exchange coupling between Bi2Se3 and YIG gave rise to zero-field
FMR in the feasible high frequency operation regime of YIG. Further improvement of interface
quality of Bi2Se3/YIG is expected to raise 𝐻𝑒𝑓𝑓 and 𝑇0 for room temperature, field free
spintronic application.
In summary, we investigate the magnetization dynamics of YIG in the presence of
interfacial exchange coupling and TSS of Bi2Se3. The significantly modulated magnetization
dynamics at room temperature is shown to be TSS-originated through the Bi2Se3 thickness
dependence study. It can be expected that the strong coupling between Bi2Se3 and YIG will
modify interface band structures and even the spin texture of TSS. The underlying mechanism
of the coexisting large in-plane IMA and pronouncedly enhanced damping calls for further
theoretical modeling and understanding. The temperature-dependent FMR results revealed rich
information, including signatures of MPE and increasing exchange effective field at low
temperature. Our study is an important step toward realization of topological spintronics.
We would like to thank helpful discussion with Profs. Mingzhong Wu and Hsin Lin. The
work is supported by MOST 105-2112-M-007-014-MY3, 105-2112-M-002-022-MY3, 106-
2112-M-002-010-MY3, 102-2112-M-002-022-MY3, and 105-2112-M-001-031-MY3 of the
Ministry of Science and Technology in Taiwan.
References
[1] M. Z. Hasan and C. L. Kane, Rev. Mod. Phys. 82, 3045 (2010).
[2]
[3]
X. L. Qi and S. C. Zhang, Rev. Mod. Phys. 83, 1057 (2011).
D. Hsieh, Y. Xia, D. Qian, L. Wray, J. H. Dil, F. Meier, J. Osterwalder, L. Patthey, J. G.
Checkelsky, N. P. Ong, A. V. Fedorov, H. Lin, A. Bansil, D. Grauer, Y. S. Hor, R. J.
Cava, and M. Z. Hasan, Nature 460, 1101 (2009).
[4]
S. Xu, Y. Xia, L. A. Wray, S. Jia, F. Meier, J. H. Dil, J. Osterwalder, B. Slomski, A.
Bansil, H. Lin, R. J. Cava, and M. Z. Hasan, Science 332, 560 (2011).
[5]
P. Wei, F. Katmis, B. A. Assaf, H. Steinberg, P. Jarillo-Herrero, D. Heiman, and J. S.
Moodera, Phys. Rev. Lett. 110, 186807 (2013).
[6] M. Lang, M. Montazeri, M. C. Onbasli, X. Kou, Y. Fan, P. Upadhyaya, K. Yao, F. Liu,
Y. Jiang, W. Jiang, K. L. Wong, G. Yu, J. Tang, T. Nie, L. He, R. N. Schwartz, Y. Wang,
C. A. Ross, and K. L. Wang, Nano Lett. 14, 3459 (2014).
[7]
F. Katmis, V. Lauter, F. S. Nogueira, B. A. Assaf, M. E. Jamer, P. Wei, B. Satpati, J. W.
Freeland, I. Eremin, D. Heiman, P. Jarillo-Herrero, and J. S. Moodera, Nature 533, 513
(2016).
[8]
Y. L. Chen, J. H. Chu, J. G. Analytis, Z. K. Liu, K. Igarashi, H. H. Kuo, X. L. Qi, S. K.
Mo, R. G. Moore, D. H. Lu, M. Hashimoto, T. Sasagawa, S. C. Zhang, I. R. Fisher, Z.
Hussain, and Z. X. Shen, Science 329, 659 (2010).
[9]
I. Garate and M. Franz, Phys. Rev. Lett. 104, 146802 (2010).
[10] C. Z. Chang, Science 340, 167 (2013).
[11] R. Yu, W. Zhang, H. J. Zhang, S. C. Zhang, X. Dai, and Z. Fang, Science 61, 329 (2010).
[12] Y. Shiomi, K. Nomura, Y. Kajiwara, K. Eto, M. Novak, K. Segawa, Y. Ando, and E.
Saitoh, Phys. Rev. Lett. 113, 196601 (2014).
[13]
J. C. Rojas-Sánchez, S. Oyarzún, Y. Fu, A. Marty, C. Vergnaud, S. Gambarelli, L. Vila,
M. Jamet, Y. Ohtsubo, A. Taleb-Ibrahimi, P. Le Fèvre, F. Bertran, N. Reyren, J. M.
George, and A. Fert, Phys. Rev. Lett. 116, 096602 (2016).
[14] H. Wang, J. Kally, J. S. Lee, T. Liu, H. Chang, D. R. Hickey, K. A. Mkhoyan, M. Wu,
A. Richardella, and N. Samarth, Phys. Rev. Lett. 117, 076601 (2016).
[15] A. R. Mellnik, J. S. Lee, A. Richardella, J. L. Grab, P. J. Mintun, M. H. Fischer, A.
Vaezi, A. Manchon, E. A. Kim, N. Samarth, and D. C. Ralph, Nature 511, 449 (2014).
[16] M. Jamali, J. S. Lee, J. S. Jeong, F. Mahfouzi, Y. Lv, Z. Zhao, B. K. Nikolic, K. A.
Mkhoyan, N. Samarth, and J. P. Wang, Nano Lett. 15, 7126 (2015).
[17] Z. Jiang, C. Z. Chang, M. R. Masir, C. Tang, Y. Xu, J. S. Moodera, A. H. MacDonald,
and J. Shi, Nat. Commun. 7, 11458 (2016).
[18]
J. Han, A. Richardella, S. Siddiqui, J. Finley, N. Samarth, and L. Liu, arXiv:1703.07470
(2017).
[19] M. Li, C. Z. Chang, B. J. Kirby, M. E. Jamer, W. Cui, L. Wu, P. Wei, Y. Zhu, D. Heiman,
J. Li, and J. S. Moodera, Phys. Rev. Lett. 115, 087201 (2015).
[20] C. Lee, F. Katmis, P. Jarillo-Herrero, J. S. Moodera, and N. Gedik, Nat. Commun. 7,
12014 (2016).
[21] Z. Jiang, C. Z. Chang, C. Tang, P. Wei, J. S. Moodera, and J. Shi, Nano Lett. 15, 5835
(2015).
[22] Y. Sun, H. Chang, M. Kabatek, Y.-Y. Song, Z. Wang, M. Jantz, W. Schneider, M. Wu,
E. Montoya, B. Kardasz, B. Heinrich, S. G. E. te Velthuis, H. Schultheiss, and A.
Hoffmann, Phys. Rev. Lett. 111, 106601 (2013).
[23] H. L. Wang, C. H. Du, Y. Pu, R. Adur, P. C. Hammel, and F. Y. Yang, Phys. Rev. Lett.
112, 197201 (2014).
[24] Y. Tserkovnyak, A. Brataas, and G. Bauer, Phys. Rev. Lett. 88, 117601 (2002).
[25] Y. Xia, D. Qian, D. Hsieh, L. Wray, A. Pal, H. Lin, A. Bansil, D. Grauer, Y. S. Hor, R.
J. Cava, and M. Z. Hasan, Nat. Phys. 5, 398 (2009).
[26] Y. Kajiwara, K. Harii, S. Takahashi, J. Ohe, K. Uchida, M. Mizuguchi, H. Umezawa,
H. Kawai, K. Ando, K. Takanashi, S. Maekawa, and E. Saitoh, Nature 464, 262 (2010).
[27] K. H. M. Chen, H. Y. Lin, S. R. Yang, C. K. Cheng, X. Q. Zhang, C. M. Cheng, S. F.
Lee, C. H. Hsu, Y. H. Lee, M. Hong, and J. Kwo, to be published in Appl. Phys. Lett.
(2017).
[28] Supplemental Materials.
[29] Y. Zhang, K. He, C.-Z. Chang, C.-L. Song, L.-L. Wang, X. Chen, J.-F. Jia, Z. Fang, X.
Dai, W.-Y. Shan, S.-Q. Shen, Q. Niu, X.-L. Qi, S.-C. Zhang, X.-C. Ma, and Q.-K. Xue,
Nat. Phys. 6, 584 (2010).
[30] G. Landolt, S. Schreyeck, S. V. Eremeev, B. Slomski, S. Muff, J. Osterwalder, E. V.
Chulkov, C. Gould, G. Karczewski, K. Brunner, H. Buhmann, L. W. Molenkamp, and
J. H. Dil, Phys. Rev. Lett. 112, 057601 (2014).
[31] M. Neupane, A. Richardella, J. Sanchez-Barriga, S. Xu, N. Alidoust, I. Belopolski, C.
Liu, G. Bian, D. Zhang, D. Marchenko, A. Varykhalov, O. Rader, M. Leandersson, T.
Balasubramanian, T. R. Chang, H. T. Jeng, S. Basak, H. Lin, A. Bansil, N. Samarth,
and M. Z. Hasan, Nat. Commun. 5, 3841 (2014).
[32] Y. Tserkovnyak, A. Brataas, G. E. W. Bauer, B. I. Halperin, and Rev. Mod. Phys. 77,
1375 (2005).
[33] K. T. Yamamoto, Y. Shiomi, K. Segawa, Y. Ando, and E. Saitoh, Phys. Rev. B 94,
204404 (2016).
[34] P. E. Seiden, Phys. Rev. 133, A728 (1964).
[35] C. L. Jermain, S. V. Aradhya, N. D. Reynolds, R. A. Buhrman, J. T. Brangham, M. R.
Page, P. C. Hammel, F. Y. Yang, and D. C. Ralph, Phys. Rev. B 95, 174411 (2017).
[36] Y. Ohnuma, H. Adachi, E. Saitoh, and S. Maekawa, Phys. Rev. B 89, 174417 (2014).
[37] L. Frangou, S. Oyarzún, S. Auffret, L. Vila, S. Gambarelli, and V. Baltz, Phys. Rev.
Lett. 116, 077203 (2016).
[38] Z. Qiu, J. Li, D. Hou, E. Arenholz, A. T. N'Diaye, A. Tan, K. Uchida, K. Sato, S.
Okamoto, Y. Tserkovnyak, Z. Q. Qiu, and E. Saitoh, Nat. Commun. 7, 12670 (2016).
[39] Y. G. Semenov, X. Duan, and K. W. Kim, Phys. Rev. B 86, 161406(R) (2012).
[40] P. K. Manna and S. M. Yusuf, Phys. Rep. 535, 61 (2014).
[41] F. B. Abdulahad, J. H. Lin, Y. Liou, W. K. Chiu, L. J. Chang, M. Y. Kao, J. Z. Liang, D.
S. Hung, and S. F. Lee, Phys. Rev. B 92, 241304(R) (2015).
[42] S. Suhl, Phys. Chem. Solids. 1, 209 (1957).
[43] R. Iguchi, K. Ando, R. Takahashi, T. An, E. Saitoh, and T. Sato, Jpn. J. Appl. Phys. 51,
103004 (2012).
Figure 1.
FIG. 1. (a), (b) FMR using the cavity and co-planar configuration for angle- and frequency-
dependent study, respectively. A dc external field 𝐻𝑒𝑥𝑡 was applied and ℎ𝑟𝑓 denotes the
microwave field. (c) FMR spectra of Bi2Se3(25)/YIG(12) and YIG(12) measured by the cavity.
(d), (e) 𝜃𝐻 and f dependence of 𝐻𝑟𝑒𝑠 of Bi2Se3(25)/YIG(12) and YIG(12), respectively.
Figure 2.
FIG. 2. (a) The 𝑑𝑌𝐼𝐺 dependence of 4𝜋𝑀𝑒𝑓𝑓 of Bi2Se3/YIG (solid triangles) and YIG
(hollow squares) obtained from 𝜃𝐻 (red) and f (blue) dependent FMR. (b) The 𝐾𝑒𝑓𝑓𝑑𝑌𝐼𝐺 vs
𝑑𝑌𝐼𝐺 plot for determining 𝐾𝑖 using a linear fit. The intercept of the y-axis corresponds to the
𝐾𝑖 value. (c) 𝑑𝐵𝑆 dependence of 𝐾𝑖. The figure is divided into two regions. For 𝑑𝐵𝑆 > 6 nm,
the Dirac cone of TSS is intact, with the Fermi level located in the bulk conduction band. For
𝑑𝐵𝑆 < 6 nm, a gap and quantum well states form. (d) The 𝑑𝐵𝑆 dependence of spin mixing
conductance 𝑔↑↓. The inset shows ∆𝐻 as a function of Bi2Se3(7)/YIG(13) and YIG(13) for
calculating 𝑔↑↓. The red dashed line indicates the typical value of 𝑔↑↓ of Pt/YIG.
Figure 3.
FIG. 3. (a), (b) 𝑓 vs 𝐻𝑟𝑒𝑠 data for various T for YIG(23) and Bi2Se3(25)/YIG(15),
respectively. Solids lines are fitted curves using Eq. (2). (c), (d), (e), and (f) T dependence of
𝐻𝑟𝑒𝑠, ∆𝐻, 4𝜋𝑀𝑒𝑓𝑓 and 𝐻𝑒𝑓𝑓 of the one YIG (23) single layer and two Bi2Se3/YIG bilayer
samples, Bi2Se3(25)/YIG(15) and Bi2Se3(16)/YIG(17), respectively. The arrows in (c) and (d)
denote the position of the "bumps". Solids line are guide of eyes obtained by properly
smoothing the experimental data.
Figure 4.
FIG. 4. (a), (b) FMR first derivative and microwave absorption spectra for various T,
respectively. The arrows indicate the 𝐻𝑒𝑥𝑡 sweep direction. The dashed line in (b) traces the
T evolution of the absorption peak 𝐻𝑝𝑒𝑎𝑘 . (c), (d), and (e) Schematics of the Bi2Se3/YIG
sample when T = 100, 25, and 5 K, respectively. (f) Zero-field FMR temperature 𝑇0 as a
function of f.
Supplemental materials
Growth and FMR characteristics of YIG films
FIG. S1. (a) and (b) AFM surface image and HAADF-STEM image of YIG/GGG.
The YIG thin films were deposited on (111)-oriented GGG substrates by off-axis
sputtering at room temperature. The GGG(111) substrates were first ultrasonically
cleaned in order of acetone, ethanol and DI-water before mounted in a sputtering
chamber with the base pressure of 2 × 10−7 mTorr. For YIG deposition, a 2-inch YIG
target was sputtered with the following conditions: an applied rf power of 75 W, an Ar
pressure of 50 mtorr and a growth rate of 0.6 nm/min. The samples were then annealed
at 800oC with O2 pressure of 11.5 mtorr for 3 hours. Fig. S1(a) displays the atomic force
microscopy (AFM) image of the YIG surface, showing a flat surface with roughness of
0.19 nm. Fig. S1(b) shows the high-angle annular dark-field (HAADF) image of
YIG/GGG. The YIG thin film was epitaxially grown on the GGG substrate with
excellent crystallinity. No crystal defects were observed at the YIG bulk and YIG/GGG
interface.
FIG. S2. (a) FMR first-derivative spectra of 23 nm YIG at various frequencies. (b) 𝑓 vs 𝐻𝑟𝑒𝑠 data fitted
to the Kittel equation (red line). (c) ∆𝐻 as a function of 𝑓 data. From the linear fit (red line) the 𝛼 value
of the sample is obtained.
Fig. S2(a) shows the representative FMR data of our YIG film measured by
coplanar waveguide. The FMR spectra exhibit Lorentzian lineshape at all measured
frequencies ranging from 3 to 9.76 GHz. To determine the 4𝜋𝑀𝑒𝑓𝑓 and 𝛼 , the
resonance fields 𝐻𝑟𝑒𝑠 and peak-to-peak widths ∆𝐻 of these spectra were plotted as a
function of 𝑓 as shown in Fig. S2(b) and (c), respectively. We obtain the 4𝜋𝑀𝑒𝑓𝑓 of
our 23 nm YIG to be 1630 Oe. We note that the 4𝜋𝑀𝑒𝑓𝑓 value is lower than the
reported values of the YIG prepared by either sputtering or pulsed laser deposition [S1-
3]. The difference may come from growth conditions dependent on different systems.
For determination of 𝛼, the data in Fig. S2(c) is fitted to the following equation,
∆𝐻 = ∆𝐻0 +
4𝜋𝑓𝛼
√3𝛾
. (S1)
Here, ∆𝐻0 and 𝛾 are the inhomogeneous broadening and gyromagnetic ratio. The
linear fit in Fig. S2(c) corresponds to 𝛼 = 1.1 × 10−3.
Structural characterizations of Bi2Se3/YIG heterostructures
FIG. S3. (a) RHEED patterns of MBE grown 7 QL Bi2Se3 on YIG/GGG(111) substrates. (b) AFM image
of a 7 QL Bi2Se3. (c) HAADF-STEM image of Bi2Se3/YIG/GGG heterostructures. (d) SR-XRD of our
Bi2Se3(25)/YIG(12) sample. Clear Pendellösung fringes of YIG and Bi2Se3 indicates excellent crystallinity.
The inset shows the radial scans of YIG before and after Bi2Se3 growth.
The YIG/GGG samples were annealed at 450oC in the MBE growth chamber for
30 min prior to Bi2Se3 growth at 280oC [27]. The base pressure of the system was kept
about 2 × 10−10 Torr. Elemental Bi (7N) and Se (7N) were evaporated from regular
effusion cells. As shown in Fig. S3a, streaky reflection high-energy electron diffraction
(RHEED) patterns of Bi2Se3 were observed. Fig. S3b displays the surface morphology
of 7 QL Bi2Se3 taken by atomic force microscopy (AFM). The image shows layer-by-
layer growth of Bi2Se3 with the step heights ~ 1nm, which corresponds to the thickness
of 1 QL. The layer structure of Bi2Se3 was also revealed by the HAADF image shown
in Fig. S3c. Despite the high quality growth of Bi2Se3, an amorphous interfacial layer
of ~ 1 nm formed. The excellent crystallinity of our samples was verified by clear
Pendellösung fringes of the synchrotron radiation x-ray diffraction (SR-XRD) data
shown in Fig. S3(d). In particular, the radial scans data of YIG/GGG(22-4) before and
after growing Bi2Se3 are perfectly matched, indicating the absence of Bi2Se3-induced
strains in YIG that might contribute additional magnetic anisotropy [S4].
Analyses of magnetic anisotropy
We express the free energy density E of the system as
𝐸 = −𝑴 ∙ 𝑯 +
1
2
𝑀𝑠(4𝜋𝑀𝑠 − 𝐻𝑎𝑛 − 𝐻𝑖𝑛𝑡)cos2𝜃𝑀 (S2)
, where 𝑴 , 𝑯 , 𝑀𝑠 , 𝐻𝑎𝑛 and 𝜃𝑀 are magnetization vector, applied field vector,
saturation magnetization, the anisotropy field induced by Bi2Se3 and magnetization
angle with respect to the surface normal, respectively. We further define the effective
demagnetization
field 4𝜋𝑀𝑒𝑓𝑓 = 4𝜋𝑀𝑠 − 𝐻𝑎𝑛 − 𝐻𝑖𝑛𝑡
, where 𝐻𝑎𝑛 is
the
magnetocrystalline anisotropy field of sputtered YIG and 𝐻𝑖𝑛𝑡 is the interfacial
anisotropy field. We have 𝐻𝑖𝑛𝑡 = 0 for the YIG single layer by definition. The first
term of Eq. (S2) is the Zeeman energy and the second term accounts for uniaxial out-
of-plane anisotropy. Here, we neglect higher order terms that are relatively small for a
strain-free cubic system. The 𝐻𝑟𝑒𝑠 can be calculated by minimizing 𝐸 and, at the
equilibrium angle of 𝑴, solving the Smit-Beljers equation [S5],
(
2
)
𝜔
𝛾
=
1
𝑀2 sin2 𝜃𝑀
[
𝜕2𝐸
2
𝜕𝜃𝑀
𝜕2𝐸
𝜕𝜑𝑀
2 − (
2
𝜕2𝐸
2 )
2 𝜕𝜑𝑀
𝜕𝜃𝑀
]. (S3)
As 𝜃𝐻 = 𝜋/2 ,
the FMR conditions reduce
to
the Kittel equation 𝑓 =
𝛾
2𝜋
√𝐻𝑟𝑒𝑠(𝐻𝑟𝑒𝑠 + 4𝜋𝑀𝑒𝑓𝑓).
We can safely assume that the 𝐻𝑎𝑛 did not change before and after the growth of
Bi2Se3 based on Figure S3(d). With this in mind, we further notice that 𝐾𝑖 can be
alternatively
expressed
as
(1/2)(4𝜋𝑀𝑒𝑓𝑓
𝑌𝐼𝐺 − 4𝜋𝑀𝑒𝑓𝑓
𝐵𝑆/𝑌𝐼𝐺)𝑀𝑠𝑑𝑌𝐼𝐺
, where
4𝜋𝑀𝑒𝑓𝑓
𝑌𝐼𝐺 (4𝜋𝑀𝑒𝑓𝑓
𝐵𝑆/𝑌𝐼𝐺) represents the 4π𝑀𝑒𝑓𝑓 of YIG (Bi2Se3/YIG) for a specific
𝑑𝑌𝐼𝐺. The calculated 𝐾𝑖 using this expression for 𝑑𝐵𝑆 = 25 nm gives an average of -
0.068 erg/cm-2, in good agreement with a 𝐾𝑖 of -0.075 erg/cm2 obtained from the linear
fit in Figure 2(b).
Temperature-dependent FMR spectra of YIG and Bi2Se3/YIG
FIG. S4. Temperature-dependent FMR first-derivative spectra of (a) YIG(23) and (b)
Bi2Se3(25)/YIG(15). The Δ𝐻 increased with decreasing 𝑇 , accompanied by decreased
𝑑𝐼/𝑑𝐻𝑒𝑥𝑡 peak magnitudes due to enhanced damping. The 𝐻𝑒𝑥𝑡 scale are fixed to clearly
show the pronounced changes of 𝐻𝑟𝑒𝑠 and Δ𝐻 induced by Bi2Se3.
Extraction of 𝟒𝝅𝑴𝒆𝒇𝒇 and 𝑯𝒆𝒇𝒇 from temperature dependence of 𝑯𝒓𝒆𝒔
FIG. S5. (a) 𝐻𝑟𝑒𝑠 vs 𝑇 data of Bi2Se3(25)/YIG(15) measured at 4 and 4.5 GHz. (b) and (c)
Comparison of extracted 4𝜋𝑀𝑒𝑓𝑓 and 𝐻𝑒𝑓𝑓 by the fitting and solving method.
Since the effective damping constant of YIG and Bi2Se3/YIG increased
pronouncedly at low 𝑇, the weakened FMR signal was inevitably accompanied by
larger uncertainties of measured 𝐻𝑟𝑒𝑠. The uncertainties are even more serious when
𝑇 < 150 K and 𝑓 > 5 GHz for our instruments. Fitting the data including points in
the 𝑓 > 5 GHz region to the Kittel equation gives large errors of 4𝜋𝑀𝑒𝑓𝑓 and 𝐻𝑒𝑓𝑓,
which obscure the temperature dependency of these two quantities. To reduce the
uncertainties in the process of extracting 4𝜋𝑀𝑒𝑓𝑓 and 𝐻𝑒𝑓𝑓, we focused on the FMR
data for 𝑓 = 4 and 4.5 GHz , from which we obtained 𝐻𝑟𝑒𝑠 with satisfactory
accuracy (Fig. S5(a)). The Kittel equation can be arranged in the following form,
𝐻𝑒𝑓𝑓
2 + (2𝐻𝑟𝑒𝑠 + 4𝜋𝑀𝑒𝑓𝑓)𝐻𝑒𝑓𝑓 + (4𝜋𝑀𝑒𝑓𝑓𝐻𝑟𝑒𝑠 + 𝐻𝑟𝑒𝑠
𝛾2 ) = 0. (S3)
With 𝛾 = 1.77 × 1011 t−1s−1 , the two sets of data in Fig. S5a provided us with
4𝜋2𝑓2
2 −
sufficient information to explicitly solve the second-order equation. Fig. S5b and S5c
compare the results of "fitting" and "solving" the Kittel equation. For 𝑇 > 150 K,
where FMR can be accurately measured up to 7 GHz, the 4𝜋𝑀𝑒𝑓𝑓 and 𝐻𝑒𝑓𝑓 obtained
from solving and fitting method agree well, demonstrating the reliability of the solving
method.
Temperature dependence of magnetization hysteresis loop
FIG. S6. Temperature dependence of magnetization hysteresis loop of Bi2Se3(16)/YIG(17)
measured by a SQUID magnetometer. The paramagnetic background of the GGG substrate has
been subtracted. No shifts of hysteresis loops were observed.
References:
[S1] O. d'Allivy Kelly, A. Anane, R. Bernard, J. Ben Youssef, C. Hahn, A. H.
Molpeceres, C. Carretero, E. Jacquet, C. Deranlot, P. Bortolotti, R. Lebourgeois,
J. C. Mage, G. de Loubens, O. Klein, V. Cros, and A. Fert, Appl. Phys. Lett. 103,
082408 (2013).
[S2]
J. C. Gallagher, A. S. Yang, J. T. Brangham, B. D. Esser, S. P. White, M. R. Page,
K.-Y. Meng, S. Yu, R. Adur, W. Ruane, S. R. Dunsiger, D. W. McComb, F. Yang,
and P. C. Hammel, Appl. Phys. Lett. 109, 072401 (2016).
[S3] H. Chang, P. Li, W. Zhang, T. Liu, A. Hoffmann, L. Deng, and M. Wu, IEEE
Magn. Lett. 5 (2014).
[S4] H. L. Wang, C. H. Du, P. C. Hammel, and F. Yang, Phys. Rev. B 89 (2014).
[S5]
J. Smit and Beljers, Philips Res. Repts. 10, 113 (1955).
|
1709.04315 | 1 | 1709 | 2017-09-13T13:23:23 | Field-free perpendicular magnetization switching through domain wall motion in Pt/Co/Cr racetracks by spin orbit torques with the assistance of accompanying Joule heating effect | [
"physics.app-ph"
] | Heavy metal/ferromagnetic layers with perpendicular magnetic anisotropy (PMA) have potential applications for high-density information storage in racetrack memories and nonvolatile magnetic random access memories. Writing and erasing of information in these devices are carried out by domain wall (DW) motion and deterministic magnetization switching via electric current generated spin orbital torques (SOTs) with an assistance of in-plane bias field to break the symmetry. Improvements in energy efficiency could be obtained when the switching of perpendicular magnetization is controlled by an electric current generated SOTs without the in-plane bias fields. Here, we report on reversible electric-current-driven magnetization switching through DW motion in Pt/Co/Cr trilayers with PMA at room temperature due to the formation of homochiral Neel-type domain, in which an in-plane effective Dzyaloshinskii-Moriya interaction field exists. Fully deterministic magnetic magnetization switching in this trilayers is based on the enhancement of SOTs from a dedicated design of Pt/Co/Cr structures with two heavy metals Pt and Cr which show the opposite sign of spin Hall angles. We also demonstrated that the simultaneously accompanying Joule heating effect also plays a key role for field-free magnetization switching through the decrease of the propagation field. | physics.app-ph | physics | Field-free perpendicular magnetization switching through domain wall
motion in Pt/Co/Cr racetracks by spin orbit torques with the assistance of
accompanying Joule heating effect
Baoshan Cui, Dong Li, Jijun Yun, Yalu Zuo, Xiaobin Guo, Kai Wu, Xu Zhang, Yupei Wang, Li Xi*,
and Desheng Xue
Key Laboratory for Magnetism and Magnetic Materials of Ministry of Education & School of Physical
Science and Technology, Lanzhou University, P. R. China
Abstract
Heavy metal/ferromagnetic layers with perpendicular magnetic anisotropy (PMA) have potential
applications for high-density information storage in racetrack memories and nonvolatile magnetic
random access memories. Writing and erasing of information in these devices are carried out by domain
wall (DW) motion and deterministic magnetization switching via electric current generated spin orbital
torques (SOTs) with an assistance of in-plane bias field to break the symmetry. Improvements in energy
efficiency could be obtained when the switching of perpendicular magnetization is controlled by an
electric current generated SOTs without the in-plane bias fields. Here, we report on reversible
electric-current-driven magnetization switching through DW motion in Pt/Co/Cr trilayers with PMA at
room temperature due to the formation of homochiral Néel-type domain, in which an in-plane effective
Dzyaloshinskii-Moriya interaction field exists. Fully deterministic magnetic magnetization switching in
this trilayers is based on the enhancement of SOTs from a dedicated design of Pt/Co/Cr structures with
two heavy metals Pt and Cr which show the opposite sign of spin Hall angles. We also demonstrated
that the simultaneously accompanying Joule heating effect also plays a key role for field-free
magnetization switching through the decrease of the propagation field.
* Corresponding author. E-mail address: [email protected] (Li Xi)
1
I. Introduction
Recently, spin-orbit torques (SOTs) in sandwich structures with perpendicular magnetic anisotropy
(PMA) where ultrathin ferromagnets (FM) is separated by a heavy metal (HM) and an oxide, have
attracted abundant research interests for highly efficient magnetization switching1-3 and fast domain
wall motion.4-10 In this kind of device, when an in-plane charge current (Je) flows through HM with
strong spin-orbit coupling (SOC) including Pt,1, 11-13 β-Ta,13-15 Hf,16 and β-W,17 etc., it can be converted
into a pure spin current (Js). Then, Js injects into FM and generates a torque to act on magnetic moments
under the assistance of an in-plane magnetic field. As a result, if the torque is sufficiently strong, the
magnetization could be switched. It is well established that the SOT switching efficiency is directly
related to the magnitude of spin Hall angle (θSH). So, considerable efforts have been devoted to obtain a
large θSH of HMs by varying the thickness of HM,18, 19 decorating the interface between HM and FM,20,
21 changing the crystallinity of HM,22 and even involving oxygen in HM.23 Besides, some reports also
achieve large effective θSH based on HM/FM/HM structures, in which two HM layers have opposite
sign of θSH.24-26 However, a deterministic magnetization switching by SOT always requires an in-plane
bias magnetic field along the current direction to break the symmetry,1 which is indispensable to achieve
the magnetization switching and has an obvious obstacle for the application in SOT-based devices.
Aiming to realize field-free SOT devices, several designs through introducing an effective in-plane field
have been experimentally demonstrated. The in-plane effective field could be induced via a wedge
structure,27, 28 or exchange coupling with another ferromagnetic layer with in-plane anisotropy,29
exchange bias with an in-plane antiferromagnetic layer,30, 31 and even in a hybrid ferromagnetic/ferroel-
ectric structure.32
In our previous work,33 we have investigated the deterministic magnetization switching in
structural inversion asymmetric Pt/Co/Cr trilayer. In which the enhanced pure spin currents generated
2
from both of Pt and Cr with opposite sign of θSH is used to work in concert to improve the SOT
switching efficiency under an assistance of the in-plane magnetic field. The enhanced PMA and SOTs in
Pt/Co/Cr structures were closely not only related to the Co/Cr, but to Pt/Co interfaces, where
Dzyaloshinskii-Moriya interaction (DMI)34-37 originates and shows influence on domain wall velocity
as reported in recent works.38-40 In our devices, there are two interfaces to contribute DMI, which could
stabilize a homochiral Néel-type domain wall and result in an in-plane DMI effective field in DW. In
this work, we report that the magnetization switching through current-driven domain wall motion also
could be achieved in a micro-sized racetrack using the in-plane DMI field in a left-handed chirality Néel
domain wall to replace an in-plane bias field. Moreover, we find that the accompanying Joule heating
effect has a significant effect on current-driven domain wall motion and it can decrease the thermally
activated barrier. Consequently, a small propagation field (HP) is required to push DWM. While most
works tend to ignore it1, 14, 30 excepting few reports considering this effect.41
II. Experimental
The Ta(3)/Pt(5)/Co(0.8)/Cr(1)/Al(1) (thickness number in nanometer) stacks with as-grown
perpendicular magnetic anisotropy were deposited on Corning glass substrate by direct current
magnetron sputtering with base pressure less than 5 × 10-5 Pa as described elsewhere.33 The deposited
films were patterned into around 300 μm long and 8 μm wide racetracks with two circular nucleation
pads and electrodes by photolithography and Ar ion milling as shown in Fig. 1(a). The typical resistance
of the whole structure is around 6.3 k. The formation of reversed magnetic domain and the creep of
domain wall were observed by the variation of magnetic fields and/or the currents passing through the
racetrack using a polar magneto-optical Kerr microscopy working in differential mode.42 A voltage
pulse generator (Tektronix: PSPL10300B, + 50/ -45 V with rising time around 300 ps) in a short pulse
regime (15 – 100 ns) and wide period ranging from 10 s to 1 s was used to generate pulse current to
3
depin the DW and control the Joule heating effect. Current strength was calculated from the voltage
measured in a serially connected real-time oscilloscope. A DC current source (Keithley: current source
220) was also used for comparison. All of the above experiments were carried out at room temperature.
III. Results and discussion
Fig. 1. Optical difference image of Pt/Co/Cr racetrack with two circular nucleation pads and electrodes saturated at -30.1
mT out-of-plane field (a) and domain wall motion velocity as a function of out-of-plane field Hz (b). The green block in
(a) shows the interested area, in which the hysteresis loops were recorded to obtain the propagation field at different
magnetic fields and/or currents. The dash line in (b) is guided to the eye. The inset of (b) shows the scaling plot of log v
vs. (0HZ)-1/4 with linear fitting.
The optical image of Pt/Co/Cr racetrack with two circular nucleation pads and electrodes is shown
in Fig.1(a). In the experimental, we found that the thermal activated nucleation events will always
happen in the left-side nucleation pads when the track was firstly saturated in one direction and then a
small reverse magnetic field was applied. Thus, the domain wall (DW) propagates along the track from
left side to right side driven by out-of-plane magnetic fields, i.e. the propagation field as shown in the
supplement movies (SI-m1). The grey level of the interested green block area was recorded as a
function of magnetic fields by the polar Kerr microscope working in the differential mode. Thus, the
4
hysteresis loop can be obtained, from which the HP was obtained. The domain wall motion (DWM)
velocity (v) was obtained by counting the time for DW travelling along the racetrack with a fixed
length.
Figure 1(b) plots v as a function of out-of-plane field Hz. The red arrow in the figure indicates the
DW depinning field (Hdep), above which the DW exhibits a linear dependence with v Hz-Hdep, which
is the typical characterization of DW motion in the flow regime (Hz > Hdep).43 Meanwhile, the creep
regime (Hz < Hdep) shows thermally activated DWM with the creep criticality ln(v) Hz
-1/4,43 as shown
by the blue dash line of the best linear fit in the inset of Fig.2. In the following measurement, we will
focus on the DW creep regime to investigate the current dependent DW velocity. In order to avoid too
much Joule heating affecting the thermally activated barrier landscape, we use a pulse generator with
pulse width lower than 100 ns and pulse period ranging from 10 s to 1 s. We found that the variation of
v not only relies on the amplitude of current, but also is related to the polarity of the current.
When the pulse current is not more than 7.9 mA with the period of 10 s, the nucleation events is
always happened in their original position (the left nucleation pad) in this device. Moreover, the
negative (positive) pulse currents favor the domain wall propagation along the track from the left (right)
side to the right (left) side whenever the DW is up-down or down-up type since the nucleation always
happens in the left side, and then DW always creeps from left to right along the track at a HP. This kind
of DWM characterization is shown in the supplementary movies (SI-m2). Thus, a negative current
avails the DWM, while a positive current hinders the DWM along the track at a fixed HP. This current
direction determined DWM characterization can be ascribed to the enhanced SOTs from the Pt and Cr
layers with opposite signs of spin Hall angles and the formation of chiral DW.4, 15, 44 It should be
mentioned that the current induced DWM by spin transfer torque45 has a contrary effect from our
experimental observation and has been ignored as reported in other SOT devices.15, 44
5
Fig. 2. The sketch of a left(right)-handed chirality Néel-type domain wall in Pt/Co/Cr sample with the illustration of
negative current availed (hindered) or positive current hindered (availed) domain wall motion whenever the domain wall
is up-down or down-up type. The anti-damping like spin Hall effective field (HSHE) in all cases is also shown.
Figure 2 shows the sketch of domain wall motion under a pulse current and a propagation field Hz
with a left-handed and right-handed chirality DW. If there is only reversed HP applied, the DW will
always creep from left side to right side as observed in our experimental. The pulse current generates a
spin Hall effective field (HSHE), which acts on the domain wall magnetization vector4, 15, 44 and is given
by,
(1)
where, , Ms, tF, Jx, and represent the effective spin Hall angle, the saturation magnetization of
the FM, the thickness of FM, the current density along x direction, the unit vector of magnetization of
FM and the unit vector of current density, respectively. Current generated effective HSHE on a left (right)
handed-chirality Néel DW is along (opposite) the external out-of-plane field Hz when a negative current
is applied, while it is opposite (along) to the direction of Hz when a positive current is applied as shown
in Fig.2. Thus, for a left-handed chirality DW a negative current induced HSHE will accelerate DWM,
and a positive current induced HSHE will accelerate DWM for a right-handed chirality DW. In this work,
the left-handed chirality of DW coincides with our experimental observations with the negative current
availing the DWM and the positive current hindering DWM. Thus, a left-handed chirality DW forms in
Pt/Co/Cr trilayers. It indicates a DMI in Pt/Co/Cr. The magnitude of DMI will be reported elsewhere.
6
The change of HP of the interested area with the variation of polarity and magnitude of the current
occurs in Pt/Co/Cr with the specified left-handed chirality of DW. Figure 3(a) shows the normalized HP
with variation of pulse current with the repeating frequency of 100 kHz, i.e. 10 s periods. HP was
obtained by measuring the hysteresis loop of the interested block area. One can see that the HP
decreases with the decrease of the negative pulse current due to the negative pulse generated field has
the same direction with the applied field during DW propagation, while at the positive current, HP firstly
increases up to a peak around 5.8 mA and then decreases quickly. The increase of HP with the increase
of positive pulse current can be ascribed to the increase of HSHE, which is opposite to the applied field.
The decrease of the HP at the high pulse current (e.g. 7.9 mA) may be ascribed to the Joule heating
effect of current, which changes the thermal activated barrier landscape and makes the nucleation area
changing to the left end of the track as shown in the supplementary videos (SI-m3). Moreover, the peak
height increases with the decrease of pulse width at the same pulse period of 10 s. It can be understood
by that the shorter of pulse width is, the weaker Joule heating effect generates and the associated lower
temperature rises, the slower DWM velocity becomes with the same strong spin Hall effective field, and
the larger HP will be required to push the DW passing the interested block area, which leads to the
increase of peak amplitude. When the pulse current is too large, the strong Joule heating effect will
change the thermally activated barrier landscape and eventually changes the nucleation area and
decreases the nucleation fields. It results in the quick decrease of HP at the large pulse current sides.
These findings are also examined by keeping the pulse width in 100 ns and changing the periods to 100
s (i.e. 10 kHz) as shown in Fig. 3(b). It can be seen that when decreasing the duty cycle, the decrease
of HP on the large pulse current sides are all restrained due to the decreased Joule heating effect.
Furthermore, we also examined our results by keeping the pulse number and pulse width are the same
(i.e. the same SHE effect) in each magnetic field steps as shown in Fig. 3(c). In this condition, the
7
function of SOTs is the same, but the Joule heating effect is larger for the shorter period one due to its
higher duty cycle. It results in a large decrease of the HP due to the heating effect when the pulse current
reaches 7.9 mA.
Fig. 3. Pulse current induced variation of normalized propagation fields with the same period and the different width of
current pulses (a), and pulse current induced variation of normalized propagation fields with the same width and different
periods of the pulse current and magnetic field steps (b, c). HP (0) represents the propagation field without any pulse
currents.
The pulse current induced variation of the DWM velocity is also investigated with the fixed magnetic
field and the variation of pulse current periods and magnitude as shown in Fig.6. In the creep region
where 0Hz < ~10 mT (see Fig. 1(b)), the DW velocity can be expressed as43, 46, 47
(2)
where, UC is a characteristic energy related to the disorder-induced pinning potential, kB is the Boltzmann
constant, T is the temperature and Hdep is the depinning field at which the Zeeman energy is equal to the DW
pinning energy. At the same static out-of-plane magnetic field, v is larger for the high frequency one due to
8
the more heating generation, which increases T in Eq. (2). Moreover, v has quite large variation at the
negative current side than that at the positive current side due to both of Joule heating effect and SOT
effective field having the same contributions to v at the negative current rather than their contrary
contribution at the positive current as shown in Fig. 4(a). Eventually, the DW velocity can be changed by
almost two orders of magnitude through negative pulse currents.
Fig. 4. The variation of DWM velocity with the different pulse currents under different magnetic fields and pulse
periods(a), and the variation of DWM velocity with the square of pulse currents under different magnetic fields and pulse
periods (b).
In our previous work, the effective antidamping-like SOT fields per unit current density is around 0.9
mT/(106 A/cm2) for Pt/Co/Cr sample.33 Thus, the variation of pulse current corresponds to the variation of the
effective HSHE acting on the DW. Thus we can quantitatively get the v variation of DWM dependence on
pulse current with a fixed reversal HP if the Joule heating effect does not exist. However, the above
experimental already prove that the larger the pulse current becomes, the stronger the Joule heating effect
9
will be. Thus, the current not only contributes to the variation of T, but also the spin Hall effective HSHE in Eq.
(1). It is hard to distinguish those effects separately at this time. However, the Joule heating caused
temperature rise is always proportional to I2.43 This temperature dependent effect was demonstrated in Fig.
4(b) which shows lnv roughly proportional to the square of current. Moreover, the Joule heating effect
becomes more evident if we change the pulse current to a DC current source (Keithley: 220). Figure 5 shows
HP of the interested area in Fig. (1) at different DC currents. One can see the propagation field decreases with
the increase of currents and is also roughly proportional to the square of the DC currents. Thus, the decrease
of HP may be mainly ascribed to the Joule heating effect rather than the SOT effective fields. It should be
mentioned that the nucleation position could be permanently changed from the left nucleation pad to the
middle of the track due to thermal annealing effect once a quite high DC currents was applied (e.g. 3.2 mA in
this case).
From the above observation, we can conclude that once a large current (no matter it is DC or pulse
current) generated SHE effective field reaches the HP, which was simultaneously reduced due to Joule
heating effect of the current, the DW will pass the interested area to achieve the magnetization switching
without the assistance of an in-plane field in our Pt/Co/Cr devices with left-handed chirality DW. Figure 6
shows snapshot of the DWM image at large current pulse around 7.1 mA without any applied magnetic
fields. In order to get strong HSHE, we first use a reversed out-of-plane magnetic field to push the DW into the
track, then the field was set to zero and a current pulse was applied. One can see DW was moved by the
current pulse from left (right) side to right (left) side at a negative (positive) current pulse. The detailed
DWM is shown in the supplementary movies (SI-m4), from which we can see that the reversible
electric-current-driven magnetization switching through DW motion could be achieved via changing the
polarity of the pulse current. Using the effective antidamping-like SOT fields per unit current density around
0.9 mT/(106 A/cm2) for Pt/Co/Cr sample with the assumption of uniformed flowing of current, the calculated
10
effective antidamping-like SOT fields is around 8.2 mT, which is in the DW creep regime and can push the
DWM. If only this HSHE existing, v is around 10.5 m/s according to Fig. 1(b), however, a quite large v (>>
10 m/s) is observed and ascribed to the Joule heating induced temperature rise, which on one hand could
decrease the propagation field, and on other hand could increase v. It coincides with our other findings that
the larger the pulse current with higher frequency, the larger the DWM velocity will be due to the current
induced spin Hall effect and the accompanying Joule heating effect.
Fig. 5. DC current dependence of propagation field of the interested area. The inset shows the square of current
dependence of propagation field. The dashed line is the best linear fitting curve.
Fig. 6. Current pulse induced domain wall motion under negative current of -7.1 mA (a – c) and positive current of 7.1
mA (d – f). (a) and (d), the snapshots of the nucleation of a reversed domain. (b) and (e), the snapshots of DW, which was
11
first pushed into the track by a reversed magnetic field, then the magnetic field was removed. (c) and (f), the snapshots of
DW, which was pushed by turning on the corresponding current pulse.
In conclusion, the magnetization switching through domain wall motion in Pt/Co/Cr racetracks is
achieved. The velocity of domain wall can be increased by more than two orders of magnitude through
applying a negative current pulse by current generated extra spin Hall effective field and Joule heating
effect in a fixed out-of-plane field comparing with that at a positive current pulse. Due to the formation
of left-handed chirality domain wall, in which an in-plane effective DMI field exists, domain wall
motion by current generated SHE effective field could be realized without applying any in-plane bias
fields. The magnetization switching can be achieved by only applying a large current since the current
generated spin Hall effective field can reach the propagation field, which was simultaneously reduced
due to Joule heating effect of the current in our Pt/Co/Cr devices. The concept presented here will be
useful in racetrack memories and logic devices and investigating the effects of SOT on the DW
displacement.
Acknowledgments
This work was supported by the Program for Changjiang Scholars and Innovative Research Team in
University PCSIRT (No. IRT16R35), the National Natural Science Foundation of China (No. 51671098),
the Natural Science Foundation of Gansu Province (No. 17JR5RA210) and the Fundamental Research
Funds for the Central Universities (lzujbky-2015-122).
Reference
[1] Liu, L., Lee, O. J., Gudmundsen, T. J., Ralph, D. C., and Buhrman, R. A. Current-induced switching of
perpendicularly magnetized magnetic layers using spin torque from the spin Hall effect. Phys. Rev. Lett. 109, 096602
(2012).
[2] Miron, I. M., Garello, K., Gaudin, G., Zermatten, P.-J., Costache, M. V., Auffret, S., Bandiera, S., Rodmacq, B.,
Schuhl, A., and Gambardella, P. Perpendicular switching of a single ferromagnetic layer induced by in-plane current
injection. Nature 476, 189 (2011).
[3] Jamali, M., Narayanapillai, K., Qiu, X. P., Loong, L. M., Manchon, A., and Yang, H. Spin-orbit torques in Co/Pd
12
multilayer nanowires. Phys. Rev. Lett. 111, 246602 (2013).
[4] Emori, S., Bauer, U., Ahn, S.-M., Martinez, E., and Beach, G. S. D. Current-driven dynamics of chiral ferromagnetic
domain walls. Nat. Mater. 12, 611 (2013).
[5] Haazen, P. P. J., Murè, E., Franken, J. H., Lavrijsen, R., Swagten, H. J. M., and Koopmans, B. Domain wall depinning
governed by the spin Hall effect. Nat. Mater. 12, 299 (2013).
[6] Moore, T. A., Miron, I. M., Gaudin, G., Serret, G., Auffret, S., Rodmacq, B., Schuhl, A., Pizzini, S., Vogel, J., and
Bonfim, M. High domain wall velocities induced by current in ultrathin Pt/Co/AlOx wires with perpendicular magnetic
anisotropy. Appl. Phys. Lett. 93, 262504 (2008).
[7] Miron, I. M., Moore, T., Szambolics, H., Buda-Prejbeanu, L. D., Auffret, S., Rodmacq, B., Pizzini, S., Vogel, J.,
Bonfim, M., Schuhl, A., and Gandi, G. Fast current-induced domain-wall motion controlled by the Rashba effect. Nat.
Mater. 10, 419 (2011).
[8] Chiba, D., Kawaguchi, M., Fukami, S., Ishiwata, N., Shimamura, K., Kobayashi, K., and Ono, T. Electric-field control
of magnetic domain-wall velocity in ultrathin cobalt with perpendicular magnetization. Nat. Commun. 3, 888 (2012).
[9] Ryu, K.-S., Thomas, L., Yang, S.-H., and Parkin, S. Chiral spin torque at magnetic domain walls. Nat. Nanotechnol. 8,
527 (2013).
[10] Yang, S.-H., Ryu, K.-S., and Parkin, S. Domain-wall velocities of up to 750 m/s−1 driven by exchange-coupling
torque in synthetic antiferromagnets. Nat. Nanotechnol. 10, 221 (2015).
[11] Zhang, W. F., Han, W., Jiang, X., Yang, S. H., and Parkin, S. S. P. Role of transparency of platinum–ferromagnet
interfaces in determining the intrinsic magnitude of the spin Hall effect. Nat. Phys. 11, 496 (2015).
[12] Nguyen, M. H., Ralph, D. C., and Buhrman, R. A. Spin torque study of the spin Hall conductivity and spin diffusion
length in platinum thin films with varying resistivity. Phys. Rev. Lett. 116, 126601 (2016).
[13] Vélez, S., Golovach, V. N., Bedoya-Pinto, A., Isasa, M., Sagasta, E., Abadia, M., Rogero, C., Hueso, L. E., Bergeret,
F. S., and Casanova, F. Hanle magnetoresistance in thin metal films with strong spin-orbit coupling. Phys. Rev. Lett. 116,
016603 (2016).
[14] Liu, L., Pai, C. F., Li, Y., Tseng, H. W., Ralph, D. C., and Buhrman, R. A. Spin torque switching with the giant spin
Hall effect of tantalum. Science 336, 555 (2012).
[15] Yu, G. Q., Upadhyaya, P., Wong, K. L., Jiang, W. J., Alzate, J. G., Tang, J. S., Amiri, P. K., and Wang, K. L.
Magnetization switching through spin-Hall-effect-induced chiral domain wall propagation. Phys. Rev. B 89, 104421
(2014).
[16] Ramaswamy, R., Qiu, X., Dutta, T., Pollard, S. D., and Yang, H. Hf thickness dependence of spin-orbit torques in
13
Hf/CoFeB/MgO heterostructures. Appl. Phys. Lett. 108, 202406 (2016).
[17] Pai, C.-F., Liu, L. Li, Y., Tseng, H. W., Ralph, D. C., and Buhrman, R. A. Spin transfer torque devices utilizing the
giant spin Hall effect of tungsten. Appl. Phys. Lett. 101, 122404 (2012).
[18] Allen, G., Manipatruni, S., Nikonov, D. E., Doczy, M., and Young, I. A. Experimental demonstration of the
coexistence of spin Hall and Rashba effects in β-tantalum/ferromagnet bilayers. Phys. Rev. B 91, 144412 (2015).
[19] Behera, N., Chaudhary, S., and Pandya, D. K. Anomalous anti-damping in sputtered β-Ta/Py bilayer system. Sci. Rep.
6, 19488 (2016).
[20] Zhang, W., Han, W., Jiang, X., Yang, S.-H., and Parkin, S. S. P. Role of transparency of platinum–ferromagnet
interfaces in determining the intrinsic magnitude of the spin Hall effect. Nat. Phys. 11, 496 (2015).
[21] Huang, K.-F., Wang, D.-S., Lin, H.-H., and Lai, C.-H. Engineering spin-orbit torque in Co/Pt multilayers with
perpendicular magnetic anisotropy. Appl. Phys. Lett. 107, 232407 (2015).
[22] Zhang, W., Jungfleisch, M. B., Freimuth, F., Jiang, W., Sklenar, J., Pearson, J. E., Ketterson, J. B., Mokrousov, Y.,
and Hoffmann, A. All-electrical manipulation of magnetization dynamics in a ferromagnet by antiferromagnets with
anisotropic spin Hall effects. Phys. Rev. B 92, 144405 (2015).
[23] Demasius, K.-U., Phung, T., Zhang, W., Hughes, B. P., Yang, S.-H., Kellock, A., Han, W., Pushp, A., and Parkin, S. S.
P. Enhanced spin-orbit torques by oxygen incorporation in tungsten films. Nat. Commun. 7, 10644 (2016).
[24] Woo, S., Mann, M., Tan, A. J., Caretta, L., and Beach, G. S. D. Enhanced spin-orbit torques in Pt/Co/Ta
heterostructures. Appl. Phys. Lett. 105, 212404 (2014).
[25] Yu, J. W., Qiu, X. P., Legrand, W., and Yang, H. Large spin-orbit torques in Pt/Co-Ni/W heterostructures. Appl. Phys.
Lett. 109, 042403 (2016).
[26] Ueda, K., Pai, C. F., Tan, A. J., Mann, M., and Beach, G. S. D. Effect of rare earth metal on the spin-orbit torque in
magnetic heterostructures. Appl. Phys. Lett. 108, 232405 (2016).
[27] Yu, G. Q., Upadhyaya, P., Fan, Y. B., Alzate, J. G., Jiang, W. J., Wong, K. L., Takei, S., Bender, S. A., Chang, L. T.,
Jiang, Y., Lang, M. R., Tang, J. S., Wang, Y., Tserkovnyak, Y., Amiri, P. K., and Wang, K. L. Switching of perpendicular
magnetization by spin-orbit torques in the absence of external magnetic fields. Nat. Nanotechnol. 9, 548 (2014).
[28] Pai, C.-F., Mann, M., Tan, A. J., and Beach, G. S. D. Determination of spin torque efficiencies in heterostructures
with perpendicular magnetic anisotropy. Phys. Rev. B 93, 144409 (2016).
[29] Lau, Y.-C., Betto, D., Rode, K., Coey, J. M. D., and Stamenov, P. Spin-orbit torque switching without an external
field using interlayer exchange coupling. Nat. Nanotechnol. 11, 758 (2016).
[30] Fukami, S., Zhang, C. L., DuttaGupta, S., Kurenkov, A., and Ohno, H. Magnetization switching by spin-orbit torque
14
in an antiferromagnet/ferromagnet bilayer system. Nat. Mater. 15, 535 (2016).
[31] Kong, W. J., Ji, Y. R., Zhang, X., Wu, H., Zhang, Q. T., Yuan, Z. H., Wan, C. H., Han, X. F., Yu, T., Fukuda, K.,
Naganuma, H., and Tung, M.-J. Field-free spin Hall effect driven magnetization switching in Pd/Co/IrMn exchange
coupling system. Appl. Phys. Lett. 109, 132402 (2016).
[32] Cai, K. M., Yang, M., Ju, H., Wang, S., Ji, Y., Li, B., Edmonds, K. W., Sheng, Y., Zhang, B., Zhang, N., Liu, S.,
Zheng, H., and Wang, K. Electric field control of deterministic current-induced magnetization switching in a hybrid
ferromagnetic/ferroelectric structure. Nat. Mater. 16, 712(2017).
[33] Cui, B., Li, D., Chen, S., Yun, J., Zuo, Y., Guo, X., Wu, K., Zhang, X., Wang, Y., Yang, D., Gao, M. and Xi, L.
Current induced magnetization switching in Pt/Co/Cr structures with enhanced perpendicular magnetic anisotropy and
spin-orbit torques. ArXiv: 1708.09174.
[34] Fert, A. Magnetic and transport properties of metallic multilayers. Mater. Sci. Forum 59-60, 439 (1990).
[35] Fert, A. and Levy, P. M. Role of anisotropic exchange interactions in determining the propertiesof spin-glasses. Phys.
Rev. Lett. 44, 1538 (1980).
[36] Dzialoshinskii, I. E. Thermodynamic theory of weak ferromagnetism in antiferromagnetic substances. Sov. Phys.
JETP 5, 1259 (1957).
[37] Moriya, T. Anisotropic superexchange interaction and weak ferromagnetism. Phys. Rev. 120, 91 (1960).
[38] Júe, E., Thiaville, A., Pizzini, S., Miltat, J., Sampaio, J., Buda-Prejbeanu, L. D., Rohart, S., Vogel, J., Bonfim, M.,
Boulle, O., Auffret, S., Miron, I. M., and Gaudin, G. Domain wall dynamics in ultrathin Pt/Co/AlOx microstrips under
large combined magnetic fields. Phys. Rev. B 93, 014403 (2016).
[39] Rojas-Sánchez, J.-C., Laczkowski, P., Sampaio, J., Collin, S., Bouzehouane, K., Reyren, N., Jaffrès, H., Mougin, A.,
and George, J.-M. Perpendicular magnetization reversal in Pt/[Co/Ni]3/Al multilayers via the spin Hall effect of Pt. Appl.
Phys. Lett.108, 082406 (2016).
[40] Khan, R. A., Shepley, P. M., Hrabec, A., Wells, A. W. J., Ocker, B., Marrows, C. H., and Moore, T. A. Effect of
annealing on the interfacial Dzyaloshinskii-Moriya interaction in Ta/CoFeB/MgO trilayers. Appl. Phys. Lett. 109, 132404
(2016).
[41] Yang, M., Cai, K., Ju, H., Edmonds, K. W., Yang, G., Liu, S., Li, B., Zhang, B., Sheng, Y., Wang, S., Ji, Y., and Wang,
K. Spin-orbit torque in Pt/CoNiCo/Pt symmetric devices. Sci. Rep. 6, 20778 (2016).
[42] Hubert, A., and Schäfer, R. Magnetic Domains: The Analysis of Magnetic Microstructures. (Springer, 1998).
[43] Emori, S., and Beach, G. S. D. Roles of the magnetic field and electric current in thermally activated domain wall
motion in a submicrometer magnetic strip with perpendicular magnetic anisotropy. J. Phys.: Condens. Matter 24, 024214
15
(2012).
[44] Khvalkovskiy, A. V., Cros, V., Apalkov, D., Nikitin, V., Krounbi, M., Zvezdin, K. A., Anane, A., Grollier, J., and Fert,
A. Matching domain-wall configuration and spin-orbit torques for efficient domain-wall motion. Phys. Rev. B 87, 020402
(2013).
[45] Yamaguchi, A., Ono, T., and Nasu, S., Miyake, K., Mibu, K., Shinjo, T. Real-space observation of current-driven
domain wall motion in submicron magnetic wires. Phys. Rev. Lett. 92, 077205 (2004).
[46] Jeudy, V., Mougin, A., Bustingorry, S., Savero Torres, W., Gorchon, J., Kolton, A. B., Lemaître, A., and Jamet, J-. P.
Universal pinning energy barrier for driven domain walls in thin ferromagnetic films. Phys. Rev. Lett. 117, 057201
(2016).
[47] Lin, W., Vernier, N., Agnus, G., Garcia, K., Ocker, B., Zhao, W., Fullerton, E. E., and Ravelosona, D. Universal
domain wall dynamics under electric field in Ta/CoFeB/MgO devices with perpendicular anisotropy. Nat. Commun. 7,
13532 (2016).
16
|
1904.10854 | 2 | 1904 | 2019-04-28T22:17:21 | Proposed high-power beta cells from MgAlB14-type icosahedral-boron semiconductors | [
"physics.app-ph",
"cond-mat.mtrl-sci"
] | Beta cells generate electric power as carrier-producing beta irradiation from incorporated radioisotopes bombard a series of p-n-junctions. However, radiation damage to the semiconductors commonly used in solar cells limits beta cells to extremely weak irradiations that generate concomitantly miniscule electric powers, e.g. micro-Watts. By contrast, beta cells that generate many orders-of-magnitude larger powers are possible with icosahedral boron-rich semiconductors since their bombardment-induced atomic displacements spontaneously self-heal. Furthermore, substitutions for Mg and Al atoms of icosahedral-boron-rich semiconductors based on the MgAlB14 structure can produce p-n junctions as electron transfers from doping-induced interstitial extra-icosahedral atoms convert some normally p-type materials to n-type. Moreover, electron-phonon interactions of the resulting readily displaceable interstitial cations with charge carriers foster their forming large polarons. Oppositely charged polarons repel one another at short range. These repulsions suppress the recombination of n-type with p-type polarons thereby increasing the beta-cell efficiency. All told, use of these icosahedral boron-rich semiconductors could enable beta cells with electric powers that are many orders of magnitude larger than those of existing beta cells. This development opens a new avenue for generating electricity from nuclear decays. | physics.app-ph | physics | Proposed high-power beta cells from MgAlB14-type icosahedral-boron semiconductors
David Emin
Department of Physics and Astronomy
University of New Mexico
Albuquerque, New Mexico 87131, USA
Beta cells generate electric power as carrier-producing beta irradiation from incorporated radioisotopes
bombard a series of p-n-junctions. However, radiation damage to the semiconductors commonly used in
solar cells limits beta cells to extremely weak irradiations that generate concomitantly miniscule electric
powers, e.g. micro-Watts. By contrast, beta cells that generate many orders-of-magnitude larger powers
are possible with icosahedral boron-rich semiconductors since their bombardment-induced atomic
displacements spontaneously self-heal. Furthermore, substitutions for Mg and Al atoms of icosahedral-
boron-rich semiconductors based on the MgAlB14 structure can produce p-n junctions as electron transfers
from doping-induced interstitial extra-icosahedral atoms convert some normally p-type materials to n-
type. Moreover, electron-phonon interactions of the resulting readily displaceable interstitial cations with
charge carriers foster their forming large polarons. Oppositely charged polarons repel one another at short
range. These repulsions suppress the recombination of n-type with p-type polarons thereby increasing the
beta-cell efficiency. All told, use of these icosahedral boron-rich semiconductors could enable beta cells
with electric powers that are many orders of magnitude larger than those of existing beta cells. This
development opens a new avenue for generating electricity from nuclear decays.
I. INTRODUCTION
The utility of terrestrial solar cells as primary power sources is challenged by the solar flux's
capricious variability arising from clouds, humidity, dirt and sand and its predictable daily and seasonal
variations. By contrast, energy fluxes from products of nuclear decays are steadier and predictable.
Solid-state conversion of the energies of beta-rays into electrical energy utilize beta cells, devices
that are analogous to solar cells. Beta cells that employ conventional semiconductors are severely limited
due to their rapid degradation due to radiation damage.1,2 As a result, current beta-cells use only very
weak beta-irradiation to generate extremely small electric powers (e.g. micro-Watts).
The combination of three distinctive properties of the small subset of icosahedral boron-rich
semiconductors based on the MgAlB14 structure enable them to overcome this severe limitation. First,
electron-bombardment damage to icosahedral boron-rich semiconductors spontaneously self-heals.3,4
Second, some substitutions for Mg and Al interstitials within the network of boron icosahedra generate
the moderate mobility n-type carriers needed for useful p-n junctions while smaller electron donation to
the electron-deficient boron network yields its p-type carriers.5 Third, interactions of the resulting
displaceable interstitial cations with carriers foster their forming large polarons.6,7 The short-range
repulsion between oppositely charged polarons impedes their recombination thereby enhancing the
carrier-separation efficiency of the beta cell's p-n junctions.8,9
As illustrated in Sec. 6, the resulting beta cells could generate electric powers and possess energy
capacities that are many orders of magnitude larger than those of existing beta cells. Such devices would
greatly increase the utility of solid-state conversion of nuclear energy to electrical energy.
II. STRUCTURE AND SELF-HEALING OF ICOSAHEDRAL BORON-RICH INSULATORS
Figure 1 illustrates the structure of the simple icosahedral boron-rich insulators, B12P2 and
B12As2.4,10 Distinctively, these solids contain boron atoms that reside at the twelve vertices of icosahedra.
Each of these atoms forms a conventional two-center bond with an atom external to its icosahedron. By
contrast, the thirteen internal-bonding molecular orbitals of an icosahedron distribute charge about the
centers of its twenty faces. A boron icosahedron has an affinity for two electrons since the sum of the 12
2
electrons donated to its external covalent bonds plus the 26 = 2 13 electrons needed to fill its internal
bonding orbitals exceeds the 36 = 3 12 second-shell electrons provided by twelve boron atoms.
Massive high-energy bombardments with electrons and ions readily knock atoms of icosahedral
boron-rich insulators from their equilibrium positions.3,4 However, as shown in Fig. 2, the amorphization
and defect clustering that occurs for other semiconductors is not observed for icosahedral boron-rich
semiconductors. Rather, such damage apparently self-heals as bombardment-induced vacancies and
interstitials recombine.3
This behavior is fostered by an icosahedron retaining its internal bonding electrons when an atom
is knocked from it. The atom then exits as a cation as the icosahedron garners an additional negative
charge. Indeed, icosahedra which are "degraded" by the loss of an atom remain structurally stable as they
each accumulate an extra negative charge.11 Self-healing occurs as degraded icosahedra recombine with
rapidly diffusing small interstitial boron cations.3
III. CHARGE CARRIERS IN SEMICONDUCTORS ARE EITHER FREE, SMALL POLARONS
OR LARGE POLARONS
The adiabatic principle governs the state of a semiconductor's charge carriers.6,7,9,12 As shown in
Fig. 3, there are two distinct adiabatic solutions for a charge carrier in an isotropic covalent material with
its short-range (e.g. deformation-potential) electron-phonon interaction. Either an electronic carrier
remains free or it collapses into a severely localized self-trapped state thereby forming a small polaron.
Unlike free carriers, small polarons generally move incoherently by low-mobility phonon-assisted
hopping.
A different situation prevails in semiconductors having significant densities of displaceable ions.
As illustrated in Fig. 4, with only the long-range electron-phonon interaction arising from an electronic
carrier's Coulomb interactions with displaceable ions, the adiabatic treatment for an isotropic material
only permits large-polaron formation. However, adding a sufficiently strong short-range component to the
long-range electron-phonon interaction drives a large polaron's collapse into a small polaron.
A large polaron's self-trapped electronic carrier is bound within the potential well produced by
carrier-stabilized shifts of surrounding ions' equilibrium positions from their carrier-free values. Since a
large-polaron's motion is contingent on movements of these ions, it moves slowly with a very large
effective mass. A coherently moving large polaron's very large effective mass usually ensures its weak-
scattering by ambient phonons.7,13 A large polaron's long scattering time then compensates for its large
effective mass to produce a moderate mobility, 1-100 cm2/V-s at 300 K.
The minimum mobility for coherently moving charge carriers occurs when their mean-free-path
falls to their de Broglie wavelength.7,14 This minimum mobility then depends on the charge carrier's
effective mass m*, its charge e, the thermal energy kT, and Planck's constant h: min eh/m*kT.7 For a
free carrier with an effective mass as large as the free-electron mass me min = 300 cm2/V-s at 300 K. By
contrast, the huge effective mass of a large polaron m* >> me gives min 1 cm2/V-s at 300 K.
Large polarons are identified by the distinctive frequency dependence of their electrical
conductivity.7,15 Large-polarons' long scattering time restricts their Drude response to frequencies below
those of the associated phonons. Meanwhile, excitation of a large polaron's self-trapped electronic carrier
from the potential well within which it is bound produces a broad absorption band at frequencies above
those of the associated phonons. The pseudo gap between these two features opens as lowering the
temperature increases the Drude conductivity's scattering time.
Measurements of the static and high-frequency dielectric constants of condensed matter indicate
the presence of displaceable ions. In covalent semiconductors, the ratio of the static to high-frequency
dielectric constants 0/, is only slightly greater than one. By contrast, the displaceable ions of simple
ionic solids, e.g. alkali halide crystals, generates dielectric-constant ratios of 0/ 2. Moreover, in
3
materials with significant densities of especially displaceable ions, e.g. cuprate superconductors and
hybrid organic-inorganic halide perovskite solar cells, 0/ >> 2.16-21
IV. RECOMBINATION OF OPPOSITELY CHARGED POLARONS
Solar-cell-type devices operate by the electric field produced near the junction between n-type
and p-type semiconductors separating photo-carriers before they can recombine. Conventional solar cells
utilize covalent materials (e.g. Si or GaAs) whose high-mobility free carriers separate rapidly enough to
forestall their rapid recombination. Unconventional solar-cells employ ionic semiconductors (e.g. metal
and organometal halide perovskites) whose large polarons have much lower mobilities than those of high-
mobility covalent solids.21-26 However, recombination of oppositely charged large polarons as well as
oppositely charged small polarons are suppressed by their mutual short-range repulsions.7-9
Figure 5 schematically depicts the net energy of an isotropic medium's oppositely charged
polarons E(s) as a function of their separation s. The net energy approaches that of two independent
oppositely charged polarons, (Ep Ep), when their mutual separation greatly exceeds the sum of their
radii, Rp and Rp. As s falls toward Rp Rp, constructive interference of the equilibrium positions of
intervening ions enhances the long-range binding of the two oppositely charged polarons. However, as
s/(Rp Rp) falls below unity the net binding energy of the oppositely charged polarons approaches zero
as the Coulomb fields that displace the equilibrium positions of surrounding ions progressively cancel one
another. At s = 0 the oppositely charged carriers merge into an exciton of energy Eex. With no net charge,
the exciton lacks the Coulomb field to displace equilibrium positions of distant surrounding ions.
A domain for which E(s)/s < 0 indicates a repulsive interaction between oppositely charged
carriers. The condition (Ep Ep) > Eex is sufficient, but not necessary, to produce such a repulsive
interaction. The energy of independent electron and hole polarons separated by the distance s and
attracted by their mutual Coulomb interaction is:
𝐸(𝑠) = − [
𝑒2
4𝑅𝑝−
1
(
𝜀∞
−
1
𝜀0
) + 𝐸𝑏−] − [
𝑒2
4𝑅𝑝+
1
(
𝜀∞
−
1
𝜀0
) + 𝐸𝑏+] −
𝑒2
𝜀0𝑠
. (1)
The electron and hole polaron binding energies are each the sum of a long-range component and a short
range component. Each of the long-range components explicitly depends on its respective polaron radius,
Rp or Rp, and on the material's static and high-frequency dielectric constants, 0 and . The short-range
components for the electron- and hole-polaron binding energies are denoted by Eb and Eb, respectively.
The energy of the exciton is
𝐸𝑒𝑥 = − (
𝑒2
2𝜀∞𝑅𝑒𝑥
+ 𝐸𝑏,𝑒𝑥), (2)
where Rex denotes the exciton radius and Eb,ex represents the correction to the exciton's binding energy,
the exciton's self-trapping energy, generated by its electronic carriers' altering the equilibrium positions
of the atoms they contact.
In the covalent limit, 0 , with s >> Rex:
𝐸(∞) − 𝐸𝑒𝑥 =
𝑒2
2𝜀∞𝑅𝑒𝑥
+ 𝐸𝑏,𝑒𝑥 − 𝐸𝑏− − 𝐸𝑏+. (3)
In covalent semiconductors, oppositely charged conventional non-polaronic carriers attract one another,
E() > Eex, since then Eb = Eb = 0. However, oppositely charged polarons will experience a short-range
repulsion when Eb Eb is sufficiently large.7,9
In semiconductors with exceptionally displaceable ions, 0 >> , the exciton's Coulomb term
tends to be overwhelmed by the long-range contributions to the polarons' binding energies:
𝐸(∞) − 𝐸𝑒𝑥 = −
𝑒2
4𝜀∞𝑅𝑝−
−
𝑒2
4𝜀∞𝑅𝑝+
4
+
𝑒2
2𝜀∞𝑅𝑒𝑥
− 𝐸𝑏− − 𝐸𝑏+ + 𝐸𝑏,𝑒𝑥, (4)
where Rex > Rp and Rp, since the Wannier exciton's reduced mass is less than the electronic effective
masses of its electron and its hole. Furthermore, since polarons are charged while an exciton is neutral,
the short-range contributions to polarons' binding energies tend to greatly exceed those for an exciton: Eb
Eb >> Eb,ex. Thus, oppositely charged polarons in materials with especially displaceable ions, 0 >> ,
tend to repel one another: E() < Eex. This short-range repulsion suppresses recombination of oppositely
charged polarons.
All told, in materials with exceptionally displaceable ions, 0 >> , oppositely charged polarons
can repel one another and polarons with the same signed charge can attract one another. In particular,
large polarons of the same charge tend to merge to form large bipolarons in materials for which the ratio
of static to high-frequency dielectric constants is exceptionally large, 0/ >> 2.7,27,28 Moreover, the novel
superconductivity of perovskite-based cuprates has been attributed to the ground-state of large bipolarons
that condense into a liquid under the influence of their mutual phonon-assisted attraction.29-32
V. ICOSAHEDRAL BORON-RICH SEMICONDUCTORS
Semiconductors generated by doping the wide-gap covalent icosahedral boron-rich insulators
B12P2 and B12As2 were suggested for use in radiation-hard beta cells.33 These are devices that are powered
by irradiation with energetic electrons emitted during radio-isotopes' beta decay. Since these icosahedral
boron-rich insulators have fourteen atoms per unit cell, their transport bands comprise many narrow
energy bands. As a result, their free-carrier effective masses are much larger and their free-carrier
mobilities are very much smaller than those of Si and GaAs, the semiconductors utilized in conventional
solar cells. Furthermore, while p-type materials are easily realized, substitutional doping has yet to
produce an n-type material.34 All told, these features dampen the prognosis for producing efficient beta
cells by substitutional doping of these covalent icosahedral boron-rich semiconductors.
By contrast, both n-type and p-type materials have been produced from icosahedral boron-rich
semiconductors based on the MgAlB14 structure. As schematically depicted in Fig. 6, the orthorhombic
unit cell of MgAlB14 structure contains four twelve-boron-atom icosahedra divided equally between two
different orientations plus four pairs of extra-icosahedral boron atoms.35,36 Each of these eight boron
atoms forms two-center bonds with three different adjacent icosahedra. This structure's two metal atoms
partially occupy sites within two inequivalent large extra-icosahedral open spaces. These materials can be
doped by making substitutions for these metal cations.5,37
The formula (RY)y(MX)xB14 describes the MgAlB14 structure having cations of metal R with
valence Y and partial occupancy y and cations of metal M with valence X and partial occupancy x.
These metals donate xX yY electrons to a formula unit's structure. Two electrons are required to fill its
icosahedron's internal bonding orbitals and two electrons are needed to produce a bond between its two
non-icosahedral boron atoms. Thus, in this idealized situation, four electrons must be donated to each
formula unit to just fill all of its bonding orbitals, thereby producing an insulator. Reducing this donation
will generate p-type materials and increasing this donation will produce n-type materials.
Electronic transport measurements were recently reported on members of the series
(Y3)0.56(Al3)xB14.5 The room-temperature dc resistivity is largest for x = 0.53, somewhat smaller for x =
0.41 and five orders-of-magnitude smaller for x = 0.63. Furthermore, measurements of the Seebeck
coefficient for x = 0.41 indicate p-type conduction while those for x = 0.63 indicate n-type conduction.
Quantitative estimates suggest moderate mobility n-type carriers and very low mobility p-type carriers.
These icosahedral-boron semiconductors have cations distributed among partially occupied sites
off the boron-network. These cations' Coulomb interactions with charge carriers generate long-range
5
components of their electron-phonon interactions. One can then envision p-type small polarons moving
on the boron network and n-type large polarons moving between the cations outside of the boron network.
Thus, p-n junctions can be formed from these icosahedral boron-rich semiconductors possessing
different values of x. Cations that are sufficiently displaceable to produce 0/ >> 1 necessarily generate
a barrier to recombination of oppositely charged polarons. The suppression of recombination of n-type
large polarons with p-type small polarons will enhance the efficiency of solar-cell-type devices based on
these p-n junctions.
VI. BETA CELLS OF ICOSAHEDRAL BORON-RICH SEMICONDUCTORS
The solar intensity on earth varies greatly and capriciously with the weather and predictably with
the time of day, latitude and season. For example, the average monthly solar intensity in London has a
minimum of less than 0.003 W/cm2 in January and maximum of about 0.02 W/cm2 in July with a yearly
average intensity of about 0.01 W/cm2.38 Each absorbed photon potentially generates an electron-hole
pair.
By contrast, the power emitted from a radioisotope falls steadily with time as it decays. For
example, the power emitted from materials containing beta-emitting 90Sr (e.g. Sr metal or SrTiO3) falls
from an initial value of 2 W/cm3 as it decays over its 28 year half-life.1 Because the average energy of
each bombarding beta electron is about 1 MeV, it induces between 105 and 106 electron-hole pairs.
Whereas a solar cell is powered by sunlight that impinges on it, the beta cell schematically
illustrated in Fig. 7 is powered by the beta decay of radioisotopes contained within it. As such, a beta cell
can employ linked p-n junctions amidst a distribution of radioisotopes. Shielding encases the beta cell to
capture radiation (e.g. Bremsstrahlung) generated within it in addition to unabsorbed radiation from its
radioisotopes. The shielding also conducts heat from this self-contained power source.
For example, if 90Sr comprises 1% of a 1 cm3 beta cell, its initial power output, 0.02 W, is
comparable to that of a 1 cm2 solar cell in London at mid-summer with similar p-n junction efficiency.
The net energy capacity of the 1 cm3 beta cell, 0.8 W-year [= 0.02 W 28 year ln(2)], is about 350
times that of a very good D-cell battery, 20 W-hr.
A beta-cell's output and capacity increase with the volume of effectively employed radioisotope.
Thus, the power and net capacity of a cubic meter of beta cells would be a million times larger than that
from a 1 cm3 beta cell.
VII. DISCUSSION
A boron icosahedron retains its internal bonding electrons when it is degraded by the loss of a
boron atom. As a result, a bombardment-induced degraded boron icosahedron takes an electron from the
departing boron atom. This bombardment-induced damage then self-heals as interstitial boron cations
spontaneously recombine with degraded icosahedra. Thus, bombardment-induced damage to icosahedral
boron-rich semiconductors self-heals. As a result, the lifetimes of beta cells based on icosahedral boron-
rich semiconductors will not be limited by bombardment-induced atomic displacements.
Beta cells require junctions between p-type and n-type icosahedral boron-rich semiconductors.
Unfortunately, since icosahedral boron networks are electron-deficient covalent networks, icosahedral
boron-rich semiconductors usually manifest p-type transport. Nonetheless, icosahedral boron-rich
semiconductors based on the MgAlB14 structure can be doped p- and n-type by altering the valences and
partial occupations of metal cations in large holes between icosahedra.
Displaceable cations generate ratios of a material's static to high-frequency dielectric constants
that greatly exceed unity, their values in covalent solids. Charge carriers' Coulomb interactions with
displaceable cations also produce the long-range electron-phonon interactions that foster large-polaron
6
formation. Large polarons are identified by their moderate mobilities and by the existence of pseudo-gaps
in their distinctive frequency-dependent absorption spectra.15
Unlike the situation for conventional charge carriers in covalent semiconductors, an energy
barrier impedes the recombination of oppositely charged polarons. This suppression of recombination can
significantly enhance the carrier-generation efficiency in polaron materials including biological matter.
Thus, beta cells that utilize MgAlB14-type icosahedral boron-rich semiconductors for their p-n junctions
offer the possibility of more efficient, longer lived and higher capacity beta cells than have heretofore
been possible.
Indeed, remarkably high-efficiency solar cells that utilize metal and organometallic halide
materials with the perovskite crystal structure have been discovered.39 Their high efficiencies have been
attributed to squelched recombination of moderate-mobility large polarons whose formation is promoted
by these materials' very large ratios of static to high-frequency dielectric constants.8,22-25 The absorption
spectra expected of large polarons have even been observed.26
Finally, just as short-range repulsions between oppositely charged polarons impede their
recombination, short-range attractions between polarons of like charge leads to their pairing as
bipolarons.7,9 In particular, materials with exceptionally large ratios of their static to high-frequency
dielectric constants foster large-bipolaron formation.27,28 The large dielectric-constant ratios of cuprate
superconductors result from displaceable ions that reside beyond the CuO2 layers within which normal
conduction occurs.16-18 The superconductivity of suitably doped perovskite-based cuprates has been
ascribed to the ground-state of the liquid of large bipolarons generated by their mutual phonon-mediated
mutual attraction.29-32
7
1. W. R. Corlis and D. G. Harvey, Radioisotopic power generation (Prentice-Hall, Englewood Cliffs, N. J., 1964)
Sec. 9.3.
2. P. Rappaport, J. Loferski, and E. Linder, The electron voltaic effect in germanium and silicon p-n junctions, RCA
Reviews 17, 100-128 (1956).
3. M. Carrard, D. Emin, and L. Zuppiroli, Defect clustering and self-healing of electron-irradiated boron-rich solids,
Phys. Rev. B 51(17), 11270-11274 (1995).
4. D. Emin, Unusual properties of icosahedral boron-rich solids, J. of Solid State Chemistry 179 2791-2798 (2006).
5. T. Mori, Perspectives of high-temperature thermoelectric applications and p-type and n-type aluminoborides, The
minerals, Metals and Materials Society, JOM 68/10, 2673-2679 (2016) DOI: 10.1007/s11837-016-2069-9.
6. D. Emin and T. Holstein, Adiabatic theory of an electron in a deformable continuum, Phys. Rev. Lett. 36 323-326
(1976).
7. D. Emin, Polarons (Cambridge Univ. Press, Cambridge, 2013) Chaps. 4, 7, 10.
8. D. Emin, Barrier to recombination of oppositely charged large polarons, J. Appl. Phys. 123, 055105 (2018).
9. D. Emin, Small polarons, Physics Today 35/6, 34-40 (1982).
10. D. Emin, Physics Today 40/1 55-62 (1987).
11. M. F. Hawthorne, D. C. Young, and P. A. Wegner, Carbametallic boron hydride derivatives, I. Apparent analogs
of ferroene and ferricinium ion, J. Am. Chem. Soc. 87 1818-1819 (1965).
12. Y. Toyozawa, Self-trapping of an electron by the acoustical mode of lattice vibration. I. Prog. Theor. Phys. 26
29-44 (1961).
13. H.-B. Schuttler and T. Holstein, Dynamics and transport of a large acoustic polaron in one dimension, Ann.
Phys. (N.Y.) 166, 93-163 (1986).
14. C. Herring, The current state of transport theory, Proc. Int. Conf. on Semiconductor Physics (Prague, 1960)
pp.60-67.
15. D. Emin, Optical properties of large and small polarons and bipolarons, Phys. Rev. B 48, 13691-13702 (1993).
16. D. Reagor, E. Ahrens, S.-W. Cheong, A. Migliori, and Z. Fisk, Large dielectric constants and massive carriers in
La2CuO4, Phys. Rev. Lett. 62, 2048-2051 (1989).
17. G. A. Samara, W. F. Hammetter, and E. L. Venturini, Temperature and frequency dependences of the dielectric
properties of YBa2Cu3O6x, Phys. Rev. B 41, 8974-8980 (1990).
18. G. Cao, J. E. O'Reilly, J. B. Crow, and L. R. Testardi, Enhanced electric polarizability at the magnetic ordering
temperature of La2CuO4x, Phys. Rev. B 47, 11510-11511 (1993).
19. E. J. Juarez-Perez, R. S. Sanchez, L. Badia, G. Garcia-Belmonte, Y. S. Kang, I. Mora-Sero, and J. Bisquert,
Photoinduced giant dielectric constant in lead halide perovskite solar cells, Phys. Chem Lett. 5, 2390-2394 (2014).
20. A. Guerrero, G. Garcia-Belmonte, I. Mora-Sero, J. Bisquert, Y. S. Kang, T. J. Jacobson, J. P. Correa-Baena, and
A. Hagfeldt, Properties of contact and bulk impedances in hybrid lead halide perovskite solar cells including
inductive loop elements, J. Phys. Chem. C 120, 8023-8032 (2016).
21. M. Bonn, K. Miyata, E. Hendry, and X.-Y. Zhu, Role of dielectric drag in polaron mobility in lead halide
perovskites, ACS Energy Lett. 2, 2555-2562 (2017).
8
22. A. J. Neukirch, W. Nie, J.-C. Blancon, K. Appavoo, H. Tsai, M. Y. Sfeir, C. Katan, L. Pedesseau, J. Even, J. J.
Crochet, G. Gupta, A. D. Mohite, and S. Tretiak, Polaron stabilization by cooperative lattice distortion and cation
rotations in hybrid perovskite materials, Nano Lett. 16, 3809-3816 (2016).
23. X.-Y. Zhu, and V. Podzorov, Charge carrier in hybrid organic-inorganic lead halide perovskites might be
protected as large polarons, J. Phys. Chem. Lett. 6, 4758-4761 (2015).
24. K. Zheng, M. Abdellah, Q. Zhu, Q. Kong, G. Jennings, C. A. Kurtz, M. E. Messing, Y. Niu, D. J. Gosztola, M.
J. Al-Marri, X. Zhang, T. Pullerits, and S. E. Canton, Direct experimental evidence for photo-induced strong-
coupling polarons in organolead halide perovskite nanoparticles, J. Phys. Chem. Lett. 7, 4535-4539 (2016).
25. K. Miyata, T. L. Atallah, and X.-Y. Zhu, Lead halide perovskites: Crystal-liquid duality, phonon glass electron
crystals and large polaron formation, Sci. Adv. 3, e1701469 (2017).
26. K. T. Munson, E. R. Kennehan, G. S. Doucetter, and J. B. Asbury, Dynamic disorder dominates delocalization,
transport, and recombination in halide perovskites, Chem. 4, 2826-2843 (2018).
27. D. Emin, Formation, motion and high-temperature superconductivity of large bipolarons, Phys. Rev. Lett. 62,
1544-1547 (1989).
28. D. Emin and M. S. Hillery, Formation of a large bipolaron: Application to high-temperature superconductivity,
Phys. Rev. B 39, 6575-6593 (1989).
29. D. Emin, Phonon-mediated attraction between large bipolarons: Condensation to a liquid, Phys. Rev. Lett. 72,
1052-1054 (1994).
30. D. Emin, Phonon-mediated attraction between large bipolarons: Condensation to a liquid, Phys. Rev. B 49,
9157-9167 (1994).
31. D. Emin, In-plane conductivity of a layered large-bipolaron liquid, Phil. Mag. 95, 918-934 (2015).
32. D. Emin, Dynamics d-symmetry Bose condensate of a planar-large-bipolaron-liquid in cuprate superconductors,
Phil. Mag. 97, 2931-2945 (2017).
33. T. L. Aselage and D. Emin, Beta cell device using icosahedral boride compounds, US Patent 6,479,919 B1
(November 12, 2002).
34. D. Emin, Bonding and doping of simple icosahedral-boride semiconductors, J. Solid State Chem. 177/4-5, 1619-
1623 (2004).
35. V. I. Matkovich and J. Economy, Structure of MgAlB14 and a brief critique of structural relationships in higher
borides, Acta Cryst. B 26, 616-621 (1970).
36. I. Higashi and T. Ito, Refinement of the structure of MgAlB14, J. Less-Common Metals 92, 239-246 (1983).
37. O. A. Golikova and I. Higashi, MgAlB14 -- Structure and doping, Proc. 11th Int. Symp. Boron, Borides and
Related Compounds, Jap. J. of Appl. Phys. Series 10, 52-53 (1994).
38. D. J. C. MacKay, Sustainable energy -- without the hot air (UIT, Cambridge, U. K., 2009) Chap. 6.
39. S. D. Stranks and H. J. Snaith, Metal-halide perovskites for photovoltaic and light-emitting devices, Nat.
Nanotechnol. 10, 391-402 (2015).
9
Fig. 1. The rhombohedral unit cell of the icosahedral boron-rich solids, B12P2 and B12 As2, has boron-
icosahedra at its corners and chains of two phosphorus or two arsenic atoms along its major diagonal.
10
Before bombardment After bombardment
Fig. 2. The HRTEM images (from L. Wang and R. Ewing in Ref. 4) of B12P2 before and after
bombardment with a beam of 400 keV electrons whose intensity, 1018 electrons/cm2-sec, is 106 times that
from undepleted 90Sr beta-emissions. No amorphization or defect clustering is seen after a net
bombardment of 1023 electrons/cm2, equivalent to that after 10,000 years of bombardment with constantly
replenished 90Sr.
11
Fig. 3. The adiabatic energy E(R) of a charge carrier with the short-range (deformation-potential-type)
electron-phonon interaction of a covalent solid is plotted against the polaron radius R in units of its
minimum value (about an atomic radius). This function's two minima correspond to the system's two
realizable states. The carrier either (1) collapses to a single site thereby forming a small-polaron or (2)
expands to an infinite radius thereby remaining a free carrier.
R1234E(R)0.000.050.100.15free carriersmall polaronR1234E(R)0.000.050.100.15
12
Fig. 4. The adiabatic energy E(R) of a charge carrier with the long-range electron-phonon interaction that
results from its Coulomb interactions with surrounding displaceable ions is plotted against the polaron
radius R in units of its minimum value (about an atomic radius). This function's solitary minimum
corresponds to the carrier forming a large polaron.
R1234E(R)-0.25-0.20-0.15-0.10-0.050.00large polaron
13
Fig. 5. The energy of two oppositely charged polarons E(s) in units of Ep Ep, the sum of their
individual binding energies, is plotted against their separation s in units of the sum of the two polaron
radii, Rp and Rp. As the inter-polaron separation is decreased from infinity, displacements of the
equilibrium positions of intervening ions induced by the two polarons causes their net energy to decrease.
However, the net inference of the ionic displacements changes from being constructive to being
destructive as the two polarons begin to overlap. The surrounding ions then increasingly see the two
polarons as being net neutral. The two polarons ultimately collapse into an exciton when s = 0. The red
portion of E(s) versus s highlights the region in which the polarons exhibit a short-range mutual repulsion.
s/(Rp++ Rp-)0123E(s)/(Ep++ Ep- )-1.4-1.0-0.6-0.2
14
Fig. 6. The orthorhombic unit cell of MgAlB14 encompasses four icosahedra divided between two
different orientations. Modifying the pictorial representations of Refs. 35and 37, hexagons' centers
indicate the centroids of icosahedra while their colors and diagonal lines both indicate their orientations.
Lightened colors indicate icosahedra that reside slightly below those of the cell's principal yz plane. Dots
indicate the locations of pairs of extra-icosahedral boron atoms which are each bonded to three
icosahedra. The large and small circles schematically indicate the idealized inequivalent partially
occupied inter-icosahedral locations for Mg (2) and Al(3) cations.
15
v
Fig. 7. Junctions between p-type material (blue) and n-type material (red) are connected in series by metal
leads (black). This beta cell is powered by encapsulated volumes of radioisotope (green). The cell is
encased in material (grey) that conducts heat to the outside while shielding it from radiation.
|
1709.02205 | 2 | 1709 | 2018-04-23T13:25:15 | High quality ultrafast transmission electron microscopy using resonant microwave cavities | [
"physics.app-ph"
] | Ultrashort, low-emittance electron pulses can be created at a high repetition rate by using a TM$_{110}$ deflection cavity to sweep a continuous beam across an aperture. These pulses can be used for time-resolved electron microscopy with atomic spatial and temporal resolution at relatively large average currents. In order to demonstrate this, a cavity has been inserted in a transmission electron microscope, and picosecond pulses have been created. No significant increase of either emittance or energy spread has been measured for these pulses.
At a peak current of $814\pm2$ pA, the root-mean-square transverse normalized emittance of the electron pulses is $\varepsilon_{n,x}=(2.7\pm0.1)\cdot 10^{-12}$ m rad in the direction parallel to the streak of the cavity, and $\varepsilon_{n,y}=(2.5\pm0.1)\cdot 10^{-12}$ m rad in the perpendicular direction for pulses with a pulse length of 1.1-1.3 ps. Under the same conditions, the emittance of the continuous beam is $\varepsilon_{n,x}=\varepsilon_{n,y}=(2.5\pm0.1)\cdot 10^{-12}$ m rad. Furthermore, for both the pulsed and the continuous beam a full width at half maximum energy spread of $0.95\pm0.05$ eV has been measured. | physics.app-ph | physics | High quality ultrafast transmission electron microscopy using resonant microwave
cavities
W. Verhoevena,∗, J. F. M. van Rensa, E. R. Kieftb, P. H. A. Mutsaersa, O. J. Luitena
aDepartment of Applied Physics, Coherence and Quantum Technology Group, Eindhoven University of Technology, P.O. Box 513, 5600
bThermo Fisher Scientific, Achtseweg Noord 5, 5651 GG Eindhoven, The Netherlands
MB Eindhoven, The Netherlands
8
1
0
2
r
p
A
3
2
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
2
v
5
0
2
2
0
.
9
0
7
1
:
v
i
X
r
a
Abstract
Ultrashort, low-emittance electron pulses can be created at a high repetition rate by using a TM110 deflection cavity
to sweep a continuous beam across an aperture. These pulses can be used for time-resolved electron microscopy with
atomic spatial and temporal resolution at relatively large average currents. In order to demonstrate this, a cavity has
been inserted in a transmission electron microscope, and picosecond pulses have been created. No significant increase of
either emittance or energy spread has been measured for these pulses.
At a peak current of 814 ± 2 pA, the root-mean-square transverse normalized emittance of the electron pulses is
εn,x = (2.7 ± 0.1) · 10−12 m rad in the direction parallel to the streak of the cavity, and εn,y = (2.5 ± 0.1) · 10−12 m rad
in the perpendicular direction for pulses with a pulse length of 1.1 -- 1.3 ps. Under the same conditions, the emittance of
the continuous beam is εn,x = εn,y = (2.5 ± 0.1) · 10−12 m rad. Furthermore, for both the pulsed and the continuous
beam a full width at half maximum energy spread of 0.95 ± 0.05 eV has been measured.
Keywords: ultrafast transmission electron microscopy, pump -- probe, microwave cavities, coherent ultrashort electron
pulses
1. Introduction
Ultrashort high quality electron pulses at energies
ranging from 30 to 200 keV have become a useful and
powerful tool to investigate dynamical systems on sub-
picosecond timescales through diffraction [1], imaging [2]
or spectroscopy [3], offering a vast amount of new informa-
tion. Typically, inside an ultrafast transmission electron
microscope (UTEM) electron pulses are extracted from a
photocathode using an intense pulsed laser. Accurately
timed with a clocking laser pulse, dynamic processes
can then be investigated with pump -- probe measure-
ments. Using photoemission, a very large operational
parameter-space can be spanned [4]. Furthermore, by
using sideways illumination of a Schottky emitter, the
emission characteristics of the source are maintained,
allowing for high quality electron pulses to be created [5].
Although photoemission is commonly used in UTEM
systems, there is an interesting alternative to use a blank-
ing method, where a continuous beam is periodically swept
across a slit or aperture [6, 7, 8]. Creating pulses in this
way has the advantages that amplified laser systems are
no longer required, and that no intrusive alterations to
the source have to be made.
Instead, the system bene-
fits from the vast amount of research done on state-of-the-
art continuous sources, including recent developments that
∗Corresponding author
Email address: [email protected] (W. Verhoeven)
Preprint submitted to Elsevier
promise a higher brightness in the future [9]. Furthermore,
any possible instabilities in electron emission due to the in-
trinsic pointing stability of a drive laser are circumvented.
Recently, it has been shown that pulsing a beam can
be done using a microwave cavity oscillating in the TM110
mode while maintaining the low emittance of a continu-
ous source [10, 11, 12]. This can be accomplished using a
conjugate blanking scheme, where the electron beam is fo-
cused at the center of the cavity, allowing for 100 fs pulses
to be created with a high beam quality. Since the power
in the cavity can easily be adjusted, the pulse length can
also be changed without influencing the electron emission
process. In Fig. 1(a) this chopping principle is shown.
In order to perform pump -- probe experiments, the
phase of these microwave cavities can be accurately
synchronized to a pump laser pulse. Using state-of-the-
art synchronization schemes, timing jitter between the
electron pulses and the laser pulses can be suppressed to
levels well below 100 fs [13, 14].
Alternatively, it has been proposed to use a microwave
signal as a pump pulse to drive electronic or semiconductor
devices for laser-free stroboscopic imaging with repetition
rates in the GHz regime [15]. This is an interesting aspect
of using microwave cavities, as they can provide a higher
repetition rate and therefore a higher average current for
samples with a fast relaxation time.
For samples with slower relaxation times, it has been
proposed to use two perpendicular deflecting modes at dif-
August 27, 2018
terms of the reduced brightness, which can be defined in
differential form as [17]
Br = 1
V ∗
∂2I
∂A∂Ω ,
(1)
Since
with I the current through an area A at a solid angle Ω,
and V ∗ = (1/2 + γ/2)V the acceleration voltage V multi-
plied by a relativistic correction term, with γ the Lorentz
factor. The reduced brightness is a conserved quantity
during acceleration of the electrons.
the differential
reduced brightness varies
throughout the beam, its maximum on-axis value is often
used, called the axial or peak brightness. Within the
typical working regime of a microscope, a large portion
of the emitted electrons is cut away at the condenser
aperture,
leading to an approximately uniform current
distribution. After focusing the beam at semi-angle α,
this then results in a uniform angular distribution and a
Gaussian position distribution within the beam waist, so
that the peak brightness can be written as
I
2π2α2σxσy
Br = 1
V ∗
qe
mec2
=
I
4π2εn,xεn,y
,
(2)
with σx and σy the root-mean-square (RMS) size of the
beam waist, qe the electron charge, me the electron mass,
c the speed of light, and εn,x and εn,y the RMS normalized
emittance in the x and y direction respectively, given by
phx2ihp2
phx2ihx02i − hxx0i2 ,
xi − hxpxi2
εn,x = 1
mec
≈ γvz
c
(3)
with vz the velocity, px the transverse momentum and
x0 = vx/vz the angular distribution of the particles. In this
equation, h. . .i indicates the averaging over a distribution.
2.2. Beam chopping
The main advantage of using a microwave cavity is that
the low emittance of the continuous beam is maintained in
pulsed mode. This is only the case when using the cavity
in a conjugate blanking scheme, in which all electrons de-
flected by the cavity originate from the same virtual image.
For a regular beam blanker conjugate blanking is achieved
by placing a crossover in the pivot point of the blanker.
Inside a microwave cavity the fields vary rapidly com-
pared to the transit time of the electrons, so that it is no
longer possible to distinguish a single pivot point. How-
ever, it can be shown that it is still possible to maintain the
virtual image by proper placement of a crossover [10, 12].
For a beam chopped by an on-axis aperture, the optimal
longitudinal position of this crossover is at the center of
the cavity.
This is also shown in Fig. 1(a), where the beam is fo-
cused at the center of the cavity. As a result, it arrives at
Figure 1: (a) General principle of the creation of pulses using a
TM110 deflection cavity, where a continuous beam is deflected over
a chopping aperture. Definition of the parameters is discussed in
section 2.2. (b) A typical cavity, with (1) the entrance aperture, (2)
the antenna, (3) the tuning stub, (4) the dielectric material, and (5)
the lid used to close the cavity. Shown left is the side of the cavity,
and right is the bottom of the cavity with the lid removed.
ferent frequencies, which can be placed in a single cav-
ity [10]. Electrons will then be created at the difference
frequency of these modes, allowing for the repetition rate
to be lowered to tens of MHz.
If lower frequencies are
desired, a fast beam blanker can be used to pick specific
pulses, which are now separated by tens of ns.
In this
way, microwave cavities can also provide lower repetition
rates for samples with slow relaxation times, allowing for
the repetition rate of the setup to be optimized for each
experiment.
In order to facilitate the implementation in a TEM col-
umn, these deflection cavities can be filled with a dielectric
material, which allows for a reduction in both the size and
power consumption [16]. Figure 1(b) shows a typical di-
electric filled cavity used for chopping an electron beam.
Shown to the left is the outside of the cavity, and to the
right is a bottom view of the cavity with the lid removed.
In this paper, the implementation of a TM110 deflection
cavity in a TEM is presented. Design considerations are
discussed, and the performance of a cavity-based UTEM
is demonstrated.
2. Theory
2.1. Brightness
An important figure of merit for a charged particle
beam is its current density per unit of solid angle, called
the transverse brightness. As the solid angle subtended
by the beam, and therefore the brightness, depends on
the beam energy, the beam quality is often expressed in
2
~FL=B0qevzsinωtexxzswl(a)(1)(1)(2)(2)(3)(3)(5)(5)(2)(2)(3)(3)(4)(4)(b) 3. Methods
For the experiments, a 200 kV FEI Tecnai TEM has
been elongated with a 203 mm long vacuum chamber be-
low the C2 aperture.
In this chamber a TM110 cavity
has been mounted, and an additional aperture holder has
been inserted at the bottom. Above the cavity an extra
set of beam deflectors has been added. Figure 2 gives a
schematic overview of the adapted column. The distance
from the center of the cavity to the chopping aperture is
l = 122.2 mm. Both apertures are 30 µm in diameter.
In order to prevent an increase in emittance in pulsed
mode, a crossover is placed at the center of the cavity by
fixing the C2 lens current. The field-of-view is controlled
with the minicondenser (MC) lens. This means that its
original functionality of altering the divergence at the ob-
jective lens is now lost. However, the appropriate choice
of apertures can mimic this functionality.
A water-cooled cavity has been designed with a reso-
nant frequency ω/2π = 2.9985 GHz, and a length Lcav =
16.67 mm. The cavity is loaded with ZrTiO4, a dielectric
material with a high permittivity and low loss tangent.
The typical magnetic field amplitude in such a cavity is
B0 = 1.2 ± 0.1 mT at an input power of 10 W [11]. For
the measurements shown in this paper, the input signal is
amplified to 16 W.
As the electron beam is swept back and forth by the
cavity, pulses are created twice every oscillation period.
However, as these leave the chopping aperture under dif-
ferent angles [10], half of these must be blocked. This is
currently done with the SA aperture.
Using a Faraday cup, the current of the beam is mea-
sured. The energy spread of the beam is measured us-
ing a Gatan ENFINA spectrometer, with a dispersion of
0.05 eV/ch. Furthermore, measurements will be compared
to particle tracking simulations using the General Particle
Tracer (GPT) code [18], in which realistic fields inside the
cavity are taken into account, including fringe fields.
4. Results
Shown in Fig. 3(a) is the pulsed electron beam focused
on the detector for varying longitudinal positions of the
crossover in the cavity. Moving from left to right, the
current through the C2 lens is increased, raising the focus
position through the point of minimal emittance growth.
At either too low or too high currents spot is elliptical,
with the long and short axes corresponding to the direction
parallel to and perpendicular to the sweeping direction of
the cavity, respectively.
With increasing C2 current, the focused spots in
Fig. 3(a) also rotate. This is due to the change in MC
current to refocus the beam into the detector, which also
rotates the beam.
Figure 3(b) shows the angular distribution of the
pulsed beam. Angles were calibrated using a known
diffraction ring at 10.6 mrad from a typical cross grating
Figure 2: Schematic overview of the adapted microscope column.
Below the C2 aperture the column has been extended by 20.3 cm, in
which a cavity, additional deflectors, and a chopping aperture have
been inserted.
the chopping aperture with a certain width w. Sweeping
this beam with a magnetic field amplitude B0 and an an-
gular frequency ω over an aperture with width s results in
a full width at half maximum (FWHM) pulse length of
τ = γme(s + w)
4qelB0 sin(cid:0)f π
(cid:1) ,
2
(4)
where l is the distance to the chopping aperture, and f =
Lcav/Lmax the fractional length of the cavity Lcav com-
pared to the maximum useful cavity length Lmax = vzπ/ω
for which electrons feel exactly half the oscillation period.
From this equation it can be seen that in order to create
short pulses, the focusing angle has to be small to restrain
w from becoming too large.
Besides deterioration of the brightness, increase of the
energy spread is also an important effect that has to be
considered. Unfortunately, electrons moving through a
cavity will probe the off-axis electric fields of the TM110
mode. This will not only cause the total beam energy to
change, but also the energy spread to increase. Focusing
the beam at the center of the cavity minimizes this increase
in energy spread, but does not completely eliminate it.
It can be shown that this additional energy spread can
be decreased further by using a shorter cavity length with
a higher field amplitude in order to maintain the same
pulse length [12]. The tradeoff is that more power has to
be dissipated by the cavity, which brings along technical
difficulties. For the results shown in this paper, a shorter
cavity length is chosen at the cost of pulse length.
3
C2lensC2aperturedeflectorsTM110cavitychoppingapertureMClensupperObj.lenssample203mmcondensersystemimagingsystem Figure 4: Measured duty cycle as a function of the cavity input
power, fitted with the expected behavior ∝ √P .
figure is the emittance found in simulations. These show
good agreement. Deviations are attributed to the error in
estimating the focus position from the lens current. The
emittance of the continuous beam under the same condi-
tions has been determined to be εn,x = 2.5 ± 0.1 pm rad.
From this, it can be seen that the beam quality is un-
affected by the cavity in the direction perpendicular to the
streak; parallel to the streak the growth of emittance can
be minimized by correct placement of the crossover. At
the minimum, both simulations and measurement give a
negligible increase in emittance.
The minimum RMS spotsize found from these measure-
ments is 0.61 by 0.56 nm, at a focusing angle of 8.74 mrad
and a peak current of 814 ± 2 pA. From Eq. (2), we find
a peak brightness of 6.6 · 106 A/(m2 sr V). As a com-
parison, the RMS spotsize of the continuous beam has
been measured to be 0.55 nm, resulting in a brightness
of 7.5 · 106 A/(m2 sr V). However, the actual brightness
is presumably larger, since the measured spotsize also in-
cludes contributions from abberations.
In order to estimate the field strength in the cavity and
the associated pulse length, the currents of both the con-
tinuous beam and the pulsed beam have been measured
with a Faraday cup at different cavity input powers. Fig-
ure 4 shows the measured current in pulsed mode divided
by the continuous current on the left y-axis as a function
of input power. The right axis shows the corresponding
pulse length acquired in the simulations with the same ra-
∝ √
tio. The solid line shows a fit with the expected behavior
P [16]. At an input power of 15.3 W, a magnetic field
strength of 1.45±0.06 mT is expected from the fit, in good
correspondence with values determined before on similar
cavities [11].
At higher input powers, the measured current is smaller
than expected. This could be due to measuring errors or
instabilities in the electron beam, or by a change in the
quality factor of the cavity as the temperature changes.
Figure 3: (a) Minimal focus size of the pulsed electron beam for vary-
ing focus position within the cavity. C2 current is increased from left
to right. (b) Angular distribution of the pulsed beam, together with
a known diffraction ring at 10.6 mrad used as calibration. (c) En-
ergy spread measured with a spectrometer for both the pulsed and
the continuous beam. (d) Emittance along the long and short axis
for each spot in (a), plotted against the focus position with respect
to the center of the cavity. Curves show the emittance expected from
simulations.
sample [19], which is also shown in Fig. 3(b). From this,
the focusing angle is determined to be 8.74 mrad.
In Fig. 3(c) the energy spread measured with the spec-
trometer for both the pulsed and the continuous beam is
shown. The measurement with the pulsed beam seems to
give a slightly lower energy spread. However, the differ-
ence is well below the resolution of the spectrometer, and
is more likely to be due to small misalignments. These can
easily arise between the two measurements as different set-
tings have to be used for a continuous beam to prevent the
spectrometer from saturating.
Figure 3(d) shows the corresponding emittance plotted
against the difference in focus position. Also shown in this
4
1nm1nm(a)10.6mrad10.6mrad(b)(c)pulsedcont.0.95eV−4−202400.51energy(eV)counts(arb.unit)−1−0.500.5102468focusposition(mm)normalizedemittance(pmrad)paralleltostreakperpendiculartostreak,simulationsIcont.=814±2pA(d)024681012141600.0050.010.015cavityinputpower(W)Ipuled/Icontinuous012345simulatedpulselength(ps) Table 1: Measured parameters of the continuous beam and the pulsed
beam, compared to simulations with the same continuous beam as
input and a magnetic field of 1.45 mT.
continuous
2.5 ± 0.1
2.5 ± 0.1
814 ± 2
0.95 ± 0.05
pulsed
2.7 ± 0.1
2.5 ± 0.1
2.8 ± 0.3
0.95 ± 0.05
εn,k (pm rad)
εn,⊥ (pm rad)
Iavg (pA)
∆Efwhm (eV)
τfwhm (ps)
GPT
2.62
2.51
3.38
1.01
1.31
In either case, a pulse length of 1.33 ± 0.06 ps is deduced
from the fitted curve, whereas a pulse length of 1.1±0.1 ps
is deduced from the current measurement.
Shown in Table 1 are the emittance, current and energy
spread measured for both the continuous and the pulsed
beam. Also shown in this table are simulation results start-
ing with a continuous beam with the same parameters,
and a magnetic field strength of 1.45 mT inside the cav-
ity. Good agreement is found between simulations and
measurements.
5. Conclusions and outlook
To summarize, it has been experimentally verified that
TM110 cavities can be used to create a pulsed electron
beam in a 200 keV TEM without a significant increase in
emittance. For pulse lengths of 1.1 -- 1.3 ps, no measur-
able increase in energy spread or deterioration in perfor-
mance of the microscope is found. This makes an RF-
based UTEM a viable alternative to photocathodes.
As a next step, cavities will be developed further, al-
lowing for synchronization to a clocking laser pulse at a
frequency of 75 MHz. Furthermore, higher input pow-
ers will be tested, and smaller chopping apertures will be
used. With increasing field strength in the cavity care
must be taken to prevent an increase in energy spread.
Reference [12] explains in more detail how this can be
achieved. In this way, the pulse length can be reduced to-
wards 100 fs, allowing for pump -- probe experiments with
both a high temporal resolution and a high transverse co-
herence.
Acknowledgement
This work is part of an Industrial Partnership Pro-
gramme of the Foundation for Fundamental Research on
Matter (FOM), which is part of the Netherlands Organi-
sation for Scientific Research (NWO). The authors would
like to thank E.H. Rietman, I. Koole, H.A. van Doorn, and
A.H. Kemper for their invaluable technical support.
References
[1] G. Sciaini, R. Miller, Femtosecond electron diffraction: herald-
ing the era of atomically resolved dynamics, Rep. Prog. Phys.
74 (2011) 096101. doi:10.1088/0034-4885/74/9/096101.
5
[2] D. Flannigan, A. Zewail, 4D electron microscopy: principles
and applications, Acc. Chem. Res. 45 (2012) 1828 -- 1839. doi:
10.1021/ar3001684.
[3] R. van der Veen, T. Penfold, A. Zewail, Ultrafast core-loss spec-
troscopy in four-dimensional electron microscopy, Struct. Dyn.
2 (2015) 024302. doi:10.1063/1.4916897.
[4] D. Plemmons, D. Flannigan, Ultrafast electron microscopy: In-
strument response from the single-electron to high bunch-charge
regimes, Chem. Phys. Lett. 683 (2017) 186. doi:10.1016/j.
cplett.2017.01.055.
[5] A. Feist, N. Bach, N. da Silva, T. Danz, M. Moller, K. Priebe,
T. Domrose, J. Gatzmann, S. Rost, J. Schauss, S. Strauch,
R. Bormann, M. Sivis, S. Schafer, C. Ropers, Ultrafast trans-
mission electron micrroscopy using a laser-driven field emitter:
femtosecond resolution with a high coherence electron beam, Ul-
tramicroscopy 175 (2017) 63. doi:10.1016/j.ultramic.2016.
12.005.
[6] L. Oldfield, A rotationally symmetric electron beam chopper for
picosecond pulses, J. Phys. E: Sci. Instrum. 9 (1976) 6. doi:
10.1088/0022-3735/9/6/011.
[7] T. Hosokawa, H. Fujioka, K. Ura, Generation and measurement
of subpicosecond electron beam pulses, Rev. Sci. Instrum. 49
(1978) 624. doi:10.1063/1.1135464.
[8] I. Weppelman, R. Moerland, J. Hoogenbeem, P. Kruit, Concept
and design of a beam blanker with integrated photoconductive
switch for ultrafast electron microscopy, Ultramicroscopy 184
(2018) 8. doi:10.1016/j.ultramic.2017.10.002.
[9] H. Zhang, J. Tang, J. Yan, Y. Yamauchi, T. Suzuki, N. Shinya,
K.ÃŚakajima, L. Qin, An ultrabright and monochromatic elec-
tron point source made of a LaB6 nanowire, Nat. Nanotechnol.
11 (2016) 273. doi:10.1038/nnano.2015.276.
[10] A. Lassise, Miniaturized RF technology for femtosecond elec-
tron microscopy, Ph.D. thesis, Eindhoven University of Tech-
nology (2012).
[11] W. Verhoeven, J. van Rens, M. van Ninhuijs, W. Toonen,
E. Kieft, P. Mutsaers, O. Luiten, Time-of-flight electron energy
loss spectroscopy using TM110 deflection cavities, Struct. Dyn.
3 (2016) 054303. doi:10.1063/1.4962698.
[12] J. van Rens, W. Verhoeven, J. Franssen, A. Lassise, X. Stragier,
E. Kieft, P. Mutsaers, O. Luiten, Theory and particle track-
ing simulations of a resonant radiofrequency deflection cavity in
TM110 mode for ultrafast electron microscopy, Ultramicroscopy
184 (2018) 77. doi:10.1016/j.ultramic.2017.10.004.
[13] G. Brussaard, A. Lassise, P. Pasmans, P. Mutsaers, M. van der
Wiel, O. Luiten, Direct measurement of synchronization be-
tween femtosecond laser pulses and a 3 GHz radio frequency
electric field inside a resonant cavity, Appl. Phys. Lett. 103
(2013) 141105. doi:10.1063/1.4823590.
[14] M. Walbran, A. Gliserin, K. Jung, J. Kim, P. Baum, 5-
femtosecond laser-electron synchronization for pump-probe
crystallography and diffraction, Phys. Rev. Appl. 4 (2015)
044013. doi:10.1103/PhysRevApplied.4.044013.
[15] J. Qui, G. Ha, C. Jing, S. Baryshev, B. Reed, J. Lau, Y. Zhu,
GHz laser-free time-resolved transmission electron microscopy:
A stroboscopic high-duty-cycle method, Ultramicroscopy 161
(2015) 130 -- 136. doi:10.1016/j.ultramic.2015.11.006.
[16] A. Lassise, P. Mutsaers, O. Luiten, Compact,
low power
radio frequency cavity for femtosecond electron micsoscopy,
Rev. Sci. Instrum. 83 (2012) 043705. doi:10.1063/1.3703314.
[17] P. Hawkes, E. Kasper, Principles of Electron Optics II: Applied
Geometrical Optics, Academic Press, 1989.
[18] See http://www.pulsar.nl/gpt for more information on the
software.
[19] Agar Scientific, Cross Grating S106. For more information, see
http://www.agarscientific.com.
|
1906.00459 | 1 | 1906 | 2019-06-02T17:56:08 | Low-loss composite photonic platform based on 2D semiconductor monolayers | [
"physics.app-ph",
"physics.optics"
] | Two dimensional materials such as graphene and transition metal dichalcogenides (TMDs) are promising for optical modulation, detection, and light emission since their material properties can be tuned on-demand via electrostatic doping. The optical properties of TMDs have been shown to change drastically with doping in the wavelength range near the excitonic resonances. However, little is known about the effect of doping on the optical properties of TMDs away from these resonances, where the material is transparent and therefore could be leveraged in photonic circuits. Here, we probe the electro-optic response of monolayer TMDs at near infrared (NIR) wavelengths (i.e. deep in the transparency regime), by integrating them on silicon nitride (SiN) photonic structures to induce strong light$-$matter interaction with the monolayer. We dope the monolayer to carrier densities of ($7.2 \pm 0.8$) $\times$ $10^{13} \textrm{cm}^{-2}$, by electrically gating the TMD using an ionic liquid. We show strong electro-refractive response in monolayer tungsten disulphide (WS$_2$) at NIR wavelengths by measuring a large change in the real part of refractive index $\Delta$n = $0.53$, with only a minimal change in the imaginary part $\Delta$k = $0.004$. The doping induced phase change ($\Delta$n), compared to the induced absorption ($\Delta$k) measured for WS$_2$ ($\Delta$n/$\Delta$k $\sim 125$), a key metric for photonics, is an order of magnitude higher than the $\Delta$n/$\Delta$k for bulk materials like silicon ($\Delta$n/$\Delta$k $\sim 10$), making it ideal for various photonic applications. We further utilize this strong tunable effect to demonstrate an electrostatically gated SiN-WS$_2$ phase modulator using a WS$_2$-HfO$_2$ (Hafnia)-ITO (Indium Tin Oxide) capacitive configuration, that achieves a phase modulation efficiency (V$_\pi$L) of 0.8 V $\cdot$ cm with a RC limited bandwidth of 0.3 GHz. | physics.app-ph | physics | Low-loss composite photonic platform based on 2D semiconductor
monolayers
Ipshita Datta1,*, Sang Hoon Chae2,*, Gaurang R. Bhatt1, Mohammad A. Tadayon1, Baichang Li2, Yiling
Yu3, Chibeom Park4 , Jiwoong Park4 , Linyou Cao3, D. N. Basov5, James Hone2 and Michal Lipson1
1Department of Electrical Engineering, Columbia University, New York, New York 10027, USA
2Department of Mechanical Engineering, Columbia University, New York, New York 10027, USA
3Department of Materials Science and Engineering and Department of Physics, North Carolina State University, Raleigh, North
Carolina 27695, USA
4Department of Chemistry, Institute for Molecular Engineering, James Franck Institute, University of Chicago, Chicago, IL 60637, USA
5Department of Physics, Columbia University, New York, New York 10027, USA
Corresponding Author -- [email protected]
*These authors contributed equally to this work.
Two dimensional (2D) materials such as graphene and transition metal dichalcogenides (TMDs)
are promising for optical modulation, detection, and light emission since their material properties
can be tuned on-demand via electrostatic doping1 -- 18. The optical properties of TMDs have been
shown to change drastically with doping in the wavelength range near the excitonic resonances19 --
22. However, little is known about the effect of doping on the optical properties of TMDs away
from these resonances, where the material is transparent and therefore could be leveraged in
photonic circuits. Here, we probe the electro-optic response of monolayer TMDs at near infrared
(NIR) wavelengths (i.e. deep in the transparency regime), by integrating them on silicon nitride
(SiN) photonic structures to induce strong light -matter interaction with the monolayer. We dope
the monolayer to carrier densities of (7.2 ± 0.8) × 1013 cm-2, by electrically gating the TMD using
an ionic liquid [P14+] [FAP-]. We show strong electro-refractive response in monolayer tungsten
disulphide (WS2) at NIR wavelengths by measuring a large change in the real part of refractive
index ∆n = 0.53, with only a minimal change in the imaginary part ∆k = 0.004. The doping induced
phase change (∆n), compared to the induced absorption (∆k) measured for WS2 (∆n/∆k ∼ 125), a
key metric for photonics, is an order of magnitude higher than the ∆n/∆k for bulk materials like
silicon (∆n/∆k ∼ 10)23, making it ideal for various photonic applications24 -- 28. We further utilize
this strong tunable effect to demonstrate an electrostatically gated SiN-WS2 phase modulator using
a WS2-HfO2 (Hafnia)-ITO (Indium Tin Oxide) capacitive configuration, that achieves a phase
modulation efficiency (VπL) of 0.8 V · cm with a RC limited bandwidth of 0.3 GHz.
In order to probe the doping induced electro-optic response of TMDs, we utilize a SiN-TMD
composite waveguide platform in which the optical mode of the composite waveguide is shared
between the TMD monolayer and the dielectric SiN waveguide. These waveguides are then
incorporated into photonic structures such as microring resonators which are sensitive to small
changes in absorption and phase induced by electrostatic doping of the TMD monolayer. The
SiO2-clad SiN waveguides10 are fabricated using standard techniques, and then planarized to
permit mechanical transfer of TMD layers and subsequent lithographic patterning to define Ti/Au
contacts for gating (see Methods). We probe the response of these structures in the NIR region (
~ 1550 nm), far below the excitonic resonances of the TMDs (~ 620 nm for WS2; 660 nm for
MoS2).
We first characterize the electro-optic response of monolayer WS2 by measuring the effect of
gating on the response of a microring resonator cavity (1.3 µm × 330 nm passive SiN waveguide
clad with 100 nm of SiO2, embedded in a ring of 60 µm radius), as depicted in figure 1a. The high-
Q (∼120,000) cavity is critically coupled to a SiN bus waveguide, such that the transmission
spectrum is extremely sensitive to small changes in phase and absorption within the cavity. A 30
µm arc length of WS2 grown by metal-organic chemical vapor deposition29
(MOCVD) is patterned
onto the ring and electrically contacted (see Methods). We tune the effective index of the optical
mode by electrostatically doping the monolayer using an ionic liquid [P14+] [FAP-], chosen due to
its stability to gating under atmospheric exposure at room temperature30. As shown in figure 1b,
we apply a potential difference between the electrode connected to WS2 and a nearby counter-
electrode, which results in the accumulation of ions at the surface of WS2. We note that ionic liquid
gated devices are dominated by the quantum capacitance (Cq)30, such that no carriers are injected
to the WS2 within the bandgap region (roughly -1 to 1 V). The maximum electron doping density
induced in the monolayer is about 7.2 ± 0.8 x 1013 cm-2 with an applied voltage of 2 V (see
Supplementary Section I). One can see from the cross-section of the platform in figure 1c, that the
optical mode is shared between the monolayer and SiN waveguide. We estimate an optical mode
overlap of 0.06 % using COMSOL Multiphysics simulations (see Methods).
Figure 1 Schematic of the ionic liquid gated SiN-WS2 platform. (a) Schematic representation of a microring
resonator with patterned monolayer WS2 and Ti/Au electrodes, in which only one of the electrodes is in contact with
the monolayer WS2. (b) Illustration of the composite SiN-WS2 waveguide with ionic liquid ([P14+] [FAP-]) cladding.
The monolayer WS2 is doped by applying a bias voltage across the two electrodes through the ionic liquid, resulting
in the accumulation of charged carriers at the interface of WS2, thereby creating image charges in the monolayer. (c)
TE mode profile of the propagating mode in the SiN waveguide with 0.06 % overlap with monolayer WS2. (d)
Normalized transmission response of the microring resonator for different voltages applied across the two electrodes.
One can see that the resonance of the microring red shifts with significantly smaller broadening of the resonance with
increasing voltage, indicating that the propagating mode undergoes strong phase shift but with minimal absorption.
Bottom right inset is the optical micrograph of the fabricated ring modulator with SiN-WS2 waveguides, before ionic
liquid is dispersed on the devices. (e) Change in real and imaginary part of the effective index (∆𝑛eff and ∆𝑘eff) of the
composite SiN-WS2 waveguide at different voltages, extracted from the normalized transmission spectra in figure 1d.
The error bars shown in figure 1e account for the root mean square (rms) error from the numerical fit and a ± 5 µm
error in the patterned length of the monolayer. (f) Change in the real and imaginary part of the refractive index of
monolayer WS2 (∆𝑛WS2 and ∆𝑘WS2) with induced carrier densities, extracted from the change in effective index in
figure 1e. The vertical error bars in figure 1f incorporates the rms error in the effective index measurements from
figure 1e, variation in the thickness of monolayer (± 0.05 nm), variation in the height of the cladding oxide by about
± 25 nm and about ± 0.02 (RIU) variation in the index of the ionic liquid included during the COMSOL simulations.
The horizontal error bars accounts for the error in the intrinsic doping of the monolayer WS2.
Figure 1d shows the measured transmission spectrum of the ring resonator as a function of voltage.
The cavity has a resonance near 1571 nm, with narrow linewidth confirming that incorporation of
WS2 introduces negligible loss and does not degrade the quality factor of the microring resonator.
At positive gate voltages above 1 V, we observe a red shift in the resonance wavelength, indicating
an increase in the effective index of the resonator. Remarkably, the resonance linewidth is largely
unchanged, thereby showing that doping does not introduce substantial loss. The measured
changes in resonance wavelength and quality factor with applied voltage can be used to derive the
changes in real and imaginary components of the effective index (∆𝑛eff and ∆𝑘eff, respectively) of
the composite waveguide (see Supplementary Section II). Figure 1e shows the measured ∆𝑛eff and
∆𝑘eff as a function of gate voltage. ∆𝑛eff increases linearly with gate voltage with the onset of n-
type doping (above 1V). In this same range, ∆𝑘eff is almost two orders of magnitude smaller. We
can then determine the underlying change in real and imaginary part of the refractive index of
WS2, ∆𝑛WS2 and ∆𝑘WS2 respectively by COMSOL modelling of the monolayer as a 2D sheet with
optical conductivity 𝜎S integrated on a SiN waveguide (see Methods). Figure 1f shows ∆𝑛WS2 and
∆𝑘WS2 as a function of gate-induced carrier density in the monolayer WS2. ∆𝑛WS2 reaches 0.53 ±
0.06 RIU (refractive index units) for maximum doping of (7.2 ± 0.8) × 1013 cm-2, while ∆𝑘WS2 is
0.004 ± 0.002. This indicates that monolayer WS2 has a unique combination of strong electro-
refractive response (∆𝑛WS2) and small electro-absorptive response (∆𝑘WS2) at NIR wavelengths,
i.e. the propagating light undergoes significant phase change with low optical loss.
Figure 2 Tuning the effective index (∆𝒏𝐞𝐟𝐟) of the propagating TE and TM mode using monolayer WS2. (a)
Diagrammatic illustration of an on-chip Mach Zehnder interferometer (MZI) with patterned monolayer WS2 on both
the arms of MZI. (b) Optical micrograph of the fabricated MZI with SiN-WS2 waveguides, before ionic liquid is
dispersed on the devices. (c) Normalized Transmission response of the MZI for the optical TE mode at different
voltages applied across the two electrodes located on the longer arm of the MZI. One can see that the MZI transmission
spectra exhibit fringes due to the path length difference between the two arms of the MZI, which red shifts with applied
voltage. The top figure shows the optical mode profile for the TE mode, with the arrows illustrating the field lines
aligned with the plane of monolayer WS2. (d) Normalized Transmission response of the MZI for the optical TM mode
at different voltages. The bottom right inset shows the optical mode profile for the TM mode with the optical field
lines perpendicular to the plane of monolayer WS2.
We show that the strong phase change occurs only when the polarization of the propagating light
is in plane with the monolayer TMD (i.e. TE mode). In order to probe the polarization effect, we
embed the SiN-WS2 composite waveguide in the arms of a Mach Zehnder interferometer (MZI),
optimized to operate both for the TM and TE mode in the S wavelength band (1460 -- 1490 nm)
and C/L band (1530 - 1630 nm), respectively (see Methods). Figure 2a illustrates the diagrammatic
representation of the MZI device while figure 2b shows an optical micrograph of the fabricated
device. The MZI is designed with a length imbalance of 200 µm between the two arms such that
the transmission spectra exhibit fringes with a visibility of 0.98 in the wavelength range spanning
from 1560 - 1620 nm for TE mode and 1455 - 1480 nm for TM mode. A 500 µm long film of
monolayer WS2 is patterned on both the arms of the MZI, followed by the deposition of metal
electrodes. We dope the monolayer by applying a voltage between the metal electrodes through
the ionic liquid [P14+] [FAP-]. We estimate from COMSOL Multiphysics simulation that the
optical mode overlap with the monolayer is 0.06% and 0.04% for the TE and TM mode,
respectively (see Methods). In Figure 2c and 2d, we show the interference pattern at the MZI
output for the TE and TM mode, respectively with different voltages applied to dope the monolayer
WS2 on the longer arm of the MZI. There is a pronounced wavelength shift in the MZI spectra for
the TE mode (Figure 2c), as compared to minimal wavelength shift for the TM mode (Figure 2d),
indicating that the doping-induced electro-refractive effect is strong only when the optical E field
is aligned with the monolayer, despite similar mode overlap for the TE/TM mode with the
monolayer. We further confirm that the doping of ionic liquid in the absence of the monolayer
does not influence the MZI transmission response (see Supplementary Section III).
To leverage the doping-induced strong electro-refractive response in monolayer TMDs for
photonic applications, we develop a fully integrated SiN-TMD platform that gates the monolayer
TMD using a parallel plate capacitor configuration. We replace the ionic liquid used above with a
stack of HfO2 (hafnia) and transparent conducting oxide ITO (indium tin oxide) to form the TMD-
HfO2-ITO capacitor on the SiN waveguide, as illustrated in figure 3a. Standard processes are used
to define the optimized 1 μm × 360 nm LPCVD SiN waveguide, clad with 260 nm of SiO2 and a
500 μm long WS2-HfO2-ITO capacitor fabricated on each arm of the MZI (see Methods). We
replace the MOCVD-grown29 monolayer WS2 used in the ionic liquid experiments with CVD-
grown21 films for the capacitive devices (see Supplementary Section IV). Supplementary figure
S4a shows the false-coloured optical micrograph of the fabricated SiN-TMD MZI. In this
configuration quantum capacitance effects are negligible,30 and we estimate a linear induced
charge density of 0.37 ± 0.05 × 1012 cm-2 per volt, based on the thickness (26 nm) and dielectric
permittivity of the HfO2. In contrast to the ionic liquid devices where the voltage swing is limited
between -2 V to 2 V, the voltage swing in these capacitive devices is determined by the thickness
of the dielectric (HfO2) and is in range of {-8V, 9V}.
Figure 3b shows the interference pattern at the MZI output for different voltages applied across
the WS2-HfO2-ITO capacitor on the longer arm of the MZI. As in the ionic liquid gated devices,
we induce a strong change in phase with applied voltage, thereby designing a phase modulator
with a modulation efficiency (VπL) of 1.33 V ∙ cm, coupled with minimal change in extinction.
Figure 3c shows the gating induced ∆𝑛eff of the composite waveguide, extracted by measuring the
wavelength shift of the MZI spectra (see Supplementary Section IV), which reaches 7×10-4 RIU
for a swing voltage of 8V. From the extremum point at -4 V in Figure 3c, we infer that the charge
neutrality point for the monolayer WS2 layer is at -4 V, which corresponds to an initial electron
doping of 1.5 ± 0.2 × 1012 cm-2. This initial doping is likely due to sulfur vacancies arising from
CVD growth but can also arise from substrate effects or adsorbates introduced in the transfer
process31.
We further enhance the performance of the capacitive SiN-WS2 platform by increasing the optical
mode overlap with monolayer WS2 using SU-8 photoresist as a high index cladding (see
Supplementary figure S4b). In figure 3c, we show an increase in the gating induced ∆𝑛eff from 7.1
× 10-4 RIU to 1.35 × 10-3 RIU which reduces the VπL from 1.33 V · cm to 0.8 V · cm (see
Supplementary Section V). We estimate that the mode overlap for the unclad device is 0.016 %
and for the SU-8 clad device is 0.03 %, based on our simulations (see Methods). The ∆𝑛eff for the
SU-8 clad device is almost double the ∆𝑛eff extracted for an unclad device, in agreement with the
improvement in the mode overlap from our COMSOL computations. Considering the relatively
small mode overlap with the monolayer CVD WS2 in this present work (0.03 %), one could
envision an even higher phase modulation efficiency using alternative waveguide geometries with
different degrees of mode confinement such as slot waveguides and photonic crystals. We measure
a 3 dB bandwidth of 0.3 GHz in our devices with minimal DC electrical power dissipation of 0.64
nW in the WS2-HfO2-ITO capacitor (see Supplementary Section VI and VII). The frequency
response of the phase modulator is measured using a 40 GHz fast photodiode and an electrical
vector network analyzer (VNA) at 1550 nm. The bandwidth is currently limited by the resistance
of monolayer WS2, and can be increased with improved electrical contacts, optimized geometry,
and control over the initial doping density.
Finally, we demonstrate that the gating induced electro-refractive effect extends to other
semiconductor TMDs by fabricating and characterizing a SiN-MoS2 composite waveguide
embedded in a MZI structure (see figure 4a). The measured ∆𝑛eff as a function of gate voltage is
shown in Figure 4b, and reaches a maximum of 6.4 × 10-4 (RIU) at 8 V (see Supplementary
Section VIII). The phase modulation efficiency VπL is 1.7 V • cm. Figure 4c compares the index
change (∆𝑛) for MoS2 and WS2 with gate voltage. In both cases the index changes strongly, with
the change in WS2 roughly double that MoS2.
The demonstrated strong light matter interaction in monolayer TMDs could open up doors for a
range of novel applications with these 2D materials and enable highly reconfigurable photonic
circuits with low optical loss and power dissipation. The 2D configuration in the composite
platform lends itself to extremely high doping of the material. The induced phase change, relative
to the induced absorption (i.e ∆n/∆k -- a key metric for photonics) is about 125 at carrier doping
densities of 7 × 1013 cm-2, about two orders of magnitude higher than in traditional bulk materials
such as silicon. The low absorption in TMDs at such high doping densities is due to the limited
phase space in a lower dimensional system. Traditional phase modulators based on either thermo-
optic or induced electro-optic χ(2) effects have similar low optical losses as our SiN-TMD device,
but suffer from either high electrical power consumption and low operation bandwidth32 -- 35 or
require a large device footprint and requires complex fabrication techniques36 -- 38, respectively. For
large scale photonic systems, wafer scale integration of TMD materials with silicon photonics can
be done either as a direct TMD growth process on silicon wafers29,39,40, or as a post processing step
where large wafer scale TMD films41 are transferred on a silicon photonics platform, fabricated in
a standard foundry. The low electrical power dissipation and moderate operation bandwidth in our
SiN-TMD platform allows for the large scale integration of these phase modulators in numerous
applications such as LIDAR, phased arrays, optical switching, quantum and optical neural
networks.
Figure 3 Phase tuning of monolayer TMD using TMD-HfO2-ITO capacitor. (a) Illustration of a composite SiN-
monolayer TMD waveguide, where the optical mode is shared between the SiN waveguide and TMD (bottom right).
The top right inset details the TMD-HfO2-ITO parallel plate capacitor configuration which electrostatically dopes the
monolayer WS2, by applying a bias voltage between the two electrodes (TMD and ITO) across the dielectric HfO2.
(b) Normalized Transmission response of the MZI for different voltages applied across the WS2-HfO2-ITO capacitor.
(c) Extracted change in effective index (∆𝑛eff) measured from the MZI spectra of the composite SiN-WS2 waveguide
with and without SU-8 cladding. The unclad device in figure 3a has a mode overlap of 0.016 % with the monolayer
WS2, while the SU-8 clad device has a mode overlap of 0.03 %. One can see that the enhanced mode overlap increases
the maximum ∆𝑛eff of the composite SiN-WS2 waveguide from 7 × 10-4 to 1.3 × 10-3 (corresponding to a VπL of 0.8
V ∙ cm). The error bars show the variation in gating induced ∆neff, extracted from multiple MZI devices with different
lengths of WS2-HfO2-ITO capacitor. (d) Frequency response (S21) of the unclad SiN-WS2 phase modulator measured
at 1550 nm.
Figure 4 Phase tuning of monolayer MoS2 in a composite SiN-MoS2 waveguide using TMD-HfO2-ITO
capacitor. (a) Illustration of a composite SiN-MoS2 waveguide, where the optical mode is shared between the
waveguide and MoS2. Figure right inset illustrates the optical mode profile for the composite SiN-MoS2 waveguide,
with an optical mode overlap of 0.054% with the monolayer MoS2. (b) Change in the effective index (∆𝑛eff) of the
propagating mode with voltage, extracted from the normalized MZI transmission spectra of the composite SiN-MoS2
waveguide (see Supplementary Section VIII). (c) Extracted change in refractive index of the monolayer TMD's
(∆𝑛TMD) such as WS2 and MoS2 with a voltage swing of 8 V across the TMD-HfO2-ITO capacitor. One can see that
∆𝑛WS2 > ∆𝑛MoS2 even though the ∆𝑛eff for the unclad SiN-WS2 composite waveguide (Figure 3c) is similar to ∆𝑛eff
of the unclad SiN-MoS2 waveguide (Figure 4b), due to the higher mode overlap with the monolayer in the SiN-MoS2
waveguide (0.054%) compared to that of the SiN-WS2 waveguide (0.016%). The error bars in the ∆𝑛 calculations
include the rms error in the effective index measurements from Figure 3c and 4b, variation in the thickness of
monolayer (± 0.05 nm), and variation in the height of the cladding oxide by about ± 25 nm included during the
COMSOL simulations
References
1. Bonaccorso, F., Sun, Z., Hasan, T. & Ferrari, A. C. Graphene photonics and optoelectronics.
Nat. Photonics 4, 611 -- 622 (2010).
2. Wang, Q. H., Kalantar-Zadeh, K., Kis, A., Coleman, J. N. & Strano, M. S. Electronics and
optoelectronics of two-dimensional transition metal dichalcogenides. Nat. Nanotechnol. 7,
699 -- 712 (2012).
3. Santos, E. J. G. & Kaxiras, E. Electrically Driven Tuning of the Dielectric Constant in MoS2
Layers. ACS Nano 7, 10741 -- 10746 (2013).
4. Xia, F., Wang, H., Xiao, D., Dubey, M. & Ramasubramaniam, A. Two-dimensional material
nanophotonics. Nat. Photonics 8, 899 -- 907 (2014).
5. Koppens, F. H. L. et al. Photodetectors based on graphene, other two-dimensional materials
and hybrid systems. Nat. Nanotechnol. 9, 780 -- 793 (2014).
6. Ross, J. S. et al. Electrically tunable excitonic light-emitting diodes based on monolayer WSe2
p -- n junctions. Nat. Nanotechnol. 9, 268 -- 272 (2014).
7. Jariwala, D., Sangwan, V. K., Lauhon, L. J., Marks, T. J. & Hersam, M. C. Emerging Device
Applications for Semiconducting Two-Dimensional Transition Metal Dichalcogenides. ACS
Nano 8, 1102 -- 1120 (2014).
8. Schwarz, S. et al. Two-Dimensional Metal -- Chalcogenide Films in Tunable Optical
Microcavities. Nano Lett. 14, 7003 -- 7008 (2014).
9. C. Ferrari, A. et al. Science and technology roadmap for graphene, related two-dimensional
crystals, and hybrid systems. Nanoscale 7, 4598 -- 4810 (2015).
10. Phare, C. T., Lee, Y.-H. D., Cardenas, J. & Lipson, M. Graphene electro-optic modulator with
30 GHz bandwidth. Nat. Photonics 9, 511 -- 514 (2015).
11. Pospischil, A. & Mueller, T. Optoelectronic Devices Based on Atomically Thin Transition
Metal Dichalcogenides. Appl. Sci. 6, 78 (2016).
12. Mak, K. F. & Shan, J. Photonics and optoelectronics of 2D semiconductor transition metal
dichalcogenides. Nat. Photonics 10, 216 -- 226 (2016).
13. Xiao, J., Zhao, M., Wang, Y. & Zhang, X. Excitons in atomically thin 2D semiconductors and
their applications. Nanophotonics 6, 1309 -- 1328 (2017).
14. Manzeli, S., Ovchinnikov, D., Pasquier, D., Yazyev, O. V. & Kis, A. 2D transition metal
dichalcogenides. Nat. Rev. Mater. 2, 17033 (2017).
15. Bernardi, M., Ataca, C., Palummo, M. & Grossman, J. C. Optical and Electronic Properties of
Two-Dimensional Layered Materials. Nanophotonics 6, 479 -- 493 (2016).
16. Bie, Y.-Q. et al. A MoTe2-based light-emitting diode and photodetector for silicon photonic
integrated circuits. Nat. Nanotechnol. 12, 1124 -- 1129 (2017).
17. Basov, D. N., Averitt, R. D. & Hsieh, D. Towards properties on demand in quantum materials.
Nat. Mater. 16, 1077 -- 1088 (2017).
18. Romagnoli, M. et al. Graphene-based integrated photonics for next-generation datacom and
telecom. Nat. Rev. Mater. 3, 392 (2018).
19. Ross, J. S. et al. Electrical control of neutral and charged excitons in a monolayer
semiconductor. Nat. Commun. 4, 1474 (2013).
20. Chernikov, A. et al. Electrical Tuning of Exciton Binding Energies in Monolayer WS2. Phys.
Rev. Lett. 115, 126802 (2015).
21. Yu, Y. et al. Giant Gating Tunability of Optical Refractive Index in Transition Metal
Dichalcogenide Monolayers. Nano Lett. 17, 3613 -- 3618 (2017).
22. Wang, G. et al. Colloquium: Excitons in atomically thin transition metal dichalcogenides. Rev.
Mod. Phys. 90, 21001 (2018).
23. Soref, R. & Bennett, B. Electrooptical effects in silicon. IEEE J. Quantum Electron. 23, 123 --
129 (1987).
24. Doylend, J. K. et al. Two-dimensional free-space beam steering with an optical phased array
on silicon-on-insulator. Opt. Express 19, 21595 -- 21604 (2011).
25. Sun, J., Timurdogan, E., Yaacobi, A., Hosseini, E. S. & Watts, M. R. Large-scale nanophotonic
phased array. Nature 493, 195 -- 199 (2013).
26. Harris, N. C. et al. Quantum transport simulations in a programmable nanophotonic processor.
Nat. Photonics 11, 447 -- 452 (2017).
27. Shen, Y. et al. Deep learning with coherent nanophotonic circuits. Nat. Photonics 11, 441 --
446 (2017).
28. Miller, S. A. et al. 512-Element Actively Steered Silicon Phased Array for Low-Power
LIDAR. in Conference on Lasers and Electro-Optics (2018), paper JTh5C.2 JTh5C.2 (Optical
Society of America, 2018). doi:10.1364/CLEO_AT.2018.JTh5C.2
29. Kang, K. et al. High-mobility three-atom-thick semiconducting films with wafer-scale
homogeneity. Nature 520, 656 -- 660 (2015).
30. Braga, D., Gutiérrez Lezama, I., Berger, H. & Morpurgo, A. F. Quantitative Determination of
the Band Gap of WS2 with Ambipolar Ionic Liquid-Gated Transistors. Nano Lett. 12, 5218 --
5223 (2012).
31. Mlack, J. T. et al. Transfer of monolayer TMD WS2 and Raman study of substrate effects. Sci.
Rep. 7, 43037 (2017).
32. Harris, N. C. et al. Efficient, compact and low loss thermo-optic phase shifter in silicon. Opt.
Express 22, 10487 -- 10493 (2014).
33. Masood, A. et al. Fabrication and characterization of CMOS-compatible integrated tungsten
heaters for thermo-optic tuning in silicon photonics devices. Opt. Mater. Express 4, 1383 --
1388 (2014).
34. Chang, Y.-C., Roberts, S. P., Stern, B. & Lipson, M. Resonance-Free Light Recycling. (2017).
35. Muñoz, P. et al. Silicon Nitride Photonic Integration Platforms for Visible, Near-Infrared and
Mid-Infrared Applications. Sensors 17, (2017).
36. Timurdogan, E., Poulton, C. V., Byrd, M. J. & Watts, M. R. Electric field-induced second-
order nonlinear optical effects in silicon waveguides. Nat. Photonics 11, 200 -- 206 (2017).
37. Wang, C., Zhang, M., Stern, B., Lipson, M. & Lončar, M. Nanophotonic lithium niobate
electro-optic modulators. Opt. Express 26, 1547 -- 1555 (2018).
38. Kieninger, C. et al. Ultra-high electro-optic activity demonstrated in a silicon-organic hybrid
modulator. Optica 5, 739 -- 748 (2018).
39. Gao, Y. et al. Large-area synthesis of high-quality and uniform monolayer WS2 on reusable
Au foils. Nat. Commun. 6, 8569 (2015).
40. McCreary, K. M., Hanbicki, A. T., Jernigan, G. G., Culbertson, J. C. & Jonker, B. T. Synthesis
of Large-Area WS2 monolayers with Exceptional Photoluminescence. Sci. Rep. 6, 19159
(2016).
41. Lee, J. et al. Crack-Release Transfer Method of Wafer-Scale Grown Graphene Onto Large-
Area Substrates. ACS Appl. Mater. Interfaces 6, 12588 -- 12593 (2014).
42. Emani, N. K. et al. Electrical Modulation of Fano Resonance in Plasmonic Nanostructures
Using Graphene. Nano Lett. 14, 78 -- 82 (2014).
43. Li, Y. et al. Measurement of the optical dielectric function of monolayer transition-metal
dichalcogenides: MoS2, MoSe2, WS2, and WSe2. Phys. Rev. B 90, 205422 (2014).
Acknowledgements: The authors thank Dr. Aseema Mohanty, Dr. Christopher T. Phare, Dr.
Xingchen Ji, Dr. Steven A. Miller and Dr. Utsav Deepak Dave for fruitful discussions. The
authors gratefully acknowledge support from Office of Naval Research (ONR) for award
#N00014-16-1-2219, from Defense Advanced Research Projects Agency (DARPA) for award
#HR001110720034 and #FA8650-16-7643. Research on tunable optical phenomena in TMD
semiconductors was supported as part of Energy Frontier Research Center on Programmable
Quantum Materials funded by the U.S. Department of Energy (DOE), Office of Science, Basic
Energy Sciences (BES), under award #DE-SC0019443. C. P. and J.P. acknowledge funding from
AFOSR (#FA9550-16-1-0031, #FA9550-16-1-0347). Y.Y. and L.C. acknowledges support from
an EFRI award from NSF (#EFMA 1741693). S.H.C. was supported by the Postdoctoral
Research Program of Sungkyunkwan University (2016). This work was done in part at the City
University of New York Advanced Science Research Center NanoFabrication Facility, in part
at the Cornell Nanoscale Facility, supported by the NSF award #EECS-1542081 and in part at
the Columbia Nano Initiative (CNI) shared labs at Columbia University in the City of New
York.
Author contributions: I.D. and M.L. conceived and proposed the TMD based photonic design
and experiments. S.H.C, J.H. and D.N.B proposed the use of WS2 for these photonic structures.
C.P. and J.P. provided the MOCVD WS2 film for the ionic liquid experiments, Y.Y. and L.C.
provided the CVD WS2 and MoS2 film for the capacitive device experiments, and S.H.C and B.L.
performed the TMD transferring and characterization (PL measurements). I.D. fabricated the
composite photonic device, with assistance from S.H.C for TMD processing and development,
G.R.B for the ITO development and M.A.T for the SU-8 cladding of the composite structures. I.D.
performed and analyzed the optical measurements of the photonic devices. I.D. and G.R.B
performed the capacitance measurement of these structures. I.D. and M.L. prepared the
manuscript. S.H.C, D.N.B, J.H. and M.L. edited the manuscript. J.H. and M.L. supervised the
project.
Methods
Device Fabrication.
TMD preparation, transfer and patterning- The PMMA/TMD stack was delaminated from the
SiO2/Si substrate by floating in a hot 1 M KOH solution with the PMMA side up. The
PMMA/TMD stack is then rinsed in water a couple of times, before transferring it onto the SiN
waveguides. After the transfer, the TMD clad waveguides are left to dry for a few hours before the
PMMA was removed by soaking in acetone and rinsing in isopropanol. The transferred TMD is
then patterned with HSQ/PMMA stack using electron beam lithography (EBL), followed by
oxygen plasma and CHF3/O2 for etching residual PMMA and TMD. After the transfer, the TMD
clad waveguides are left to dry for a few hours before the PMMA was removed by soaking in
acetone and rinsing in isopropanol. The transferred TMD is then patterned with HSQ/PMMA stack
using electron beam lithography (EBL), followed by oxygen plasma and CHF3/O2 for etching
residual PMMA and TMD.
Device Fabrication for SiN-TMD Ionic Liquid Photonic Devices. We lithographically defined
1.3 μm wide waveguides on 330 nm high Silicon nitride (SiN), deposited using Low Pressure
Chemical Vapor Deposition (LPCVD) at 800 ᵒC and annealed at 1200 ᵒC for 3 hours, on 4 µm
thermally oxidized SiO2, with a combination of e-beam lithography to define the waveguides and
deep UV lithography to pattern the CMP marks. We etch the SiN waveguides and CMP patterns
using CHF3/O2 chemistry in Oxford 100 Plasma ICP RIE, followed by the deposition of 600 nm
of Plasma Enhanced Chemical Vapor Deposition (PECVD) silicon dioxide (SiO2) on the
waveguides. We planarize the SiO2 to about 100 nm above the SiN waveguides using standard
chemical planarization (CMP) techniques to create a planar surface for the transfer of the MOCVD
WS2 monolayer and to prevent the TMD film from breaking at the waveguide edges. The power
splitter (combiner) at the input (output) of the MZI structure is designed using a 1 × 2 (2 × 1)
multimode interferometer (MMI). Due to the disproportionate ratio of the height (330 nm) to the
width (1.3 µm) of the waveguide and sensitivity of the MMI structure, the MZI supports
wavelength fringes from 1520 -- 1630 nm for the TE mode and 1455 -- 1480 nm for the TM mode.
Following the TMD preparation, transfer and patterning mentioned above, the metal contacts are
lithographically patterned using the DUV mask aligner and 50/80 nm of Ti/Au is deposited using
electron beam evaporation, followed by liftoff in acetone.
Device Fabrication for SiN-TMD Capactitive Devices. We lithographically defined 1 μm wide
waveguides on 360 nm high Silicon nitride (SiN), deposited using Low Pressure Chemical Vapor
Deposition (LPCVD) at 800 ᵒC and annealed at 1200 ᵒC for 3 hours, on 4 µm thermally oxidized
SiO2, using 248 nm deep ultraviolet lithography. We etch the SiN waveguides using CHF3/O2
chemistry in Oxford 100 Plasma ICP RIE, followed by the deposition of 600 nm of Plasma
Enhanced Chemical Vapor Deposition (PECVD) silicon dioxide (SiO2) on the waveguides. We
planarize the SiO2 to about 260 nm {40 nm} above the SiN waveguides using standard chemical
planarization (CMP) techniques to create a planar surface for the transfer of the WS2 {MoS2}
monolayer and to prevent the TMD film from breaking at the waveguide edges. The 260 {40 nm}
nm SiO2 layer prevents the lossy ITO layer of the capacitor from interacting with the optical mode
too strongly. A 45 nm {5 nm} of thermal ALD alumina is deposited on top of SiO2 to isolate the
waveguides from the subsequent fabrication steps for the TMD-HfO2-ITO capacitor. The TMD is
prepared, transferred and patterned using the aforementioned process. The metal contacts to the
TMD layer are then patterned using EBL and 30 nm Ti/ 50 nm Au is deposited using electron
beam evaporation, followed by lift-off in acetone. 26 nm {37 nm} of thermal ALD Hafnia at 200
C is then deposited to form the dielectric for the TMD-HfO2-ITO capacitor. The other electrode
of the capacitor is first patterned using EBL and then Indium Tin oxide (ITO) is sputtered at room
temperature (24 ᵒC), at a chamber base pressure of 12 mTorr, with an Argon flow of 30 sccm and
5 sccm oxygen flow, followed by lift-off in acetone. To reduce the resistivity of the sputtered ITO
film, the substrate is heated to a temperature of 200ᵒC and annealed in vacuum with the chamber
pressure at 10-6 Torr, and an oxygen flow of 5 sccm for 30 minutes. We characterize the complex
refractive index of ITO and thereby its doping using visible and infrared ellipsometry (Woollam
VASE) and fit to a Drude-Lorentz relation (see Supplementary Section XI). Finally, the metal
contacts to the ITO layer are patterned using EBL and 30nm Ti/ 50 nm Au is then deposited using
e-beam evaporation, followed by lift-off in acetone. To define devices with SU-8 on the composite
SiN-WS2 waveguides, SU-8 is photolithographically patterned on the waveguide and developed,
by spinning SU-8 to a thickness of 3.5 μm.
Note. The numbers in curly brackets {} indicate the dimensions for the composite SiN-MoS2
devices.
Experimental Setup. We couple TE/TM polarized light from a tunable NIR laser (1450 nm --
1600 nm) to the SiN microring/MZI input using a tapered single mode fiber, which is then
collected at the SiN bus/MZI output, using a similar tapered fiber. We record the microring/MZI
transmission spectrum for different DC bias voltages applied across the WS2-HfO2-ITO capacitor,
to measure the phase shift and absorption change. We normalize each of the MZI/ring transmission
spectra first by the maximum output power across the measured wavelength range and then by the
optical power spectral profile measured at the output of one of the MZI arms by coupling light to
a similar SiN waveguide with only one MMI splitter at the input of MZI. The above normalization
process eliminates the wavelength dependent optical loss experienced by the incident laser light in
the tapered optical fiber, the fiber polarizer, SiN waveguide and the MMI, which are all designed
to operate optimally at 1550 nm with the TE polarization. It also accounts for the wavelength
dependent power fluctuations of the laser, if any.
For small-signal RF bandwidth measurement using the Vector Network Analyzer (VNA),
we apply a DC bias voltage of 0 V across the WS2-HfO2-ITO capacitor, combined with an RF
signal of -10 dBm with a source impendance of 50 Ω.
Optical Sheet Conductivity of Monolayer TMD. We use the 2D sheet conductivity model to
extract the ∆𝑛WS2 and ∆𝑘WS2, as is commonly done when modeling graphene monolayers42. We
extract the complex index change of the monolayer (∆𝑛 + 𝑖∆𝑘) with carrier densities by comparing
the measured ∆𝑛eff and ∆𝑘eff (Figure 1e)), with the simulated change obtained using COMSOL
Multiphysics finite element model (FEM). We model the monolayer TMD as a conductive sheet
with a surface charge density (J = σs · E) and complex optical conductivity43 𝜎𝑠 = 𝜎𝑅 + 𝑗𝜎𝐼. σs(ω)
is related to the dielectric permittivity (𝜀(𝜔)) through the equation
𝜎𝑠(𝜔) = 𝑗𝜔𝑡𝑑𝜀0 (𝜀(𝜔) − 1)
(1)
where, 𝑡𝑑 defines the thickness of the monolayer TMD (0.85 nm ± 0.05 nm)21. We estimate the
complex index change of the monolayer from the computed change in conductivity (∆σ) using the
equation
Where, ∆𝜀 defines the change in dielectric permittivity (𝜀(𝜔)), which is related to ∆𝑛 and
𝜎𝑠(𝜔) + ∆𝜎 = 𝑗𝜔𝑡𝑑𝜀0 (𝜀(𝜔) + ∆𝜀 − 1) (2)
∆𝑘 through
𝜀(𝜔) + ∆𝜀 = (𝑛 + ∆𝑛 + 𝑖(𝑘 + ∆𝑘))2
(3).
The change in the real part of effective index of the mode (∆neff) signifies a change in the imaginary
part of the optical conductivity (∆𝜎𝐼), whereas the change in the imaginary part of the effective
index (∆𝑘eff) of the mode is reflected in the real part of the optical conductivity (∆𝜎𝑅) through
equation (1), (2) and (3). Since the change in complex effective index of the composite SiN-WS2
waveguide is contingent on the spatial overlap between the propagating optical mode and the
monolayer WS2, the true measure of the efficiency of the phase modulation lies in the change of
its in-plane optical sheet conductivity (S) and thereby its real and imaginary part of the refractive
index (∆n+ 𝑖∆k) with electrostatic gating. We calculate (∆nWS2+ 𝑖∆kWS2) by modeling the
monolayer as a boundary condition with surface current density (J = σs · E), which eliminates the
approximation of the thickness of the monolayer in our COMSOL model.
We quantify the mode overlap by replacing the surface current density (J = σ · E) in the
refractive index calculations, with a monolayer thickness (𝑡𝑑) of 0.85 ± 0.05 nm thickness, and
finding the surface integral to fraction of the energy flux (E × H) of the mode in monolayer WS2,
compared to the total mode flux in the composite waveguide.
Comparison of the electro-refractive change in WS2 and MoS2 - Even though the ∆𝑛eff for the
air clad SiN-WS2 composite waveguide (Figure 3c) is similar to the ∆𝑛eff of the air clad SiN-MoS2
(Figure 4a), the extracted ∆𝑛WS2 is higher than ∆𝑛Mos2, due to the lower optical mode overlap of
0.016 % in the SiN-WS2 waveguide, as compared to that of SiN-MoS2 waveguide.
|
1906.12254 | 2 | 1906 | 2019-07-18T08:10:18 | Zero-field propagation of spin waves in waveguides prepared by focused ion beam direct writing | [
"physics.app-ph",
"cond-mat.mes-hall"
] | Metastable face-centered-cubic Fe78Ni22 thin films grown on Cu(001) substrates are excellent candidates for focused ion beam direct writing of magnonic structures due to their favorable magnetic properties after ion-beam-induced transformation. The focused ion beam transforms the originally nonmagnetic fcc phase into the ferromagnetic bcc phase with additional control over the direction of uniaxial magnetic in-plane anisotropy. The magnetocrystalline anisotropy in transformed areas is strong enough to stabilize the magnetization in transverse direction to the long axis of narrow waveguides. Therefore, it is possible to propagate spin waves in these waveguides without the presence of an external magnetic field in the favorable Demon-Eshbach geometry. Phase-resolved micro-focused Brillouin light scattering yields the dispersion relation of these waveguides in zero as well as in nonzero external magnetic fields. | physics.app-ph | physics | Zero-field propagation of spin waves in waveguides prepared by focused ion beam direct writing
Lukáš Flajšman1*, Kai Wagner2, Marek Vaňatka1, Jonáš Gloss3, Viola Křižáková4, Michael Schmid3, Helmut Schultheiss2 and
Michal Urbánek1,4*
1 CEITEC BUT, Brno University of Technology, Brno, Czech Republic
2 Institute of Ion Beam Physics and Materials Research, HZDR, Dresden, Germany
3 Institute of Applied Physics, TU Wien, Vienna, Austria
4 Institute of Physical Engineering, Brno University of Technology, Brno, Czech Republic
*E-mail: [email protected], [email protected]
Metastable Fe78Ni22 thin films are excellent candidates for focused ion beam direct writing of magnonic structures
due to their favorable magnetic properties. The focused ion beam transforms the originally nonmagnetic fcc phase
into the ferromagnetic bcc phase with an additional control over the direction of uniaxial magnetic in-plane
anisotropy and the saturation magnetization. The induced anisotropy allows to stabilize transverse direction of
magnetization in narrow waveguides. Therefore, it is possible to propagate spin waves in these waveguides in the
favorable Demon-Eshbach geometry without the presence of any external magnetic field.
Nowadays, the vibrant field of magnonics stands on the edge between development of elementary building blocks of magnonic
circuitry and envisioned all magnon on-chip devices [1,2]. The magnonic devices, utilizing physics of spin waves, are
recognized to have potential in information processing in the frequency range from gigahertz to terahertz. High frequencies,
together with low energy of elementary excitations render the magnonic devices suitable for beyond-CMOS computational
technologies. Many concepts of future devices used for steering and manipulating spin waves have been presented recently [3-
6]. To allow further advances in this field, new types of materials possessing additional means of control over their magnetic
properties together with good spin wave propagation are needed.
Here we show, that magnonic waveguides allowing for fast spin-wave propagation at zero magnetic field can be
directly written into metastable Fe78Ni22 thin films by focused ion beam (FIB). The local dose and scanning strategy controls
both the saturation magnetization and the magnetocrystalline anisotropy (direction, type and strength) of irradiated areas. The
unique possibilities of this material system allow to overcome the shape anisotropy of long magnonic waveguides and stabilize
1
the magnetization along the waveguide's short side even in the absence of any external magnetic field. This is of great
importance for using magnons as information carriers because the geometry where the magnetization is aligned perpendicularly
to the propagation direction has the maximum group velocity.
The Fe78Ni22 layers grow on Cu(001) substrate in the metastable nonmagnetic fcc phase and can be transformed by ion
irradiation into the ferromagnetic bcc phase [7]. Using FIB the fcc->bcc transformation can be precisely controlled which
results in unprecedented local control over magnetic properties [8]. This approach removes the need for complicated multi-step
lithography processing and allow rapid prototyping of individual structures on the sample with additional possibility to control
the magnetic properties of each structure [8-12]. Classical approaches of nanostructuring of magnetic materials such as optical
or electron beam lithography combined with lift-off processing [13], wet [14] or dry [15] etching or ion implantation [16,17]
do not offer the same control over local material properties as the fabrication processes rely on binary selection of adding or
removing the magnetic materials.
The FIB-written waveguides together with a microwave antenna for spin-wave excitation are shown in Fig. 1 a). The
waveguides are 30 µm long with nominal widths of 3, 2 and 1.5 µm and are clearly visible in the SEM image.
2
Fig. 1: a) SEM micrograph of FIB prepared waveguides (bcc waveguides appear dark grey). The crystallographic orientations of the fcc
matrix and bcc waveguides are indicated (for the latter, inferred from the magnetic anisotropy, double-headed arrow). The microwave
antenna is highlighted by ocher color. b) hysteresis loops of three waveguides for two different orientations of the magnetic field [blue and
red arrows showing the respective orientation of the magnetic field are shown in a)]. c) 2D frequency maps of thermal spin-wave spectra
measured by micro-focused BLS scanning over the waveguides in the transverse direction along the red lines in (a) in zero external magnetic
field (top row) and in the applied magnetic field 𝐵𝑒𝑥𝑡−0° = 50 mT (bottom row). The spatially and magnetic field invariant mode at approx.
3 GHz is a spurious laser mode.
Prior to the FIB processing the metastable fcc Fe78Ni22 thin film of nominal thickness 12 nm was grown under UHV
conditions on a Cu(001) single crystal substrate using the procedure described in [7]. After deposition, the sample was
transferred to the FIB-SEM microscope for FIB irradiation. We oriented the long axis of the waveguides along the fcc[010]
direction of the Cu substrate in order to obtain the highest possible magnetocrystalline anisotropy. To imprint the anisotropy
direction perpendicularly to the long axis of the waveguide, we wrote the structures with a single pass of a 30 keV Ga+ ion
3
beam (30 nm spot, beam current of 150 pA and 5 s dwell time) with the fast scanning direction rotated by 80° from the
waveguide's long axis. The resulting ion dose 4 × 1015 ions/cm2 has given reliable growth conditions for all the waveguides.
Nucleation of the bcc structure was facilitated by starting the growth in triangular 5 µm wide region with the doubled ion dose
of 8 × 1015 ions/cm2 (again in single FIB scan).
After the irradiation, the magnetic waveguides were investigated by Kerr magnetometry [18] and Kerr microscopy.
When the field 𝐵ext−90° is applied parallel to the long axis of the waveguide, hard-axis hysteresis loops [blue lines in Fig. 1 b)]
with effective anisotropy fields [19] in the range of 20 − 24 mT are observed for all waveguides. When the field 𝐵ext−0° is
applied perpendicular to the waveguide, easy-axis loops [red lines in Fig. 1 b)] with coercive fields in the range of 4 − 8 mT
are observed.
Thermal spin-wave spectra obtained by micro-focused Brillouin light scattering microscopy [20] are shown in Fig. 1
c) for zero magnetic field and for the external field 𝐵ext−0° = 50 mT. The signal is proportional to the density of spin-wave
states at the detection frequency (𝑦-axis) and shows pronounced spin-wave spectral intensity localized solely in the areas
irradiated by the FIB. The bandgap of the spin-wave band structure for the middle of the waveguide (𝑥 = 0 µm) can be clearly
seen as a sudden increase in the density of spin-wave states for frequencies higher than approx. 6 GHz in zero field and at
approx. 10 GHz in the external field of 50 mT.
All three waveguides show qualitatively the same behavior. The data also reveals the local change in spin-wave spectra
towards the sides of the waveguide. Here localized low frequency modes appear at approx. 4 GHz in zero field and at approx.
8 GHz in the external field of 50 mT. This is a clear indication of the transverse orientation of the magnetization and its inherent
demagnetizing field leading to lower effective fields at the waveguide edges and thus directly resulting in the localized spin-
wave edge modes [21,22]. The overall analysis is depicted for three vertically oriented waveguides only as the behavior of the
horizontally oriented waveguides show qualitatively same behavior (the anisotropy is again imprinted perpendicular to the long
axis of the waveguides). This also demonstrate unique potential of our approach when compared to other less versatile
approaches or materials with global magnetocrystalline anisotropy.
In the following experiments, we extract the magnetic field-dependence of the spin-wave dispersion relation. By fitting
the measured dispersion, we were able to obtain the full set of magneto-dynamic parameters of the material. The µBLS does
not directly sense the phase of the detected spin waves and thus it does not allow to determine the wavelength 𝜆 (or equivalently
4
the wave propagation vector 𝑘⃗ ) of the spin waves. In order to extract the wave-vector information we employed the phase-
resolved µBLS technique [20,23]. The method directly reveals the spatial profile of the spin-wave phase by letting the scattered
photons interfere with a reference spatially invariant signal created by an electro-optic modulator (EOM).
We recorded the interference signal along the 1.5 µm wide waveguide (𝑥 = 0 µm) with a step size of 120 nm from
the edge of the exciting antenna up to 7 µm distance at the microwave frequency of 10.2 GHz. The phase-resolved
measurements are shown in Fig. 2.
Fig. 2: Phase-resolved BLS microscopy interference intensity map for various external magnetic fields at fixed excitation frequency of 10.2
GHz. Individual linescans have been normalized and space-invariant background was subtracted. The bottom graph shows line profiles
extracted from the intensity map for 15 mT (red line) and for zero external magnetic field (blue line).
We measured the BLS interference linescans in zero and applied external magnetic field (perpendicular to the waveguide long
axis) up to 15 mT (1 mT step). The BLS interference intensity map shows gradual transition of the spin-wave wavelength from
the lowest value found at zero field up to the longest wavelengths at 15 mT. This is additional evidence for the presence of the
Damon-Eshbach geometry even at zero field since otherwise a decrease in wavelength with increasing field would be expected.
To extract the spin-wave wavelength, we fitted the measured data with the simple interference model:
5
𝐼(𝑦) = 𝐼SW(𝑦) + 𝐼EOM + 2√𝐼SW(𝑦)𝐼EOMcos(𝜃(𝑦)),
(1)
where we assume 𝐼SW = 𝐼maxe−𝑦/𝐿att.. The parameters 𝐼EOM and 𝐼max are EOM and spin-wave maximum intensities, 𝐿att. is a
spin-wave attenuation length and the total phase difference is 𝜃 = 2𝜋
𝑦
𝜆𝑙
+ 𝜃0 with 𝜆𝑙 representing the longitudinal spin-wave
wavelength, 𝑦 is the distance from the excitation antenna, and 𝜃0 is an arbitrary phase offset in-between the EOM and the spin-
wave excitation. We fitted the experimental data using Eq. (1) with all five parameters unconstrained. The two important fit
parameters are the wavelength [shown in Fig. 3 a) as a function of 𝐵ext] and the attenuation length. The largest attenuation
length 𝐿att. = 3.1 ± 0.4 µm is measured for zero external magnetic field and the frequency of 10.2 GHz. The spin-wave
dispersion at zero field obtained from these fits is shown in Fig. 3 b).
From the dispersion, we can determine the magnetic parameters using the model of Kalinikos and Slavin [24], while taking
into account the finite width of our waveguides by assuming the effective boundary conditions as described by Guslienko et al.
[25]. The effective dipolar conditions yield an effective width 𝑤eff = 𝑤𝑑/(𝑑 − 2) of the waveguide, determined by the
geometric width 𝑤 and the pinning parameter 𝑑 = 2𝜋/(𝑝 + 2𝑝ln(1/𝑝)). The parameter 𝑝 is given by 𝑝 = 𝑡/𝑤, where 𝑡 is the
thickness of the magnetic material. For the spin wave dispersion model, we assumed that the external magnetic field 𝐵ext points
in the direction of the magnetic anisotropy, i.e. perpendicularly to the long edge of the waveguide (magnetic anisotropy is
introduced in the form of effective magnetic field 𝐵ani). With all the terms in place the model is following:
2
𝑓2/𝛾2 = (𝐵ext + 𝐵ani + 𝜇0𝑀s𝑃 sin2( ∡𝑘) + 𝐴ex𝑘2)( 𝐵ext + 𝐵ani − 𝜇0𝑀s(1 − 𝑃) + 𝐴ex𝑘⃗
),
(2)
𝑤ℎ𝑒𝑟𝑒 𝑓 is the spin-wave frequency and 𝛾 is the gyromagnetic ratio. 𝑀s is the saturation magnetization and 𝐴ex is an exchange
stiffness. Since the thickness of the waveguide is small with respect to the other dimensions and the spin-wave wavelength, we
only calculate the first branch of spin waves along the thickness, yielding 𝑃 = 1 − (1 − exp(−𝑘⃗ 𝑡)/𝑘⃗ 𝑡). The total
propagation vector 𝑘⃗ = √𝑘∥
2 + 𝑘⊥n
2 comprises the longitudinal component 𝑘∥ (parallel to the long axis of the waveguide) and
transverse quantized component 𝑘⊥n. The angle of the spin wave propagation vector with the long waveguide axis is then
∡𝑘 = atan (𝑘∥/𝑘⊥n). The amplitude of the transverse component 𝑘⊥n is calculated from the width quantization condition 𝑘⊥n =
𝑛𝜋/𝑤eff for 𝑛 = 1, 2, …. In the micro-focused BLS experiment, as we record only the interference pattern along the long
waveguide axis, we see only the longitudinal component of the propagation vector 𝑘∥ = 2𝜋/𝜆l.
6
Fig. 3: a) the dependence of the spin-wave wavelength 𝜆𝑙 [extracted from equation (1)] on the external magnetic field for three different RF
frequencies fitted with a model given by equation (2). b) experimental and calculated (blue line) spin-wave dispersions together with
calculated group velocity (red line) at zero external magnetic field. The error bars have been calculated from the 95% fit confidence bounds.
We performed the fit using Eq. (2) [considering 95% confidence intervals obtained by fitting of the Eq. (1)] for various width
modes (and their linear combinations [26,27]) with one set of unconstrained universal parameters and we observed the total
minimal residuals of the fit. The best fit was found solely for single mode of 𝑛 = 1 (in contrast to experiments in e.g. permalloy
waveguides [26,27]), with magnetic parameters of 𝑀s = 1.41 ± 0.03 MA/m, 𝛾 = 29.3 ± 0.1 GHz/T, 𝐵ani = 24 ± 1 mT, 𝑡 =
9.5 ± 1.0 nm, 𝐴ex = 11 ± 5 pJ/m and 𝑤 = 1.62 ± 0.05 µm. The obtained fit parameters lie close to the bulk values of single
crystal iron films [28]. The saturation magnetization 𝑀s = 1.41 MA/m is expectedly lower then bulk value of iron (𝑀s
Fe = 1.7
MA/m). If we consider 22% of nickel (𝑀s
Ni = 0.51 MA/m) in our films, we estimate the expected saturation magnetization to
𝑀s
FeNi = 1.45 MA/m which is very close to the obtained value 𝑀s = 1.41 MA/m. The large saturation magnetization together
with the Damon-Eshbach geometry result in a high group velocity 𝑣𝑔 = 𝜕𝑓/𝜕𝑘∥ of the spin waves reaching almost 6 km/s [see
Fig. 3 b)]. The anisotropy field resulting from the fit perfectly reproduces the value measured by Kerr microscopy, which
further supports the validity of the model. We also see that the film is transformed to the magnetic bcc phase throughout the
whole thickness. The fitted thickness is very close to the nominal thickness of 12 nm (the few topmost layers are expected to
oxidize when performing ex-situ experiments). This is also supported by the Monte Carlo ion stopping range simulations using
7
the SRIM/TRIM package [29], where more than 90% of the 30 keV Ga+ ions penetrate to the copper substrate. The width of
the waveguide 𝑤 = 1.62 µm is slightly larger than the nominal value (1.5 m). First and presumably the major contribution
increasing the width of the waveguide is the finite size and shape of the focused ion beam spot. Furthermore, in our previous
work [30] it was shown that the bcc crystallites protrude to the fcc phase slightly further from the ion impact spot and thus they
again effectively increase the width of the waveguides (the protrusion length is approx. 50 nm). Both effects effectively create
a gradient in the magnetization affecting the dynamic boundary conditions of our waveguides, which differs from the
discontinuous boundary conditions found e.g. in structures prepared by classical lithography techniques [26,27]. We performed
micromagnetic simulations in mumax3 [31] to study effects of the magnetization gradient at the waveguide edges on the spin-
wave dispersion (for the approach see supplementary material [32]). In the simulations we continuously decreased the
magnetization from the bulk value to zero in a defined region at the edge of the waveguide. The introduced profile of the
magnetization was chosen as an error function as it resembles the convolution of the nominal shape of the waveguide with a
FIB spot. As the gradient of the saturation magnetization is introduced it is expected that the effective magnetic field, the major
driving force affecting the local spin-wave dispersion, will differ from discontinuous case where the saturation magnetization
changes abruptly. The transverse profiles of the saturation magnetization together with simulated effective (internal) magnetic
field at zero external magnetic field are shown in Fig. 4 a).
Fig. 4: a) shows the transverse profiles of the saturation magnetization (grey) and effective magnetic field (red) for discontinuous (left panel)
and continuous (right panel) transition of saturation magnetization on the edges of the waveguide at zero external magnetic field for material
parameters obtained by the fitting procedure. The continuous transition was implemented as a quasi-continuous modulation of 𝑀𝑠 expressed
by an equation shown above the plots. Transition width (𝑤) of the complementary error function is designated in each plot. b) dispersion
relations of the spin-wave modes for both magnetization profiles extracted from micromagnetic simulations. The density of states (represented
by color scale) has been normalized to the maximum value.
8
The micromagnetic simulations reveal, that in the case of the infinite gradient in the magnetization leads to abrupt increase of
the effective magnetic field (𝐵eff) on the edges of the waveguide. In the case of finite magnetization gradient, the 𝐵eff profile
is significantly changed. The maximal value of the 𝐵eff is lower and the central region is significantly broadened, as the lower
gradient of magnetization leads to lower and less localized demagnetizing field. The longitudinal spin-wave dispersion
extracted from the micromagnetic simulations is plotted in Fig. 4 b). There is a very apparent change in the modal structure of
the dispersion when we compare the case with 𝑤 = 0 nm and the case where 𝑤 = 50 nm. As the gradient is introduced to the
edge region, the boundary conditions for the dynamic magnetization are altered. The quantization of the modes with higher
transverse mode numbers is deteriorated. This leads to less effective excitation of the higher order modes when compared to
e.g. fundamental mode (as can be seen from a lower modal density of states). The first waveguide mode does not significantly
change even for 𝑤 = 200 nm. This is likely the main mechanism explaining the absence of any higher order waveguide modes
seen in the linescans. This statement is further supported by careful analysis of the linescans (see Fig. 2). In additional analysis
we readjusted the equation (1) to allow for more spatial frequencies to be detected since we expect from the modal profiles of
the analytical model presented by equation (2) to excite multiple odd modes [27,33] at certain frequency by the excitation
antenna. The analysis confirmed best agreement in the absence of any higher spatial frequencies.
In conclusion we studied the spin-wave propagation in the waveguides prepared by FIB direct writing into metastable
fcc Fe78Ni22 thin films. We have shown that in these high-aspect ratio waveguides we can propagate spin waves without the
presence of an external magnetic field, and with high group velocities reaching almost 6 km/s. This unique feature comes from
the possibility of the local uniaxial magnetic anisotropy control. The spin-wave dispersion relation has been determined by
using phase-resolved BLS microscopy and the magnetic properties of the waveguides were extracted. The relatively large
saturation magnetization together with high (controllable) magnetic anisotropy render the material suitable for high frequency
spin-wave circuits operational even at zero external magnetic field. Moreover, the extracted material properties of the system
will allow us to design more complex spin-wave devices by utilizing the possibility to spatially control both the saturation
magnetization and the direction of the uniaxial magnetic anisotropy in a single magnetic structure. Our unique approach paves
the way towards many other possibilities to develop and study spin-wave propagation in magnetization landscapes that are
unattainable in any conventional magnetic system.
9
Acknowledgements:
This research has been financially supported by the joint project of Grant Agency of the Czech Republic (Project No. 15-
34632L) and Austrian Science Fund
(Project
I 1937-N20) and by
the CEITEC Nano+ project
(ID
CZ.02.1.01/0.0/0.0/16013/0001728). Part of the work was carried out in CEITEC Nano Research Infrastructure (ID
LM2015041, MEYS CR, 2016 -- 2019). L.F. was supported by Brno PhD talent scholarship.
References:
[1]
[2]
[3]
S. Neusser, and D. Grundler, Adv. Mater. 21, 2927-2932 (2009).
A. V. Chumak, V. I. Vasyuchka, A. A. Serga, and B. Hillebrands, Nat. Phys. 11, 453-461 (2015).
E. Albisetti, D. Petti, G. Sala, R. Silvani, S. Tacchi, S. Finizio, S. Wintz, A. Calò, X. Zheng, J. Raabe, E. Riedo, and
R. Bertacco, Commun. Phys. 1, 56 (2018).
[4]
K. Wagner, A. Kákay, K. Schultheiss, A. Henschke, T. Sebastian, and H. Schultheiss, Nat. Nanotechnol. 11, 432 -- 436
(2016).
[5] M. Vogel, R., Assmann, P. Pirro, A. V. Chumak, B. Hillebrands, and G. von Freymann, Sci Rep. 8(1), 11099 (2018).
[6]
[7]
A. V. Chumak, A. A. Serga, and B. Hillebrands, Nat. Comm. 5, 4700 (2014).
J. Gloss, S. S. Shah Zaman, J. Jonner, Z. Novotný, M. Schmid, P. Varga, and M. Urbánek, Appl. Phys. Lett. 103,
262405 (2013).
[8] M. Urbánek, L. Flajšman, V. Křížáková, J. Gloss, M. Horký, M. Schmid, and P. Varga, APL Matter. 6, 060701 (2018).
[9]
C. Chappert, H. Bernas, J. Ferre, V. Kottler, J.-P. Jamet, Y. Chen, E. Cambril, T. Devolder, F. Rousseaux, F. Mathet,
and H. Launois, Science 280, 1919 (1998).
[10]
R. Bali, S. Wintz, F. Meutzner, R. Hubner, R. Boucher, A. A. Unal, S. Valencia, A. Neudert, K. Potzger, J. Bauch, F.
Kronast, S. Facsko, J. Lindner, and J. Fassbender, Nano Lett. 14, 435 -- 441 (2014).
[11]
F. Roder, G. Hlawacek, S. Wintz, R. Hübner, L. Bischoff, H. Lichte, K. Potzger, J. Lindner, J. Fassbender, and R.
Bali, Sci. Rep. 5, 16786 (2015).
[12] W. Rupp, A. Biedermann, B. Kamenik, R. Ritter, Ch. Klein, E. Platzgummer, M. Schmid, and P. Varga, Appl. Phys.
Lett. 93, 063102 (2008).
[13]
K. Vogt, F.Y. Fradin, J.E. Pearson, T. Sebastian, S.D. Bader, B. Hillebrands, A. Hoffmann, and H. Schultheiss, Nat.
Comm. 5, 3727 (2014),
10
[14]
C. L. Ordonez-Romero, Z. Lazcano-Ortiz, A. Drozdovskii, B. Kalinikos, M. Aguilar-Huerta,J. L. Dominguez-Juarez,
G. Lopez-Maldonado, N. Qureshi, O. Kolokotsev, and G. Monsivais, J. Appl. Phys. 120, 043901 (2014).
[15]
C. Bayer, J. Jorzick, B. Hillebrands, S. O. Demokritov, R. Kouba, R. Bozinoski, A. N. Slavin, K. Y. Guslienko, D. V.
Berkov, N. L. Gorn, and M. P. Kostylev, Phys. Rev. B 72, 064427 (2005).
[16]
B. Obry, P. Pirro, T. Brächer, A. V. Chumak, J. Osten, F. Ciubotaru, A, A. Serga, J. Fassbender, and B. Hillebrands,
Appl. Phys. Let. 102, 202403 (2013).
[17]
J. Fassbender, J. von Borany, A. Mucklich, K. Potzger, W. Moller, J. McCord, L. Schultz, and R. Mattheis, Phys. Rev.
B 73, 184410 (2006).
[18]
[19]
[20]
[21]
[22]
[23]
L. Flajšman, M. Urbánek, V. Křižáková, M. Vaňatka, I. Turčan, and T. Šikola, Rev. Sci. Instrum. 87, 053704 (2016).
The anisotropy field is called effective as we see only the result of the two competing contributions -- crystalline
uniaxial magnetic anisotropy coming from the FIB writing process and from the uniaxial shape anisotropy having its
origin in the shape of the waveguide (approx. 7 mT for 1.5 µm wide waveguide). The total anisotropy of the structure
is then obtained as the difference of the two contributions.
T. Sebastian, K. Schultheiss, B. Obry, B. Hillebrands, and H. Schultheiss, Front. Phys. 3, 1-23 (2015).
V. E. Demidov, S. O. Demokritov, K. Rott, P. Krzysteczko, and G. Reiss, Appl. Phys. Lett. 92, 1-4 (2008).
J. P. Park, P. Eames, D. M. Engebretson, J. Berezovsky, and P. A. Crowell, Phys. Rev. Lett. 89, 277201 (2002).
K. Vogt, H. Schultheiss, S. J. Hermsdoerfer, P. Pirro, A. A. Serga, and B. Hillebrands, Appl. Phys. Lett. 95, 0-3 (2009).
[24]
B. A. Kalinikos, and A. N. Slavin, J. Phys. C: Solid State Phys. 19, 7013 (1986).
[25]
[26]
K. Yu. Guslienko, S. O. Demokritov, B. Hillebrands, and A. N. Slavin, Phys. Rev. B 66, 132402 (2002).
V. E. Demidov, M. P. Kostylev, K. Rott, P. Krzysteczko, G. Reiss, and S. O. Demokritov, Appl. Phys. Lett. 95, 112509
(2009).
[27]
V. E. Demidov, S. O. Demokritov, K. Rott, P. Krzysteczko, and G. Reiss, Phys. Rev. B 77, 064406 (2008).
[28]
O. Gladii, D. Halley, Y. Henry, and M. Bailleuel, Phys. Rev. B 96, 174420 (2017).
[29]
[30]
J. F. Ziegler, M. D. Ziegler, and J. P. Biersack, Nucl. Instrum. Methods Phys. Res. B 268, 1818 (2010).
S. S. Zaman, P. Dvořák, R. Ritter, A. Buchsbaum, D. Stickler, H. P. Oepen, M. Schmid, and Peter Varga, J. Appl.
Phys. 110, 024309 (2011).
[31]
A. Vansteenkiste, J. Leliaert, M. Dvornik, M. Helsen, F. Garcia-Sanchez, and B. Van Waeyenberge, AIP Adv. 4 (10),
107133 (2014).
11
[32]
See Supplemental Material at LINK for additional information on the micromagnetic simulations.
[33]
C. Kittel, Phys. Rev. 110, 1295 (1958).
12
|
1708.05434 | 1 | 1708 | 2017-08-17T20:44:32 | Biaxial magnetic field setup for angular magnetic measurements of thin films and spintronic nanodevices | [
"physics.app-ph",
"cond-mat.mes-hall"
] | The biaxial magnetic-field setup for angular magnetic measurements of thin film and spintronic devices is designed and presented. The setup allows for application of the in-plane magnetic field using a quadrupole electromagnet, controlled by power supply units and integrated with an electromagnet biaxial magnetic field sensor. In addition, the probe station is equipped with a microwave circuitry, which enables angle-resolved spin torque oscillation measurements. The angular dependencies of magnetoresistance and spin diode effect in a giant magnetoresistance strip are shown as an operational verification of the experimental setup. We adapted an analytical macrospin model to reproduce both the resistance and spin-diode angular dependency measurements. | physics.app-ph | physics |
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
Biaxial magnetic-field setup for angular magnetic
measurements of thin films and spintronic nanodevices
Piotr Rzeszut,1, a) Witold Skowroński,1 Sławomir Ziętek,1 Piotr Ogrodnik,2, b) and
Tomasz Stobiecki1, 3
1)AGH University of Science and Technology, Department of Electronics,
Al. Mickiewicza 30, 30-059 Kraków, Poland
2)University of Michigan, Department of Electrical Engineering and Computer Science,
Ann Arbor, MI 48109, USAc)
3)AGH University of Science and Technology, Faculty of Physics
and Applied Computer Science, Al. Mickiewicza 30, 30-059 Kraków,
Poland
(Dated: 21 August 2017)
The biaxial magnetic-field setup for angular magnetic measurements of thin
film and spintronic devices is designed and presented. The setup allows for
application of the in-plane magnetic field using a quadrupole electromagnet,
controlled by power supply units and integrated with an electromagnet biax-
ial magnetic field sensor. In addition, the probe station is equipped with a
microwave circuitry, which enables angle-resolved spin torque oscillation mea-
surements. The angular dependencies of magnetoresistance and spin diode
effect in a giant magnetoresistance strip are shown as an operational verifica-
tion of the experimental setup. We adapted an analytical macrospin model to
reproduce both the resistance and spin-diode angular dependency measure-
ments.
Keywords: quadrupole electromagnet, magnetometry of thin magnetic films
and nanodevices, magnetoresistance, spin diode effect
I.
INTRODUCTION
Ferromagnetic resonance (FMR)1 in thin ferromagnetic films measured typically in a mi-
crowave regime can deliver much useful information important for spintronics applications2
such as magnetization saturation, magnetic anisotropy constant and Gilbert damping.
In patterned devices of micro- to nanometer sizes, FMR can be detected electrically us-
ing the spin-torque diode (SD) effect.3,4 In this effect, a radio frequency (RF) current
passes through an anisotropic magnetoresistance5,6, giant magnetoresistance7–9 or tunnel
magnetoresistance3,10 device, which, due to spin transfer or field torque effect, induces the
resistance oscillations. These oscillations mix with RF current and as a result produce an
output DC voltage.3,11
The SD effect can potentially be used in various applications, such as microvave
detectors10,12,13, modulators14 and demodulators15.
In addition, by performing the analysis of angular dependence of SD measurements,
one can determine the dynamic properties of the investigated magnetic system, such as
resonance fields or frequencies of both FMR6 and standing spin wave modes16.
An experimental approach often requires a characterization of prototype devices by ap-
plying an external magnetic field at different angles in plane or to measure angular relations
of electric and magnetic parameters (e.g. resistance vs. magnetic field angle)7. Such mea-
surements can be done by mechanically rotating a sample in a dipole electromagnet, using
perpendicular crossed Helmholtz coil pairs or using a quadrupole electromagnet.
a)Electronic mail: [email protected]
b)Electronic mail: [email protected]
c)permanent address: Warsaw University of Technology, Faculty of Physics, , ul. Koszykowa 75, 00-662
Warszawa, Poland
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
2
The first method requires a complex mechanical system, that is difficult to realize for
microwave measurements. In perpendicularly crossed Helmholtz coils an arbitrary magnetic
field angle can be easily applied by setting suitable currents for each pair of coils. However
the magnitude of the field in such non-superconducting electromagnet is limited to a few
kA/m. A quadrupole electromagnet enables the increase of the maximum field magnitude,
while still allowing to control the field angle electrically. However, due to the non-linear
magnetizing curve of the ferromagnetic cores and strong influence of coil pairs on each other,
a precise magnetic-field control system is needed.
In this work, we present a complete microwave and magnetoresistance measurement sys-
tem that allows for varied angle-resolved determination of spintronic device parameters.
Application of a magnetic field up to 160 kA/m is possible at any angle. The field may
be controlled using a proportional–integral–derivative (PID) algorithm or without a feed-
back loop. The measurement setup consists of the following: RF generator, RF spectrum
analyser, bias tee, sourcemeter, and lock-in voltmeter. In addition in this work we present
a specially designed electromagnet-pole-shape which can increase field homogeneity inside
and extend the area where field is homogeneous, without increasing the outer dimensions
of the magnet. Dedicated software enables measurement using the presented hardware,
especially DC voltage vs. RF current characterisation.
II A we describe the generation and measurements of magnetic field by a
quadrupole electromagnet and electronics implementation of a magnetic field vector con-
troller. In Sec. II B we present the measurement setup used to examine SD voltage and in
Sec. II C a sample structure. Later we discuss the results of using the designed system and
present measurement of SD voltage.
In Sec.
II. EXPERIMENTAL
A. Magnetic field generation and measurement
1. Quadrupole electromagnet
To apply an arbitrary in-plane magnetic field vector, a quadrupole electromagnet is used.
Fig. 1. presents a drawing of the magnet consisting of: two pairs of coils placed orthogonally
to each other and four ferromagnetic cores mechanically connected to the outer frame. Two
voltage-controlled current sources (VCCS) are used for application of current in order to
generate the required magnetic field vector in the area between the poles.
A key issue in biaxial magnetic field sources is the homogeneity of the field, because it
is not always possible to place the element under measurement in the geometrical center
between the poles.
Therefore, finite element simulations were used to determine the region suitable for spin-
tronic measurements, taking field homogeneity under consideration. Thin film samples
inside the magnet can be treated as two-dimensional objects placed in the geometrical cen-
ter between the poles, therefore the problem of homogeneity measure can be simplified as
two-dimensional. Additionally, as will be mentioned afterwards, the magnetic field vector
( H) in the center of the magnet can be measured and controlled precisely, so H(0, 0, θset)
can be used as a reference value.
The representation of the field vector in polar coordinates (Eq. 1) enables calculation of
the error of the magnetic field angle and magnitude, where x and y represent the position
in the electromagnet, θset expected angle and θH (x, y, θset) actual angle at (x, y) position.
H(x, y, θset) =
= xH(x, y, θset) sin θH (x, y, θset)+
+ yH(x, y, θset) cos θH (x, y, θset)
(1)
Equations 2 and 3 represent the measure of homogeneity using a normalised maximum
difference in magnetic field magnitude (or angle) in the geometrical center between the poles
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
3
FIG. 1. Design of the quadrupole magnet. 1 - coil, 2 - outer frame, 3 - sample holder with bi-axial
magnetic field sensor, 4 - cylindrical core with spherical-shaped pole.
and field at the (x, y) position inside the circle of radius r.
∆H(r) =
max(x2+y2)(cid:54)r2
(cid:12)(cid:12)H(0, 0, θset) − H(x, y, θset)(cid:12)(cid:12)
H(0, 0, θset)
(2)
(3)
∆θH (r) =
max(x2+y2)(cid:54)r2 θH (0, 0, θset) − θH (x, y, θset)
π
Initially cores with flat poles were prepared, but the achieved homogeneity was not sat-
isfactory. After finite element simulations, the poles with spherical hollows were selected as
the best option to improve homogeneity. Both types of poles are presented in Fig. 2.
2. Magnetic field control algorithm
The magnetic field control system contains two VCCS, connected to two independent
pairs of coils. In the center between the poles a bi-axial magnetic field sensor is used for
magnetic field vector measurements.
For two perpendicularly crossed pairs of Helmholtz coils, the magnetic field could be
controlled by applying current according to simple trigonometric formulas, however, for
quadrupole electromagnet these relations are not true, as two pairs of cores mounted in the
electromagnet influence the magnetic field produced by each other. Therefore, the angle and
magnitude of the field do not follow the rules described above. Moreover, a step response
of the electromagnet was measured using a magnetometer in order to determine the time
constant of the electromagnet, which is equal to 600 ms. This relatively high value of the
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
4
FIG. 2. Overview of the poles. A - flat, B - with spherical holes.
FIG. 3. Field control and measurement device overview.
time constant adds complexity to controlling the magnetic field. The solution for these
issues is to use a feedback loop to control the field in the area between the poles. For this
purpose, two independent discrete PID controllers with anti-windup were used.
3. Magnetic field controller
The overview of the control and measurement device is presented on Fig. 3.
The hall sensor MLX90363 with SPI interface was used as a 3-axial magnetic field sensor.
This sensor provides data via a digital SPI interface. It allows for a sampling rate of 250
times per second with 14-bit resolution (13 bits + sign) when filtering and automatic gain
control is switched off. The measurement magnetic field range of this sensor is approx.
±35.8 kA/m for each axis. The sensor is placed on a separate printed circuit board utilizing
a cable connection to the device. The board is suitable to be installed inside a sample holder.
In this application we use two of the three available axes. Although the magnetic fields used
in the setup exceed the range of the sensor, a simple linear relation of the magnetic field
vs. current is used for higher magnetic field values. This approach is justified in the region
above the magnetic remamnence and below the saturation of the electromagnet core.
4. Control and signal processing
The ARM microcontroller (µC) STM32F407 was used as the main processing unit. It has
an embedded Floating-Point Unit (FPU), which enables performing high-speed mathemat-
ical calculations and is necessary to calculate the field parameters and run two independent
PID controllers with sufficient speed.
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
5
The communication interface is based on an FT232RL USB to UART converter. The
chip provides serial communication (Virtual Com Port) between the PC and the ARM
microcnotroller. The interface is powered via USB and optoisolated from the rest of the
hardware in order to ensure that the PC will not be damaged, even when some power issues
occur. The interface was also used to upgrade the firmware.
5. Analog output
Analog driving output signals are generated using DAC8531 (16 bit Digital to Analog
Converter with serial interface). The reference voltage is provided by REF193 which outputs
3 V. Using operational amplifiers (OP07CR) -1.5 V voltage is generated, and then added to
the output of each DAC. It results in shifting the range of voltage from 0∼3 V to ±1.5 V.
3 to provide output in the range of ±11 V.
Then a second set of amplifiers uses a gain of 22
Digital control signals are also optoisolated from control logic and powered from a separate
voltage regulator to minimize the noise generated by the microcontroller.
The circuit provides output voltage resolution of approx. 0.34 mV. Considering noise, a
real resolution of 1 mV was obtained. Two VCCS are provided by two Kepco power supplies
BOP 50-8M. In this configuration a field resolution of 40 A/m was obtained.
6. Power supply for magnetic field controller
The power supply is located on an external printed circuit board to allow easy replace-
ment. To avoid introducing high voltages in the system, the controller is supplied by a 5
V and 2 A power supply. Then two AM1P-0512SZ DC converters were used do provide
±12 V for the analogue output part. Also a low-dropout regulator LM1117-3.3 was used to
provide 3.3 V for the microcontroller.
7. User interface
The display and control module was also made as a separate printed circuit board to
allow mounting on a front panel. A HD44780 compatible 20×4 character display was used.
The display presents measurement results in each axis, error messages and mathematical
functions of field vector: projections to various planes, total magnitude, angles.
8. Software
As the measurement process is controlled by the PC, the quadrupole system is also
integrated with software designed using the LabVIEW environment. The software, together
with the necessary equipment, enables taking the following measurements:
• R(H)θH =const. - field resolved magnetoresistance
• R(θH )H=const. - angle resolved magnetoresistance
• VDC(H)θH =const.,f =const. - field resolved spin-torque diode
• VDC(θH )H=const.,f =const - angle resolved spin-torque diode
• VDC(f )H=const.,θH =const - frequency resolved spin-torque diode
The software also supports parametric sweeps allowing to prepare a set of measurements.
This is very useful when preparing for example a set of linear magnetic field sweeps with
different angles. A graphical interface is provided to select the equipment used and control
its parameters.
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
6
B. Spin diode measurement setup
The example experimental setup presented in Fig. 4. enables angular depedence mea-
surements of both magnetoresistance and SD voltage7. The bias Tee separates the RF input
signal from the output DC voltage generated by the examined spintronics device. Also, by
replacing the voltmeter with a sourcemeter, simultaneous resistance measurements can be
performed.
The setup utilizes an Agilent RF generator Model PSG E8257D with output power set
to 10 dBm, while for DC voltage measurements - an Agilent 34401A voltmeter. The sample
is connected to the system using a signal-ground RF probe provided by Picoprobe with
200 µm pitch. 20 GHz bandwidth RF cables are used to connect the probe, Bias Tee and
generator.
FIG. 4. Setup for angular SD effect measurement.
C. Sample
For test purposes, a spin valve giant magnetoresistance (SV-GMR) sample7, fabricated
by means of magnetron sputtering, was used (Fig. 5.) It consists of a composite NiFe/CoFe
(6 nm) free layer (FL) and CoFe (2.1 nm) reference layer (RL) separated by a Cu (2.1 nm)
spacer. RL is antiferromagnetically coupled through a thin Ru layer with a CoFe (2.0 nm)
pinned layer (PL) that is deposited on an antiferromagnetic (AFM) PtMn (18 nm). After
the deposition process the GMR stack was annealed in the in-plane magnetic field in order
to induce an exchange bias between AFM and PL, which is oriented paralelly to the stripe
axis. Next, micro-stripes of 5× 70 µm2 were patterned using direct write laser lithography,
ion-etching and lift-off process. A representative microscope image of the fabricated device
is presented in Fig. 5.
III. RESULTS AND DISCUSSION
A. Step response of the system
Measurements of magnetic field (H) versus time are presented in Fig. 6. Black lines
represent a response to change in power supply control voltage from 0 V to 0.2 V and 0.3
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
7
FIG. 5. Sample overview and layer structure (layer thickness in nm).
V, respectively. Red lines represent the results of running a PID controller to obtain the
same magnetic field change, as during the application of a voltage step.
FIG. 6. System response comparison for magnetic field H=7.9 kA/m (a) and H=12 kA/m (b) step,
tr are the time constants.
The initial value of the magnetic field for all measurements is greater than zero due to
remanence of cores, which is impossible to eliminate when using a voltage step, therefore
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
8
the initial value for the PID controller was set to the corresponding magnetic field obtained
using a voltage step.
To verify the strength and angle of the magnetic field vector applied in the center between
poles the steady state error (offset field and angle difference between the requested and
obtained values) was measured (Fig. 7.) when using the PID controller and without the
feedback loop.
FIG. 7. Measurements of the steady state error of angle (a) and magnitude of magnetic field H (b)
for cases without a feedback loop and with a PID feedback loop.
As a result a significant improvement of rise time was observed, as well as the ability to
cancel the remanence field of cores (steady state error). The maximum field difference for
the case with PID is 78 A/m and the angle difference shows a value of 0.13◦, compared to
1024 A/m and 1.65◦ without the feedback loop.
Measurements of the magnetic field vectors in an area between the cores were performed
to determine the uniformity of the field. Results of the measurements were analysed as
described in Sec. II A 1 and are presented on Figs. 8 and 9.
B. Angular measurements of spin valve GMR
In this section we show the reliability of the presented experimental setup by performing
static (GMR) and dynamic (SD) measurements of our spintronic device described in details
in Sec. II C.
First we show the angular-dependent resistance measurements. The external magnetic
field was set to H=31.8 kA/m. Next, the direction of the magnetic field was changed from
-180 to 180 degrees with respect to the long axis of the GMR microstrip. The results
are shown in Fig.10(a). The measured resistance R(θ) follows the cosine dependence as
predicted for the GMR structures7,17,18:
R(θ) = RP +
(1 − cos θ)
∆R
2
(4)
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
9
FIG. 8. Magnetic field magnitude error for cases with flat and spherical poles.
In the above Eq.(4) ∆R ≡ RAP − RP is the magnetoresistance, i.e. the difference in a
resistance of parallel(P) and antiparallel(AP) states (c.f. Fig.10(a)). Next, we measured
the SD voltage dependence on external magnetic field direction, and the results are presented
in Fig.10(b).
In order to verify our experimental results, we adapted the analytical model that was
previously developed for an exchange-biased GMR structure7. Here, the model accounts
for the exchange bias direction parallel to the strip's long axis.
First, we compared the experimental and theoretical dependence of resonance frequency
on magnetic field angle. This dependence is visible in Fig.10(b) as a frequency shift of the
experimental VDC spectra for different magnetic field angles. According to the macrospin
model (cf. Ref.7), the resonance frequency in the limit of small damping (α2 → 0) can be
expressed as:
(cid:118)(cid:117)(cid:117)(cid:116)(cid:32)
(cid:21)2(cid:33)
(cid:20) ∂2U
∂φ∂θ
f0 ≈ 1
2π
γe
µ0MS sin θM
∂2U
∂φ2
∂2U
∂θ2 −
(5)
where MS stands for saturation magnetization, γe is the gyromagnetic ratio, U (θ, φ) is the
magnetic total energy, and partial derivatives are calcualted at stationary magnetization
angles θM and φM, which are determined by the local minimum of U. Similarly as in Refs.7
and 11, the total magnetic energy U includes the shape anisotropy, the magnetocrystalline
anisotropy in the x direction, as well as the interlayer exchange coupling energy term. It
was also assumed that the Oersted field has only one component, i.e. HOe,x perpendicular
to the long axis. The calculated angular dependence of the resonance frequency is shown
in Fig.10(b) as a solid line.
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
10
FIG. 9. Magnetic field angle error for cases with flat and spherical poles.
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
11
FIG. 10. (a) Angular dependence of resistance: measured experimentally (black points) and pre-
dicted by the macrospin model (blue solid line), P and AP denote the parallel and antiparallel
configuration of magnetizations of RL and FL, (b) the two-dimensional map of the VDC signal:
examples of experimental resonance frequency (f0) angular dependence (open circles) are compared
with the theoretical prediction (black solid lines), (c) angular dependence of the amplitude of the
VDC signal: experimental data (black points) and theoretical dependence described by Eq.6. In-
set: example of the SD antisymmetric spectrum measured at an angle 45◦: f0 denotes resonance
frequency, the A is the experimental peak-to-peak amplitude of the VDC signal: the sign of A is
negative (positive) when the first extremum of the VDC spectrum is minimum (maximum).
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
12
As one can see the theoretical prediction agrees well with the experimental result. In
both cases the resonance frequency changes by 0.7 GHz while the external field rotates by
180◦.
Next, we used our model to compare the theoretical and experimental angular dependence
of the VDC signal amplitude. Based on the results presented in Ref.7, we can express the
VDC signal as an antisymmetric resonance curve:
VDC ≈ A sin 2θ(f 2 − f 2
0 )
0 )2 − σ2f 2
(f 2 − f 2
(6)
where f0 and σ denote the resonance frequency and resonance curve width (FWHM) re-
spectively, θ is the external magnetic field angle, and A stands for the amplitude of the VDC
signal given by:
A ≈ −η
γe
2
πIHOe∆R (HK + H + MS(Nz − Nx))
(7)
In the above Eq.(7) I and HOe are the amplitudes of the alternating current and associated
Oersted field, respectively, while Nz(x) denotes the demagnetizing factor in the z(x) direc-
tion. The phenomenological factor η was introduced in order to account for all possible
losses of microwave signal in the experimental setup11. Similarly as in the case of resonance
frequency calculations, we assumed the following values of magnetic parameters: saturation
magnetization µ0MS = 1.06 T , uniaxial anisotropy field Hk = 2.8 kA/m, demagnetizing
factors: Nz = 0.997311, Nx = 0.00246728 and Ny = 1 − (Nz + Nx).
According to the Eq.(6), the spin-diode signal follows sin 2θ dependence, what is also visible
in the case of experimental data shown in Fig.10(c).
IV. SUMMARY
The experimental setup for static and dynamic angular magnetic measurements has been
developed and presented. As a result of application of the quadrupole electromagnet with
a dedicated feedback control system, we have been able to generate an arbitrary in-plane
magnetic field vector. Moreover, in order to increase the area with the uniform magnetic field
vector, we optimized the shape of the electromagnet cores. Therefore we have eliminated a
possible magnetic measurement inaccuracy that may occur in a standard sample-rotating
setup. The experimental system was tested by measuring the angular dependence of the
resistance and spin diode effect in the spin-valve GMR micro-stripe. Both static (resistance)
and dynamic (spin-diode effect) angular dependencies were compared with the theoretical
predictions that confirmed the reliability of the presented measurement setup.
ACKNOWLEDGEMENT
We acknowledge the Polish National Center for Research and Development grant No.
LIDER/467/L-6/14/NCBR/2015.
P.O. acknowledges Dekaban Fund at the University of Michigan for the financial support.
1C. Kittel, "Interpretation of anomalous larmor frequencies in ferromagnetic resonance experiment," Phys-
ical Review 71, 270 (1947).
2A. Chumak, V. Vasyuchka, A. Serga, and B. Hillebrands, "Magnon spintronics," Nature Physics 11,
453–461 (2015).
3A. Tulapurkar, Y. Suzuki, A. Fukushima, H. Kubota, H. Maehara, K. Tsunekawa, D. Djayaprawira,
N. Watanabe, and S. Yuasa, "Spin-torque diode effect in magnetic tunnel junctions," Nature 438, 339–
342 (2005).
4M. Harder, Y. Gui, and C.-M. Hu, "Electrical detection of magnetization dynamics via spin rectification
effects," Physics Reports 661, 1–59 (2016).
5A. Yamaguchi, H. Miyajima, T. Ono, Y. Suzuki, S. Yuasa, A. Tulapurkar, and Y. Nakatani, "Rectification
of radio frequency current in ferromagnetic nanowire," Applied Physics Letters 90, 182507 (2007).
Biaxial magnetic-field setup for angular magnetic measurements of thin films and spintronic nanodevices
13
6S. Ziętek, P. Ogrodnik, W. Skowroński, F. Stobiecki, S. van Dijken, J. Barnaś, and T. Stobiecki, "Electric-
field tunable spin diode fmr in patterned pmn-pt/nife structures," Applied Physics Letters 109, 072406
(2016).
7S. Ziętek, P. Ogrodnik, M. Frankowski, J. Chęciński, P. Wiśniowski, W. Skowroński, J. Wrona, T. Sto-
biecki, A. Żywczak, and J. Barnaś, "Rectification of radio-frequency current in a giant-magnetoresistance
spin valve," Phys. Rev. B 91, 014430 (2015).
8S. Ziętek, P. Ogrodnik, W. Skowroński, P. Wiśniowski, M. Czapkiewicz, T. Stobiecki, and J. Barnaś,
"The influence of interlayer exchange coupling in giant-magnetoresistive devices on spin diode effect in
wide frequency range," Applied Physics Letters 107, 122410 (2015).
9J. Kleinlein, B. Ocker, and G. Schmidt, "Using giant magneto resistance stripes to efficiently generate
direct voltage signals from alternating current excitations," Appl. Phys. Lett. 104, 153507 (2014).
10W. Skowroński, M. Frankowski, J. Wrona, T. Stobiecki, P. Ogrodnik, and J. Barnaś, "Spin-torque diode
radio-frequency detector with voltage tuned resonance," Appl. Phys. Lett. 105, 072409 (2014).
11S. Ziętek, P. Ogrodnik, W. Skowroński, P. Wiśniowski, M. Czapkiewicz, T. Stobiecki, and J. Barnaś,
"The influence of interlayer exchange coupling in giant-magnetoresistive devices on spin diode effect in
wide frequency range," Applied Physics Letters 107, 122410 (2015).
12C. Wang, Y.-T. Cui, J. Sun, J. Katine, R. Buhrman, and D. Ralph, "Sensitivity of spin-torque diodes
for frequency-tunable resonant microwave detection," Journal of Applied Physics 106, 053905 (2009).
13X. Fan, R. Cao, T. Moriyama, W. Wang, H. Zhang, and J. Q. Xiao, "Magnetic tunnel junction based
microwave detector," Applied Physics Letters 95, 122501 (2009).
14M. Pufall, W. Rippard, S. Kaka, T. Silva, and S. Russek, "Frequency modulation of spin-transfer oscil-
lators," Applied Physics Letters 86, 082506 (2005).
15A. Yamaguchi, H. Miyajima, S. Kasai, and T. Ono, "Self-homodyne rf demodulator using a ferromagnetic
nanowire," Applied Physics Letters 90, 212505 (2007).
16S. Ziętek, J. Chęciński, M. Frankowski, W. Skowroński, and T. Stobiecki, "Electric-field tunable spin
waves in pmn-pt/nife heterostructure: experiment and micromagnetic simulations," Journal of Magnetism
and Magnetic Materials 428, 64–69 (2017).
17L. Steren, A. Barthelemy, J. Duvail, A. Fert, R. Morel, F. Petroff, P. Holody, R. Loloee, and P. Schroeder,
"Angular dependence of the giant magnetoresistance effect," Physical Review B 51, 292 (1995).
18J. Barnaś, O. Baksalary, and A. Fert, "Angular dependence of giant magnetoresistance in magnetic
multilayers," Physical Review B 56, 6079 (1997).
|
1801.01298 | 1 | 1801 | 2018-01-04T10:28:41 | Adsorption and binding dynamics of graphene-supported phospholipid membranes using the QCM-D technique | [
"physics.app-ph",
"cond-mat.mtrl-sci"
] | We report on the adsorption dynamics of phospholipid membranes on graphene-coated substrates using the quartz crystal microbalance with dissipation monitoring (QCM-D) technique. We compare the lipid vescle interaction and membranne formation on gold and silicon dioxide QCM crystal surfaces with their graphene oxide (GO) and reduced (r)GO coated counterparts, and report on the different lipid structures obtained. We establish graphene derivative coatings as support surfaces with tuneable hydrophobicity for the formation of controllable lipid structures. One structure of interest formed are lipid monolayer membrannes which were formed on rGO, which are otherwise challenging to produce. We also demonstrate and monitor biotin-avidin binding on such a membranne, which will then serve as a platform for a wide range of biosensing applications. The QCM-D technique could be extended to both fundamental studies and applications of other covalent and non-covalent interactions in 2-dimensional materials. | physics.app-ph | physics | Adsorption and binding dynamics of graphene-supported phospholipid membranes
using the QCM-D technique
D. A. Melendreza, T. Jowittb, M. Iliuta, A.F. Verrea, S. Goodwinc and A. Vijayaraghavan†a,c
We report on the adsorption dynamics of phospholipid membranes on graphene-coated substrates using the quartz crystal microbalance with dissipation
monitoring (QCM-D) technique. We compare the lipid vescle interaction and membranne formation on gold and silicon dioxide QCM crystal surfaces with their
graphene oxide (GO) and reduced (r)GO coated counterparts, and report on the different lipid structures obtained. We establish graphene derivative coatings
as support surfaces with tuneable hydrophobicity for the formation of controllable lipid structures. One structure of interest formed are lipid monolayer
membrannes which were formed on rGO, which are otherwise challenging to produce. We also demonstrate and monitor biotin-avidin binding on such a
membranne, which will then serve as a platform for a wide range of biosensing applications. The QCM-D technique could be extended to both fundamental
studies and applications of other covalent and non-covalent interactions in 2-dimensional materials.
a. School of Materials and National Graphene Institute, University of Manchester, Manchester M13 9PL UK.
b. Biomolecular Analysis Core Facility, Faculty of Life Sciences, University of Manchester, M13 9PL UK.
c. School of Computer Science, University of Manchester, Manchester M13 9PL UK.
† Corresponding author email: [email protected]
Introduction
Graphene is a versatile 2-dimensional (2-d) nanomaterial which has
attracted particular attention of scientists in the field of biological
sensors [1] and biotechnology [2]. Graphene's large surface area [3],
biocompatibility [4] and ease of functionalization [5], [6] provide
opportunities for biomedical applications [7], [8]. Graphene
derivatives can solubilize and bind drug molecules and thus have the
potential to be drug delivery vehicles [9]. Such properties can be
exploited to increase the sensitivity of biosensors [10] and may act
as a supporting platform for the construction of biological detection
arrays [11], [12]. Graphene oxide (GO) has played an important role
in the development of electrochemical [10], [13] and mass-sensitive
sensors [14], [15] and inclusively it has shown strong antibacterial
activity [16] by affecting the integrity of the cell membrane [17],
which is formed by a continuous lipid bilayer. In GO, the main surface
functional groups are hydroxyls and epoxies [18], with carboxylic
acids and other keto groups on the edges [19]. GO retains sp 2-carbon
domains as well as sp 3-carbon groups, endowing it with both
hydrophobic and hydrophilic domains, respectively [20]. A majority
of such functional groups can be removed upon treatment with a
reducing agent, such as L-ascorbic acid [21] or hydrazine [22], or
thermally [23] to form reduced (r)GO, which is overall significantly
more hydrophobic, comparable to pristine graphene.
Exploring different routes for the construction of supported lipid
membranes (SLMs) (Scheme 1a), such as supported lipid bilayers
(SLBs) or monolayers is pertinent since they are useful platforms for
the study of fundamental cell functions and signaling, cell-cell
interactions, drug delivery and biosensing [24]. SLMs help to
establish a model for the cell membrane and can serve as the key
component of biosensor devices. The most common membrane
lipids are phospholipids; amphiphilic molecules composed of
hydrocarbon chain tails attached to a phosphate head group.
Here we present the study of the adsorption and rupture of small
unilamellar vesicles (SUVs) (typically <100 nm) on GO- and rGO-
substrates, monitored using the quartz crystal microbalance with
dissipation monitoring (QCM-D) technique [25]. Substrates used
here were produced by spin coating GO on QCM-D crystals. This
coating technique is a fast and reliable way that renders a uniform
surface coverage with small quantities of GO dispersion (<100
µL/crystal) and serves as the first step to obtain an rGO film upon
thermal reduction of selected substrates. Moreover, both the
automation of sample injection and the real-time monitoring
(Q-Sense Pro, Biolin Scientific,
capabilities of
Gothenburg,
the experimental
the QCM-D
improve
Sweden)
system
Scheme 1. Schematic representation of lipid membranes from the basic
lipid unit. a) The three possible lipid structures studied b) Depiction (not
drawn to scale) of a lipid monolayer on reduced graphene oxide and intact
vesicles supported on graphene oxide on top of a Quartz Crystal
Microbalance chip c) DOPC molecular structure d) Biotinyl cap PE molecular
structure.
performance and simplifies the interpretation of the dynamics of
membrane formation.
The
interaction of graphene derivatives and phospholipid
membranes using different techniques, such as AFM, ellipsometry
and electrochemical cells has been previously reported [26]–[34] and
the results are promising in the development of highly versatile
biosensors. However, the formation of different lipid structures on
commonly used substrates as well as on graphene is still not
completely understood. Changing the physiochemical properties and
the topology of the substrate that acts as a mediator of the
interaction between the phospholipids and the surface is key in
establishing possible mechanistic scenarios for the adsorption and
spreading into specific structures (Scheme 1a-b). On this regard, the
different chemistries present in GO and rGO provide certain
hydrophilic and hydrophobic degrees and surface net charge which
favor the formation of specific lipid structures. The hydrophilicity of
Table 1 Mean values of wetting contact angles measured on the bare
and graphene coated SiO2 and Au chip surfaces.
SiO 2
Au
Bare
GO
rGO
17.34º ± 2.79º
41.32º ± 0.83º
32.12º ± 3.18º
38.12º ± 1.43º
88.44º ± 1.40º
94.74º ± 1.39º
hydrophobic interactions play in the support of lipid membranes
[26], [50]–[52]. One disadvantage of the growth of graphene
through the CVD technique and transfer onto a supporting substrate
is that it involves a harsh atmosphere. In this regard, the annealing
temperature to obtain an rGO film reported in this work is relatively
low (180 °C), in contrast to the high temperatures (> 900 °C) reached
for thermal reduction [20], [53]. In contrast, using graphene
dispersions, like GO, for the formation of thin film coatings offers a
fast route to achieve a surface with tunable hydrophobicity through
thermal reduction, which in consequence helps to recover most of
the properties of pristine graphene. Therefore, we propose a friendly
method to obtain graphene-based acoustic wave biosensors.
In this report, we present the adsorption dynamics of
zwitterionic lipid vesicles on two traditional substrates, SiO2 and Au,
whose hydrophobicity is modified by two graphene derivatives: GO
and rGO. We additionally investigated the utility of the formed lipid
membranes through a biomolecular binding event, specifically
between the biotin-avidin complex which is the strongest known
non-covalent biological interaction [54], [55] and is used for the
development of robust and highly sensitive assays useful in protein
detection [56], [57]. Non-covalent
interactions
involving π-systems are pivotal to the stabilization of proteins,
enzyme-drug complexes and functional nanomaterials [5], [58]. One
possible route to experimentally accomplish this binding event is by
presenting the avidin protein dispersed in buffer to lipid membranes
formed from biotinylated lipid vesicles that have been adsorbed, and
in some cases ruptured and reorganized on a supporting substrate.
intermolecular
2. Experimental methods
2.1 Solutions and reagents preparation
GO dispersions were prepared by oxidizing graphite flakes
according to a modified Hummers method [59] followed by
exfoliation and purification, as described in the electronic supporting
information (ESI).
Detailed protocol for the preparation and assembly of
phospholipid membranes is presented in the ESI. Briefly, the buffer
solution was prepared with 10 mM HEPES (4-(2-hydroxyethyl)-1-
piperazineethanesulfonic acid) buffer from powder (Sigma Aldrich),
100 mM NaCl, and 5 mM MgCl2, all then diluted in ultrapure water
(18.5 MΩ, MilliQ). The pH was corrected to 7.4, dropwise with a
solution of NaOH. This buffer solution was filtered with 0.22 𝜇m
pore-size nylon membranes before each experiment and stored in
the fridge for up to two weeks.
To prepare 1 mL of DOPC lipid vesicles 1 mL of 2.5 mg/mL DOPC
(1,2-dioleoyl-sn-glycero-3-phosphocholine) lipid (Avanti Polar Lipids)
dispersed in chloroform, was dried under a stream of nitrogen before
resuspending in 10 Mm HEPES. For the binding studies, 10 % biotinyl
cap PE lipid were mixed to the final concentration of 10% total lipid
in chloroform before drying. The functionalized lipid was hydrated
with the HEPES buffer solution and subsequently extruded more
than 23 times, as recommended by Cho's protocol [36], using 50 nm
pore-size polycarbonate membranes to obtain vesicles with a
diameter size distribution of ~80-110 nm, analyzed via dynamic light
scattering (DLS, Malvern Zetasizer Nano-S) (data not shown).
Fig. 1 Optical images of the drop shape obtained during contact angle
measurements on the bare and graphene coated SiO2 and Au chip surfaces.
SiO2 makes it the ideal surface for SLB formation via vesicle fusion
which on plain Au has been difficult to achieve using the same
technique. Therefore, finding adequate modifiers to obtain different
surface chemistries is crucial in the development of biocompatible
platforms. Accordingly, our investigation is primarily focused on the
study of biomimetic lipid membranes, namely, those that fully or
partially mimic the structure of the naturally occurring containment
unit of the cell as well as equivalent lipidic structures with potential
biomedical applications.
The formation of lipid membranes on bare and modified SiO2, Au,
mica and TiO2 substrates have been previously reported and
discussed in depth [35]–[38]. Promoting adsorption of SUVs onto a
substrate followed by spontaneous rupture and the formation of a
uniform membrane involves control over various parameters such
as: vesicle size [37], electrostatic force [36], ionic properties of the
buffer solution [39], [40], as well as vesicle-vesicle and vesicle-
substrate interactions [41]. The latter will strongly depend on the
properties of the substrate therefore varying the surface chemistry
will lead to different lipidic conformations which can be monitored
in real time with the help of acoustic wave sensors. Specifically, the
QCM system has emerged as the primary instrument for ultra-
sensitive mass detection ( <1 ng/cm2) [25]. The QCM-D technique
has drawn great interest as reported in studies on lipid adsorption
kinetics [36], [42], measuring vitamin-protein binding events [43],
[44], studying interactions in layer-by-layer graphene-lipid structures
[45] and formaldehyde detection using GO as a sensing layer [46].
Modifying the supporting substrate to obtain different lipidic
structures is one common approach, e.g. creating a hydrophobic
layer through the deposition of -thiol self-assembled monolayers
(SAMs) [41], using polymers as cushions [47] or using SAMs of
organosulfates and organophosphates [48] have been reported.
These modifiers are mostly used to provide specific functional groups
that confer hydrophilic or hydrophobic domains, to act as spacers for
further protein insertion into the membrane or to change the charge
of the substrate.
Graphene's high surface area, biocompatibility, ease of
functionalization and electrical properties makes it a great candidate
as a surface modifier with similar simplicity of SAMs [47]. On this
regard, GO is particularly interesting from its multiple functionalities.
A large amount of research is devoted to understand how lipids
interact with graphene to form well-defined lipidic structures and is
an emerging area of investigation. Recently, Tabaei et al., [49]
reported the formation of a lipid monolayer on pristine graphene
grown via the chemical vapor deposition (CVD) technique and then
transferred to a SiO2 coated QCM-D chip. Then, through the vesicle
fusion technique and via a solvent-assisted lipid bilayer method the
authors reported the formation of a supported layer of unruptured
vesicles and a lipid bilayer on oxidized CVD graphene, respectively.
Their findings are
line with previous reports which have
highlighted the pivotal role that both the hydrophilic and
in
Fig. 2 Frequency (purple, top) and dissipation (orange, bottom) response from the adsorption of DOPC on bare substrates: a) on bare SiO 2 and b) bare Au.
Steps: (A) DOPC injection, (B) buffer rinse, (C) SDS rinse, (D) final buffer rinse.
Avidin from egg white (Thermo Fischer Scientific) from a stock
concentration of 1mg/mL was diluted to 50 µg/mL. Finally, the
extruded lipid was aliquoted with a ratio of 1:10 (lipid:buffer) and
stored in the freezer at -20 ℃ for up to two weeks to ensure vesicle
stability.
2.2 QCM chips preparation and equipment setup
AT-cut piezoelectric quartz crystals from Q-Sense (Biolin
Scientific, Gothenburg, Sweden) with a fundamental frequency of
5 MHz were used in all experiments. SiO 2 and Au coated crystals
were first immersed in a 2% solution of sodium dodecyl sulfate (SDS)
and sonicated in an ultrasonic bath during 25 minutes, rinsed with
copious amounts of ultrapure water, then soaked in 99% ethanol for
10 min, then dried under a soft beam of nitrogen gas and
immediately placed inside a UV/Ozone system to be treated for 30
min. Crystals were spin coated at 3500 rpm during 2 min (Laurell
Tech., WS-650MZ) using the final dilution of GO (0.5 mg/mL).
For the thermal reduction of GO coated crystals to obtain an rGO
film, GO coated sensors were placed in the oven at 180 °C under
vacuum (Towson + Mercer, EV018) during 20 hours.
The Q-Sense Omega Auto (Q-Sense, Gothenburg, Sweden)
system was used to monitor the adsorption kinetics of lipid vesicles
on bare and coated sensors and record the frequencies (3rd to the 7th
harmonics) and dissipation values. We report the value for the 3rd
overtone for both frequency (∆𝑓𝑛=3/3) and energy dissipation
(∆𝐷𝑛=3/3) components. Values at other overtones can be found in
the ESI.
For the wetting properties analysis, a Kruss DSA100 (Hamburg)
system was used measure the wetting contact angles (WCAs) for the
bare, GO- and rGO-coated QCM chips. A manually controlled syringe
was used to cast sessile drops (~5 µL) of DI water on top of the
substrates under study. An ImageJ plugin for contact angle
measurement developed by Marco Brugnara [60] was used to
compute the WCAs (Table 1) through a manual ellipse/circle fitting
method.
3. Results and discussion
3.1 Membrane formation overview
Figure 1 shows the DI water droplets formed on the surface of
the QCM crystals (see ESI for extended wetting analysis). The
corresponding contact angles are summarized in Table 1. Preference
for intact vesicle adsorption and/or spreading into a mono- or bilayer
is
influenced by the hydrophobicity or hydrophilicity of the
deposition substrate. Removing organic contaminants by means of
chemical agents and UV/Ozone treatment is crucial to render a
is substantially more hydrophilic.
substrate hydrophilic [61]. On this regard, the formation of a SLB
requires a hydrophilic substrate, where a critical coverage of
adsorbed vesicles must be reached first, since vesicle-substrate
interactions usually do not commence individual vesicle rupture [36].
In particular, SiO2 has been the most widespread substrate for the
formation of SLBs and the adsorption dynamics is well understood.
On the other hand, the hydrophobic vesicle-substrate interaction has
been considered as the primary driving force in the formation of a
lipid monolayer [35]. With respect to gold, in an early study Smith
[62] stressed the effects of carbonaceous contamination on the
hydrophilicity of a clean gold substrate turning it hydrophobic. This
claim is supported by Gardner and Woods [63] who demonstrated
that when organic species are present, the surface of gold is
hydrophobic. Therefore, we stress the importance of an adequate
cleaning procedure of this sensor substrate to minimize undesirable
effects of contaminants present on the working electrode of the
QCM crystal. In spite of the fact that avoiding organic and inorganic
contamination on the surface of gold is a challenging task, our
analysis of the WCA (Table 1) shows that both the Au and SiO2 QCM
chips are hydrophilic as a result of our cleaning procedure, where the
SiO2substrate
In particular,
spreading of vesicles on Au is not straightforward since it has
produced conflicting results [35], [41], [64], [65]. Amongst the
properties of gold, we find biocompatibility, inert nature, affinity
with -thiol group and low electrostatic repulsion to zwitterionic lipids
making this noble metal a good candidate for the formation of SLMs,
therefore, whether clean Au promotes only vesicle adsorption or
adsorption followed by spontaneous rupture was investigated here
as a control and was also used as a support for the graphene
coatings. Finally, the adsorption of vesicles is influenced by the
presence of electrolytes and the ionic strength of the buffer solution.
Both parameters have shown to affect the lipid-substrate interaction
by means of the charge of the lipids head groups and the net
substrate charge [39], [52]. These characteristics altogether
constitute the main driving forces to promote vesicle adsorption and
eventual fusion of DOPC vesicles to form a uniform membrane [52].
We have investigated the effect of GO and rGO on the assembly
of different lipid membrane structures and in combination with the
QCM-D technique we proposed them as suitable platforms for the
detection of biomolecular
interactions. All experiments were
performed by following an adapted version of the protocol proposed
by Cho and coworkers [36] (for detailed protocol, see ESI). Bare
control QCM-D crystals were used in parallel on each of the reported
measurements. Results obtained from the control chips were
consistent throughout the experimental routines.
Fig. 3 Frequency (purple, top) and dissipation (orange, bottom) response from the adsorption of DOPC on a) GO-coated SiO 2 and b) GO-coated Au. Steps:
(A) DOPC injection and (B) buffer rinse.
3.2 SLMs on bare substrates
3.2.1 Formation on bare SiO2
First, we report the lipid adsorption kinetics on bare SiO2, as
shown in Figure 2(a). After lipid injection (step A), a high frequency
downshift indicates that vesicles are adsorbed intact on the surface
reaching a critical coverage before they break and spread into a
uniform membrane. Within less than 2 minutes the frequency
increases stabilizing at -26.65 Hz with a dissipation of 0.99 × 10-6
confirming the formation of a bilayer. This energy dissipation value
is somewhat higher than the typical reported values during the
formation of a SLB (< 0.5 × 10-6) [35], [42], [49], [61] denoting that
our bilayer is slightly less rigid than previous reports. We emphasize
that the observed fast rupture is due to the effect of the Mg2+ ions
on the reduction of critical coverage thus accelerating the phase of
vesicle fusion [39]. The adsorbed lipid remains stable even after a
buffer rinse (step B) confirming the formation of a rigid and uniform
membrane. After washing out the QCM chip with SDS solution (step
C), an increase in the energy dissipation to a median value of 8.52 ×
10-6 shows that some unabsorbed lipid was removed from the
surface. Finally, after rinsing with buffer (step D) the initial baseline
is recovered, indicating that the crystal has been fully cleansed. The
values for the frequency shift and stabilization throughout the
control experiment are in line with the results reported by Keller [35]
and Cho [36] on the formation of supported lipid bilayers on clean
SiO 2 and SiO, respectively.
3.2.2 Formation on bare Au
Similarly, we investigated the adsorption of DOPC vesicles on
clean Au. We point at the hydrophilic-hydrophilic interaction
between the lipid heads and the substrate as the main force in the
formation of a vesicular layer. The lower hydrophilicity from this
substrate does not favor any noticeable rupture and a monotonic
vesicular adsorption is obtained. The adsorption of intact vesicles on
bare Au is corroborated by the report by Liu and Chen [17] in which
a supported vesicular layer was required for the evaluation of the
rupture of vesicles in the presence of GO. It should be noted that in
both reports we used identical Au QCM chips (QSX-301). In addition,
the ionic strength from our buffer is not enough to promote vesicle
fusion and liposomes adsorbed intact without rupture. This is in
contrast with previous reports that have stressed the importance of
divalent ions, as Ca2+, as mediators in the formation of SLBs on gold
[36], [41] i.e. by increasing the deformation of DOPC lipid vesicles
[40]. Interestingly, Marques et al. [41] reported the inhibitory effect
of NaCl on the formation of lipid bilayers on gold, leading to tubular
structures. We indeed tested their experimental conditions (data not
shown) using a gold QCM substrate, preparing buffer solution
without NaCl and keeping the Mg2+ ions. However, the results did not
match, perhaps because both the lipid composition (binary lipid vs
single lipid) and surface topography differ between our studies. In
their study, Marques and coworkers annealed the gold substrate at
direct high temperature to obtain smooth micro-domains, while we
used the QCM crystals as- received, namely without any other
treatment than thorough cleaning, therefore keeping the inherent
roughness of the substrate. It has been established that controlling
the roughness of a substrate has direct impact on the structure of
the membrane to be deposited [66].
The event of intact vesicle adsorption is shown in Figure 2(b) as
a high frequency shift (the highest amongst all experiments) of -
164.27 Hz and an energy dissipation of 14.08 × 10-6, values
attributable to the size of vesicles suggesting the high mass loading
of the crystal with a non-homogeneous vesicular membrane that
releases considerable amounts of energy during its formation and
stabilization. In addition, we must consider the added mass of buffer
trapped within and between the vesicles. The high dissipation value
indicates that the ad layer is not rigid, rather viscoelastic. Finally,
after rinsing with buffer (step B) the slight decrease of ∆𝐷3/3, whilst
the frequency remains constant, shows that the vesicles that
conform the membrane are close packed and during this stage a
lateral shift occurs, as indicated by the energy dissipation variation,
with negligible loss or gain of mass. As mentioned before, we
highlight the role of the roughness of the Au surface to favor the
placement of intact vesicles. AFM topography of our Au QCM crystals
is included in the ESI and shows a less smooth surface compared to
the SiO2 substrates. In a report from Li et. al. [64] it was established
that surface roughness plays a pivotal role in the formation of SLBs
on a gold substrate through AFM studies on annealed gold
electrodes. During a flame annealing process, clean and large gold
grains with atomically flat terraces are produced which help in the
promotion of vesicle fusion after lipid deposition. Such atomically
smooth surfaces are not present in our crystals. Similar to our results,
they indicated having obtained unfused vesicles on rough gold QCM
surfaces.
As mentioned before, the electrostatic interaction plays a pivotal
role on the formation of SLBs. On this regard, previous reports
denoted that negatively charged intact vesicles adsorb onto a
titanium oxide or a gold substrate without spontaneous rupture to
form SLBs in the absence of divalent ions, like Ca2+ [36], [67]. In our
study, however, a similar situation occurs with a zwitterionic
(neutrally charged) single lipid on bare gold even under the effects of
Fig. 4 Frequency (purple, top) and dissipation (orange, bottom) response from the adsorption of DOPC on a) rGO-coated SiO 2 and b) rGO-coated Au. Steps:
(A) DOPC injection and (B) buffer rinse.
Mg2+ ion, which has shown similar strength in the promotion of
vesicle rupture [48]. In their protocol, Cho et. al. [36] stressed the
necessity for a negatively charged lipid in combination with calcium
ions for the formation of SLBs on titanium oxide. As a matter of fact,
they stated that their procedure is not suitable for forming single-
component zwitterionic SLBs (as DOPC) on titanium oxide, which
possesses similar biocompatibility and electrical properties to gold.
To overcome this limitation, they proposed and tested the use of an
amphipathic AH peptide as a vesicle-destabilizing agent which
successfully promoted vesicle deformation and subsequent rupture
into a uniform SLB. Overall, SLB formation demands specific surface
properties such as surface charge density and hydrophilicity.
Therefore, the use of specific agents, including surface modifiers, is
a common way to help the promotion of vesicle rupture and GO
naturally emerges as a good candidate due to its richness in
functional groups that confer this material its distinctive properties.
Accordingly, we discuss next the interaction between GO-coated
substrates and lipid vesicles.
3.3 SLMs on GO-coated substrates
According to the widely accepted Lerf-Klinowski model [20], GO
sheets have a high number of oxygen groups that confer the material
interesting physicochemical properties. Particularly, on the basal
plane hydroxyl (-OHs) and epoxy groups give GO its hydrophilic
nature [20]. In addition, ionized carboxyl groups (-COOH) at the
edges make GO sheets negatively charged [68]. We confirm the
presence of these groups in our GO through XPS (see ESI). Previous
studies have reported that lipid head groups control the interaction
between charged lipids and GO, showing that these have a strong
electrostatic interaction with the negatively charged carboxyl groups
of GO [28], [69]. Some studies consider that even van der Waals
forces might participate on the association between GO and lipids
[68]. Specifically, neutrally charged liposomes (as DOPC) associate
with the oxidized hydrophilic regions of GO sheets. In addition,
system hydration is believed to importantly contribute to the
interaction, since water molecules mediate the hydrogen bonding
between the carboxyl and the phosphate oxygen present in the lipid
headgroup [69]. In a recent work, Willems et. al. [70] have shown, via
coarse-grained molecular dynamics simulations, that preformed lipid
bilayers and inverted lipid monolayers supported on GO sheets when
immersed in water rapidly reorganize into bicelle-like structures.
Such rearrangement is thought to be driven by lipid headgroup
interactions and was explained by the polarity from the oxygen-
containing functional groups in GO. This finding highlights the effect
of hydration of the system, thus supporting the vastly accepted
hypothesis of the hydrogen bonding between the carboxyl groups of
GO and the phosphate head group from DOPC [71]. Willems' and
coworkers work preceded a similar study by Rivel et. al. [72] that
stressed the competition between the amphiphilic nature of the
phospholipid, the hydrophobicity of graphene (also present as
hydrophobic domains on GO sheets) and the hydrophilicity of water
during the formation of single and multilayer of lipids on graphene.
3.3.1 Formation on GO-SiO2
The formation of lipid membranes supported on GO-coated SiO 2
quartz crystals is shown in Figure 3(a). The frequency shift (∆𝑓3/3)
reaches an initial stabilization value of -68.31 Hz after vesicle
injection (step A). Similar to the previous result, we assume the
considerable mass uptake due to buffer trapped within the aqueous
phase of liposomes and the mass of buffer between them. The
energy dissipation value of 3.56 × 10-6 indicates the formation of a
soft membrane conformed by lipid vesicles which are strongly
adhered to the edges of the GO flakes and sparsely distributed on
their surface, as it has been previously proposed [69], [73]. From the
frequency dissipation response, vesicles adhere without noticeable
rupture, as a possible effect of bare SiO2 regions during the coating
procedure thus indicating a good coverage of the substrate. In this
regard, Furukawa et. al [74] reported that GO blocks the formation
of SLBs on a SiO2 substrate where GO flakes are present. This effect
is explained by the amphiphilic nature of this graphene derivative
due to the presence of both sp2 and sp3 domains. Additionally, it was
established by Frost et. al. [45], [75] that liposomes do not rupture
when adsorbing to GO. They found that rupture of vesicles is affected
by both the dimension of the GO flakes and the diameter of the
liposomes. They observed that liposomes fully rupture upon further
addition of GO flakes, after they have adsorbed to large GO sheets
(0.5-5 µm) obtaining multilayered structures of lipid membranes and
GO. Interestingly, their hypothesis is that for vesicle rupture to occur,
lipid vesicles must be exposed to two GO sheets, one on each side,
where GO sheets are of the same size or larger that the cross-
sectional area of the liposome. On this regard, it is evident that in our
investigation we exposed only one side of the liposomes to GO flakes
when they were adsorbed to the GO-coated substrate hence no
rupture would be expected. Our result is in good agreement with a
previous report on the adsorption of intact vesicles without fusion
on oxidized CVD graphene transferred to a SiO2 substrate [49],
however the adsorbed mass in our experiment was higher perhaps
due to the higher hydrophilicity of our GO coated crystals, promoting
decreases from a maximum of 5.71 × 10 − 6 to a value of 2.28 × 10 − 6,
before rinse. These factors indicate a considerable release of energy
during the initial stabilization stage, where vesicles stack then squash
and the weakly adsorbed lipid distributed on the surface detaches
after buffer rinse (step B), suggesting the gradual compaction of the
membrane and a spatial redistribution of the initial vesicular
membrane into a different lipidic structure. In contrast to the
previous result on GO-coated SiO 2 chip, the lower value of adsorbed
mass on the GO-coated Au chip after stabilization indicates that after
leaking buffer from within the aqueous phase, some ruptured
vesicles reorganize to more likely form bicelle-like structures, as
previously discussed from the results on GO in water by Willems and
coworkers. Therefore, our results point toward a mixed membrane
conformed of intact vesicles and bicelle-like islands.
Fig. 5 Frequency (purple, top) and dissipation (orange, bottom) response from the biotin-avidin binding event a) on a lipid bilayer supported on bare SiO 2
and b) on a vesicular lipid membrane supported on bare Au. Steps: (A) DOPC with biotin caps injection, (B) buffer rinse, (C) avidin injection, (D) final buffer
rinse.
then a higher attraction vesicles. Interestingly, our results are in
opposing direction to the findings of Okamoto et al. [73] who
reported the formation of single and double bilayers via vesicle
fusion on GO flakes supported on SiO2 in the presence of divalent
ions, like calcium. We point out at the topographic differences
between the substrates used in our studies and their effect on the
adsorption and fusion of lipid vesicles. In their study, SLBs were
formed by incubation of DOPC vesicles on a GO/SiO2/Si substrate and
analyzed using AFM. During the substrate preparation stage, large
blank SiO2 regions were still present after the deposition of GO
solution, while in our study we aimed to obtain a uniform coverage
of the substrate. Based on their results, the formation of bilayers and
double bilayers on GO sheets is explained as an effect of its surface
heterogeneity, however their findings are not conclusive on this
matter. As we discussed before, SiO2 is well known to promote SLB
formation due to its hydrophilic property, therefore the strong effect
that SiO2 has on the fusion of DOPC vesicles might be the leading
interaction in that scenario. We reason that after the vesicle rupture
events originated at the blank SiO2 regions, fragmented lipid is
rapidly attracted by and mobilized onto the regions covered with GO.
This hypothesis is supported by previous reports by Hirtz et. al. who
have shown the self-limiting spreading of DOPC on graphene [34],
[76] and, as previously discussed, lipids rapidly reorganize into
different
in aqueous media [70], [72]. The
contrasting homogenous hydrophilicity of SiO2 and the distributed
hydrophobicity from the sp2 domains from GO sheets might lead
such interaction. On this regard, the QCM-D technique excels the
AFM topographical analysis on the capability to monitor both
dynamically and in real-time the adsorption of lipids on specific
substrates. It is important to highlight the relevance of vesicular lipid
membranes for certain applications, e.g. where lipid vesicles are
used as drug carriers and the release of drugs trapped in the aqueous
phase must be time controlled through the addition of vesicle
destabilizers that promote rupture.
lipidic structures
3.3.2 Formation on GO-Au
On the other hand, on the GO-coated gold chip (Figure 3(b)),
vesicles are adsorbed intact after lipid injection (step A). The
frequency shift reaches a critical coverage point at a value of -53.75
Hz (arrow) indicating the formation of a membrane of intact vesicles
which in less than one minute is followed by partial vesicle rupture,
shown as a subsequent frequency increase and stabilization to a
value of -44.74 Hz before the buffer rinse step (B). After reaching the
critical coverage point, the energy dissipation exponentially
In both cases, lipid vesicles interact with a heterogeneous surface
chemistry present on the GO sheets on these GO-coated chips. We
reason that the presence of both hydrophilic/hydrophobic domains
establishes an equilibrium on the vesicle-substrate interaction forces
which might be altered by the differences on the intrinsic topography
of the substrates. Based on our WCAs (Table 1), the GO coating on
the smoother SiO2 substrate is not hydrophilic enough to promote
rupture like its bare counterpart while GO on gold, showing slightly
more hydrophilic regions than bare gold, leads the adsorption of
vesicles followed by partial rupture, as shown in Figure 3(b). In
addition, following the previously discussed hypothesis posed by
Frost and coworkers, such rupture might occur at specific sites where
some vesicles are partially wrapped by GO flakes present on some
valleys of the rough gold substrate.
In addition to the hydrophilic vesicle-substrate interaction, we
point at the electrostatic force between the negatively charged GO
regions and the dipole headgroup of the zwitterionic lipid as the main
driving forces for the adsorption of intact vesicles. From our results,
it is evident that the ionic strength of the buffer and the cation
bonding with the phosphate group of DOPC only promotes partial
vesicle rupture upon completion of the critical coverage which in the
case of the GO-SiO 2 (Figure 4(a)) the mass loss after rupture is
negligible in comparison to the desorbed lipid on the GO-Au chip
after partial rupture and stabilization.
3.4 SLMs on rGO-coated substrates
Removing the oxygen groups by the thermal reduction of GO
coated substrates effectively changes the surface chemistry leading
to a highly hydrophobic surface, as seen on the substantial increase
on the contact angle after the thermal treatment of the QCM-D
Fig. 6 Frequency (purple, top) and dissipation (orange, bottom) response from biotin-avidin binding event on a lipid monolayer supported a) on rGO-coated
SiO 2 and b) on rGO-coated Au. Steps: (A) DOPC with biotin caps injection, (B) buffer rinse, (C) avidin injection, (D) final buffer rinse.
sensor set (Table 1). Thus, a new group of measurements were
carried out to study the adsorption dynamics of DOPC vesicles on
rGO substrates.
Considering the variations between the initial and final values
obtained for the frequency shift in both samples, rGO-SiO2 and rGO-
Au, our results are within range of previous reports on the formation
of lipid monolayers on hydrophobic graphene [49].
3.4.1 Formation on rGO-SiO2
Figure 4(a) shows the monotonic response for the formation of a
lipid monolayer on rGO-coated SiO2 chip. After injecting the lipid
vesicles (step A), vesicle rupture and spreading into a monolayer
occurs. This distinctive adsorption and instantaneous rupture is due
to the interaction between the hydrophobic regions of rGO [77] and
the hydrophobic fatty acid chains from lipid tails. It has been
reported that in the formation of this type of membrane via vesicle
fusion on hydrophobic substrates such as a methyl-terminated SAM
[35] and CVD graphene [49] a frequency shift of around -13 Hz is
expected. In our case ∆𝑓3/3 varies from an initial stabilization value
of -15.76 Hz to a final frequency of -18.20 Hz with a dissipation of
2.55 × 10 − 6 before buffer rinse (step B). These values somewhat
differ from previous reports on the formation of uniform lipid
monolayers, however they point toward the formation on a non-
homogeneous monolayer. On this regard, we hypothesize that
individual lipid molecules are attracted during vesicle rupture to the
reduced GO sheets with higher hydrophobicity, forming small islands
that support an additional lipid membrane on top on them, as it was
pictured by Tsuzuki et.al. [27]. Furthermore, the higher mass uptake
can be attributed to a wetting film present in the interface. Interfacial
water layers have been observed under graphene membranes
adhered to sapphire substrate, uniformly trapping water that lifted
the edges of graphene sheets, in consequence adding more mass to
the sensor [78].We draw attention to the structural and chemical
differences between a transferred CVD graphene sheet and a
thermally reduced GO coating. While the first can be considered as a
highly crystalline film that might be mono- or few-layer graphene,
the second cannot be regarded as a single continuous film, rather a
group of overlapping sheets randomly arranged on the surface
creating multilayer graphene-like platelets that may preserve
different functionalities due to an imperfect thermal reduction
process.
3.4.2 Formation on rGO-Au
Similarly, on an rGO-coated Au chip (Figure 4(b)), after lipid
injection (step A) instantaneous rupture of vesicles occurs with an
initial ∆𝑓3/3 value of -13.48 Hz reaching a final frequency shift of -
17.70 Hz indicating the formation of a lipid monolayer membrane.
The final dissipation value of 2.85 × 10 − 6 shows that the membrane
has viscoelastic properties.
3.5 Graphene-SLMs as biomolecular interactions platforms
Finally, we examined the biomolecular interaction between the
Biotin-Avidin complex supported by lipid membranes formed on bare
and rGO-coated substrates. The biomolecular interaction associated
with a vesicular layer, as those from the GO-coated substrates, is out
of the scope of this study and has no biotechnical relevance from the
perspective of biomimetic membranes.
Our aim was to investigate the kinetics of the lipid adsorption and
binding event on the SLMs, especially on graphene due to the
similarities between rGO and pristine graphene. On this regard, Hirtz
el al. [79] have reported the assembly of inverted phospholipid
bilayers (where the hydrophobic tails are facing towards the
water/air media and the supporting layer holds the hydrophilic
heads) on exfoliated graphene in air, via the dip-pen nanolithography
technique. Interestingly, after immersion in buffer of the lipid bilayer
they observed a rearrangement
into a monolayer with the
hydrophilic headgroups facing outwards, more likely happening due
to a strong interaction between the hydrophobic surface of rGO and
the lipid tails. Obtaining right oriented lipid monolayers is crucial in
biomolecular studies in liquid media such as the insertion of
peripheral proteins or
the present biotin-avidin binding
measurement since the interaction can only occur with the biotin
molecules attached to the lipid heads.
We have obtained the experimental conditions for real-time
monitoring of the detection of biomolecular interactions that can
take place in biomimetic membranes supported on graphene
through the vesicle fusion technique for the formation of SLMs and
employing the QCM-D system. These experiments were performed
by following the same process described before and the only
difference is the use of 10% biotinylated DOPC lipid vesicles for the
formation of the lipid membranes. In addition, we included two extra
steps: the injection of avidin protein dispersed in HEPES followed by
a final rinse with clean buffer for the elimination of any residual lipid
or untied protein.
3.5.1 Avidin-biotin binding on bare SiO2 SLM
The first binding event was carried out on a lipid bilayer formed
on bare SiO2, shown in Figure 5(a). Initially, the formation of a
homogeneous lipid bilayer follows the same adsorption kinetics as
described before. After lipid injection (step A) followed by buffer
Table 2 Summary of results
Crystal
Coating
𝜟𝒇𝟑/𝟑 [Hz]
𝜟𝑫𝟑/𝟑 [× 𝟏𝟎−𝟔]
Stabilisation time
SiO2
Au
Bare
GO
rGO
Bare
GO
rGO
-26.73
-67.76
-18.20
-169.96
1.00
3.41
2.55
13.31
-53.75 -44.74
5.71 2.28
-17.70
2.85
10:15
10:00
9:53
10:12
10:24
14:58
Structure type
Lipid bilayer
Intact vesicles layer
Monolayer
Intact vesicles layer
Intact vesicles + Bicelle-like structures
Monolayer
rinse (step B), both values for the frequency and dissipation of -27.25
Hz and 0.63 × 10 − 6, respectively, validate the formation of the
bilayer. At step C, avidin injection is followed by an instantaneous
binding to the biotin caps attached to the lipid heads. A clean and
monotonic mass uptake after protein injection occurs, with a
frequency of -56.46 Hz. In addition, a low energy dissipation of
0.95 × 10 − 6 during binding (steps C-D) indicates that the attachment
of avidin molecules happens without altering the structure of the
previously formed rigid bilayer. Here the difference between the
stable frequency value before protein injection and the final
frequency at the binding event completion (step D) is ≈ 29 𝐻𝑧 for
the bilayer support.
3.5.2 Avidin-biotin binding on bare Au SLM
In contrast, the binding event supported on a bare Au chip
(Figure 5(b)) shows a frequency a value of -153.30 Hz at the end of
the stabilization region (steps B-C) after the adsorption of intact
vesicles (step A). Then, the frequency reaches a value of -217.79 Hz
after avidin injection (step C) with a final frequency value of -223.86
Hz before rinse. This variation corresponds to ≈ 70 𝐻𝑧. We validate
this high value from the surface area of the vesicular membrane,
where a myriad of biotin molecules is present at the outer layer of
each liposome for the protein to bind. The overtones considerably
diverged (see ESI) throughout the experiment indicating that there is
frequency dependence during the shear mode of the crystal. A high
dissipation value of 12.46 × 10 − 6 is reached after a slight peak during
the formation of the supported lipid vesicles while ∆𝑓3/3 varies
monotonically. This behavior
indicative of vesicle-vesicle
compaction prior to the formation of a non-homogeneous soft
membrane. In contrast to the vesicular lipid membrane formed on
bare Au (Figure 2(b)) the frequency shift during stabilization is lower
and no partial rupture occurs, perhaps due to the hydrophobic
nature of biotin [55] present in the lipid heads, changing the vesicle-
substrate interaction by the reduction of the hydrophilic attracting
force, and thus reducing the number of adsorbed vesicles. Finally,
after rinse (step D), the slight increase in both ∆𝑓3/3 and ∆𝐷3/3 to
values of -224.54 Hz and 8.42 × 10 −6, respectively, indicates that
some remaining unattached protein is dragged from saturated sites
to available biotin sites to attach, similar to the previous result.
is
3.5.3 Avidin-biotin binding on rGO-SiO2 SLM
Next, the lipid monolayer formed on rGO-SiO2 was used as a
supporting platform for the protein binding (Figure 6(a)). The
frequency dissipation before avidin injection (step A) stabilized at -
10.38 Hz. We consider this value within the lower limit for the
formation of a lipid monolayer on graphene, according to previous
reports [49]. The low energy dissipation value of 0.44 × 10 −6 with a
momentary increase to 0.68 × 10 −6 during the monolayer formation
(steps A-C), indicates that the membrane is being compacted to be
finally closely attached to the surface. The noticeable unsteady value
of ∆𝐷3/3 throughout the lifetime of the experiment (~38 min)
suggests that the membrane is settling onto the surface. However,
the uniformity of ∆𝑓3/3 throughout the stabilization period (steps B-
C) indicates that the membrane is rigid and the movement is parallel
to the shear force of the QCM-D sensor. After injection of avidin (step
C) the instantaneous frequency shift indicates a successful binding
event reaching an initial value of -39.07 Hz then stabilizing at -49.55
Hz after buffer rinse (step D). These values represent a variation of
≈ 40 𝐻𝑧.
3.5.4 Avidin-biotin binding on rGO-Au SLM
Finally, Figure 6(b) shows the adsorption kinetics of the binding
event on a lipid monolayer supported on a rGO-Au chip. At the end
of the membrane formation stage (steps A-C) the frequency
stabilizes to a value of -9.77 Hz which is close to the lower limit of the
formation of a monolayer. We reason that this low value of ∆𝑓3/3
indicates an incomplete coverage of the substrate and the formation
of rGO patches supporting islands of lipid monolayer. In this case, the
frequency variation is of ≈ 42 𝐻𝑧, which is slightly higher than the
values obtained for bilayer and monolayer on SiO2. This difference
can be explained from a structural point of view of the distribution
of the monolayer patches on the substrate. We consider that they
are spatially separated by the topography of the Au substrate,
leaving exposed biotin caps attached to the lipid heads that are
located at the margins of the lipid membrane, therefore increasing
the available binding sites, whereas for the monolayer and bilayer on
SiO2 the avidin binds to the biotin present in the superficial layer of
the continuous membranes. In addition, the
increase in the
dissipation after the binding event (step C) and buffer rinse (step D)
to a final value of 6.79 × 10 −6 shows that the membrane is not
homogeneous. This behavior might corroborate our hypothesis for
the structural distribution of lipid patches since the buffer flow hits
them generating lateral movement, therefore increasing the energy
dissipation.
Conclusions
We have described the construction and characterization of
graphene supported biomimetic lipid membranes and the
biomolecular interactions that can take place on different
substrates. We elucidated that both the chemical heterogeneity
of GO and the nature of supporting substrate lead to different
lipid structures obtaining either a membrane of intact vesicles
or a mixed layer composed of intact vesicles and ruptured
vesicles that reorganised into structures that resemble lipidic
bicelles.
lipid monolayers were successfully
obtained on both substrates coated with reduced GO, where
the hydrophobicity of the supporting material reached its
highest point. In general, the formation of the range of lipidic
structures presented on this work occurred in less than 15
minutes in terms of the time to become stable or for the time
the response showed a monotonic behaviour.
In contrast,
From a general standpoint, the variability of the responses
obtained for the
lipids and coated
substrates in the four cases under study that involved graphene
is appreciable as presented in Table 2, i.e., GO-SiO2, GO-Au,
interaction between
rGO-SiO2 and rGO-Au. This contrast of responses is explained
not only from the perspective of the lipid-graphene interface
but also from the effect that the underlying layer has on the
graphene support. The wetting transparency effect [80] has
been demonstrated in terms of the permeability of the
graphene film to the hydrophilicity/hydrophobicity of the
support substrate and the role that the interfacial water layer
plays, transferring these forces from the substrate to the
graphene surface [78]. This fact may explain with enough
consistency the differences on the formation of SLMs using the
same support substrate and attributable to the difference in
surface chemistries.
Graphene stands as an effective route for the chemical
modification of the selected substrates and as the underlying
platform for the development of a mass sensitive biosensor.
The interaction between lipids and graphene is strongly led by
electrostatic and hydrophobic interactions between them,
therefore is necessary to investigate the crucial experimental
parameters to obtain reproducible and defect free membranes.
The techniques for the formation of lipid monolayers have
evolved, moving from the Langmuir-Blodgett technique [81]
towards the vesicle fusion [35], [50] on substrates with inherent
or modified hydrophobicity. Performing the latter on graphene
sheets
is a straightforward option for the site-selective
formation of lipid monolayers due to the strong interaction
between the lipid tails and the graphene plane.
Overall, our results shed light on the interaction between
zwitterionic lipids and the dynamics of the physisorption to
graphene platforms for potential biotechnical applications.
Despite lipid monolayers do not resemble the complexity of
biological membranes due to their structural simplicity, they
have shown great utility on the evaluation of the interfacial
organization of lipid membrane constituents and the changes in
the interfacial organization upon the insertion of amphipathic
compounds [82] and due to their homogeneity, stability and
planar geometry lipid monolayers have been proposed over
bilayers as suitable models to characterize protein-membrane
interactions [83].
incorporation of amino acids,
Following the monolayer technique, lipid monolayers have
been used for the
like
antimicrobial peptides, or proteins, like cardiotoxins as their site
of actuation is at the cell membrane level, binding and
disrupting the outer membrane [83]. This specific affinity to
monolayers could potentially increase the sensitivity of the
binding detection in comparison to a bilayer membrane.
Is of our interest to use the proposed rGO-QCM-D platforms
as a biomimetic device to study the insertion and binding
mechanism of tail-anchored proteins and Odorant-Binding
Proteins (OBPs), a soluble protein secreted in the nasal mucus
of animal species and in the sensillar lymph of chemosensory
sensilla of insects. Our results will serve as the basis to achieve
such biosensing system. In the future, our work will be
undertaken to study the viscoelastic properties of the adsorbed
membranes and binding events using models such as the
Sauerbrey equation and the Voigt model to characterize the
thickness and mass of the ad layers.
Acknowledgements
DM acknowledges The National Council for Science and
Technology (CONACyT), Mexico for the financial support. AV,
AFV and SG acknowledge funding from the Engineering and
Physical
grants
EP/K016946/1 and EP/G03737X/1. The authors acknowledge E.
W. Hill and B. Grieve for helpful discussions.
Sciences Research
Council
(EPSRC)
Notes and references
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
X. Gan and H. Zhao, "A Review: Nanomaterials Applied in
Graphene-Based Electrochemical Biosensors," Sensors
Mater., vol. 27, no. 2, pp. 191–215, 2015.
J. Wang, Analytical electrochemistry. John Wiley & Sons,
2006.
A. Geim and K. Novoselov, "The rise of graphene," Nat.
Mater., 2007.
A. M. Pinto, I. C. Gonçalves, and F. D. Magalhães,
"Graphene-based materials biocompatibility: A review,"
Colloids Surfaces B Biointerfaces, vol. 111, pp. 188–202,
2013.
V. Georgakilas, M. Otyepka, A. B. Bourlinos, V. Chandra, N.
Kim, K. C. Kemp, P. Hobza, R. Zboril, and K. S. Kim,
"Functionalization of Graphene : Covalent and Non-
Covalent Approaches , Derivatives and Applications," 2012.
K. Yang, L. Feng, H. Hong, W. Cai, and Z. Liu, "Preparation
and functionalization of graphene nanocomposites for
biomedical applications.," Nat. Protoc., vol. 8, no. 12, pp.
2392–403, 2013.
H. Shen, L. Zhang, M. Liu, and Z. Zhang, "Biomedical
applications of graphene," Theranostics, vol. 2, no. 3, pp.
283–294, 2012.
C. Chung, Y. K. Kim, D. Shin, S. R. Ryoo, B. H. Hong, and D.
H. Min, "Biomedical applications of graphene and
graphene oxide," Acc. Chem. Res., vol. 46, no. 10, pp.
2211–2224, 2013.
K. S. Novoselov, V. I. Fal, L. Colombo, P. R. Gellert, M. G.
Schwab, K. Kim, and others, "A roadmap for graphene,"
Nature, vol. 490, no. 7419, pp. 192–200, 2012.
B. Unnikrishnan, S. Palanisamy, and S.-M. Chen, "A simple
electrochemical approach to fabricate a glucose biosensor
based on graphene-glucose oxidase biocomposite,"
Biosens. Bioelectron., vol. 39, no. 1, pp. 70–75, 2013.
P. Labroo and Y. Cui, "Graphene nano-ink biosensor arrays
on a microfluidic paper for multiplexed detection of
metabolites," Anal. Chim. Acta, vol. 813, pp. 90–96, 2014.
J. Choi, C. Lim, Y. Jung, D. Heon Shin, S. Bae, S. Kyung Kim,
and C. Kim, "A Microfluidic-Channel Regulated, Electrolyte-
Gated Graphene FET Biosensor Array for Repeatable and
Recalibrated Detection of Thrombin," Biophys. J., vol. 110,
no. 3 Supplement 1, p. 334a, 2016.
J. Li, D. Kuang, Y. Feng, F. Zhang, Z. Xu, and M. Liu, "A
graphene oxide-based electrochemical sensor for sensitive
determination of 4-nitrophenol," J. Hazard. Mater., vol.
201, pp. 250–259, 2012.
B. Cai, K. Hu, C. Li, J. Jin, and Y. Hu, "Bovine serum albumin
bioconjugated graphene oxide: Red blood cell adhesion
and hemolysis studied by QCM-D," Appl. Surf. Sci., vol. 356,
pp. 844–851, 2015.
Z. Yuan, H. Tai, Z. Ye, C. Liu, G. Xie, X. Du, and Y. Jiang,
"Novel highly sensitive QCM humidity sensor with low
hysteresis based on graphene oxide (GO)/poly
(ethyleneimine) layered film," Sensors Actuators B Chem.,
vol. 234, pp. 145–154, 2016.
Y. Tu, M. Lv, P. Xiu, T. Huynh, M. Zhang, M. Castelli, Z. Liu,
Q. Huang, C. Fan, H. Fang, and R. Zhou, "Destructive
extraction of phospholipids from Escherichia coli
membranes by graphene nanosheets," Nat Nano, vol. 8,
no. 8, pp. 594–601, 2013.
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
X. Liu and K. L. Chen, "Interactions of graphene oxide with
model cell membranes: Probing nanoparticle attachment
and lipid bilayer disruption," Langmuir, vol. 31, no. 44, pp.
12076–12086, 2015.
K. Erickson, R. Erni, Z. Lee, N. Alem, W. Gannett, and A.
Zettl, "Determination of the local chemical structure of
graphene oxide and reduced graphene oxide," Adv. Mater.,
vol. 22, no. 40, pp. 4467–4472, 2010.
J. P. Rourke, P. A. Pandey, J. J. Moore, M. Bates, I. A.
Kinloch, R. J. Young, and N. R. Wilson, "The real graphene
oxide revealed: Stripping the oxidative debris from the
graphene-like sheets," Angew. Chemie - Int. Ed., vol. 50,
no. 14, pp. 3173–3177, 2011.
D. R. Dreyer, S. Park, C. W. Bielawski, and R. S. Ruoff, "The
chemistry of graphene oxide," Chem. Soc. Rev., vol. 39, no.
1, pp. 228–240, 2010.
J. Zhang, H. Yang, G. Shen, P. Cheng, J. Zhang, and S. Guo,
"Reduction of graphene oxide via L-ascorbic acid.," Chem.
Commun., vol. 46, no. 7, pp. 1112–4, 2010.
S. Stankovich, D. A. Dikin, R. D. Piner, K. A. Kohlhaas, A.
Kleinhammes, Y. Jia, Y. Wu, S. T. Nguyen, and R. S. Ruoff,
"Synthesis of graphene-based nanosheets via chemical
reduction of exfoliated graphite oxide," Carbon N. Y., vol.
45, no. 7, pp. 1558–1565, 2007.
S. Pei and H. M. Cheng, "The reduction of graphene oxide,"
Carbon N. Y., vol. 50, no. 9, pp. 3210–3228, 2012.
S. G. Boxer, "Molecular transport and organization in
supported lipid membranes," Curr. Opin. Chem. Biol., vol.
4, no. 6, pp. 704–709, 2000.
[25] M. Rodahl, F. Höök, A. Krozer, P. Brzezinski, B. Kasemo, F.
H????k, A. Krozer, P. Brzezinski, and B. Kasemo, "Quartz
crystal microbalance setup for frequency and Q-factor
measurements in gaseous and liquid environments," Rev.
Sci. Instrum., vol. 66, no. 7, pp. 3924–3930, 1995.
X. Hu, H. Lei, X. Zhang, and Y. Zhang, "Strong hydrophobic
interaction between graphene oxide and supported lipid
bilayers revealed by AFM," Microsc. Res. Tech., vol. 79, no.
8, pp. 721–726, 2016.
K. Tsuzuki, Y. Okamoto, S. Iwasa, R. Ishikawa, a Sandhu,
and R. Tero, "Reduced Graphene Oxide as the Support for
Lipid Bilayer Membrane," J. Phys. Conf. Ser., vol. 352, p.
12016, 2012.
S. Li, A. J. Stein, A. Kruger, and R. M. Leblanc, "Head groups
of lipids govern the interaction and orientation between
graphene oxide and lipids," J. Phys. Chem. C, vol. 117, no.
31, pp. 16150–16158, 2013.
L. Wu, L. Zeng, and X. Jiang, "Revealing the nature of
interaction between graphene oxide and lipid membrane
by surface-enhanced infrared absorption spectroscopy," J.
Am. Chem. Soc., vol. 137, no. 32, pp. 10052–10055, 2015.
L. Rui, J. Liu, J. Li, Y. Weng, Y. Dou, B. Yuan, K. Yang, and Y.
Ma, "Reduced graphene oxide directed self-assembly of
phospholipid monolayers in liquid and gel phases,"
Biochim. Biophys. Acta - Biomembr., vol. 1848, no. 5, pp.
1203–1211, 2015.
S.-J. Liu, Q. Wen, L.-J. Tang, and J.-H. Jiang, "Phospholipid--
graphene nanoassembly as a fluorescence biosensor for
sensitive detection of phospholipase D activity," Anal.
Chem., vol. 84, no. 14, pp. 5944–5950, 2012.
[26]
[27]
[28]
[29]
[30]
[31]
[32] M. A. Ali, K. Kamil Reza, S. Srivastava, V. V. Agrawal, R.
John, and B. D. Malhotra, "Lipid-lipid interactions in
aminated reduced graphene oxide interface for biosensing
application," Langmuir, vol. 30, no. 14, pp. 4192–4201,
2014.
[33] M. Pittori, M. G. Santonicola, L. Ortolani, D. Gentili, V.
Morandi, and R. Rizzoli, "Graphene-lipids interaction:
Towards the fabrication of a novel sensor for biomedical
uses," IEEE-NANO 2015 - 15th Int. Conf. Nanotechnol., pp.
850–853, 2016.
[34] M. Hirtz, A. Oikonomou, N. Clark, Y.-J. Kim, H. Fuchs, and A.
[35]
[36]
[37]
[38]
[39]
Vijayaraghavan, "Self-limiting Multiplexed Assembly of
Lipid Membranes on Large-area Graphene Sensor Arrays,"
Nanoscale, p. (to be submitted), 2015.
C. A. Keller and B. Kasemo, "Surface specific kinetics of lipid
vesicle adsorption measured with a quartz crystal
microbalance," Biophys. J., vol. 75, no. 3, pp. 1397–1402,
1998.
N.-J. J. Cho, C. W. Frank, B. Kasemo, and F. Hook, "Quartz
crystal microbalance with dissipation monitoring of
supported lipid bilayers on various substrates," Nat.
Protoc., vol. 5, no. 6, pp. 1096–1106, May 2010.
E. Reimhult, F. Höök, and B. Kasemo, "Vesicle adsorption
on SiO2 and TiO2: Dependence on vesicle size," J. Chem.
Phys., vol. 117, no. 16, pp. 7401–7404, 2002.
R. P. Richter and A. R. Brisson, "Following the formation of
supported lipid bilayers on mica: a study combining AFM,
QCM-D, and ellipsometry.," Biophys. J., vol. 88, no. 5, pp.
3422–3433, 2005.
B. Seantier and B. Kasemo, "Influence of mono-and
divalent ions on the formation of supported phospholipid
bilayers via vesicle adsorption," Langmuir, vol. 25, no. 10,
pp. 5767–5772, 2009.
[40] M. Dacic, J. A. Jackman, S. Yorulmaz, V. P. Zhdanov, B.
Kasemo, and N. J. Cho, "Influence of divalent cations on
deformation and rupture of adsorbed lipid vesicles,"
Langmuir, vol. 32, no. 25, pp. 6486–6495, 2016.
J. T. Marquês, R. F. M. de Almeida, and A. S. Viana,
"Biomimetic membrane rafts stably supported on
unmodified gold," Soft Matter, vol. 8, p. 2007, 2012.
C. A. Keller, K. Glasmästar, V. P. Zhdanov, and B. Kasemo,
"Formation of supported membranes from vesicles," Phys.
Rev. Lett., vol. 84, no. 23, p. 5443, 2000.
S. Kim and S.-J. Choi, "A lipid-based method for the
preparation of a piezoelectric DNA biosensor," Anal.
Biochem., vol. 458, pp. 1–3, 2014.
[41]
[42]
[43]
[44] M. Kasper, L. Traxler, J. Salopek, H. Grabmayr, A. Ebner,
and F. Kienberger, "Broadband 120 MHz Impedance Quartz
Crystal Microbalance (QCM) with Calibrated Resistance
and Quantitative Dissipation for Biosensing Measurements
at Higher Harmonic Frequencies," Biosensors, vol. 6, no. 2,
p. 23, 2016.
R. Frost, G. E. Jo?nsson, D. Chakarov, S. Svedhem, and B.
Kasemo, "Graphene oxide and lipid membranes:
interactions and nanocomposite structures," Nano Lett.,
vol. 12, no. 7, pp. 3356–3362, 2012.
[45]
[46] M. Yang and J. He, "Graphene oxide as quartz crystal
microbalance sensing layers for detection of
formaldehyde," Sensors Actuators B Chem., vol. 228, pp.
486–490, 2016.
J. Munro and C. Frank, "In Situ Formation and
Characterization of Poly (ethylene glycol)-Supported Lipid
Bilayers on Gold …," Langmuir, no. 7, pp. 10567–10575,
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
2004.
J. Ekeroth, P. Konradsson, and F. Höök, "Bivalent-ion-
mediated vesicle adsorption and controlled supported
phospholipid bilayer formation on molecular phosphate
and sulfate layers on gold," Langmuir, vol. 18, no. 21, pp.
7923–7929, 2002.
S. R. Tabaei, W. B. Ng, S.-J. Cho, and N.-J. Cho, "Controlling
the Formation of Phospholipid Monolayer, Bilayer, and
Intact Vesicle Layer on Graphene," ACS Appl. Mater.
Interfaces, vol. 8, no. 18, pp. 11875–11880, 2016.
R. P. Richter, R. Bérat, and A. R. Brisson, "Formation of
solid-supported lipid bilayers: An integrated view,"
Langmuir, vol. 22, no. 8, pp. 3497–3505, 2006.
R. Tero, H. Watanabe, and T. Urisu, "Supported
phospholipid bilayer formation on hydrophilicity-controlled
silicon dioxide surfaces," Phys. Chem. Chem. Phys., vol. 8,
no. 33, p. 3885, 2006.
G. Edward Gnana Jothi, S. Kamatchi, and a.
Dhathathreyan, "Adsorption of DOPC vesicles on
hydrophobic substrates in the presence of electrolytes: A
QCM and reflectometry study," J. Chem. Sci., vol. 122, no.
3, pp. 341–348, 2010.
S. Pei and H. Cheng, "The reduction of graphene oxide,"
Carbon N. Y., vol. 50, no. 9, pp. 3210–3228, 2011.
H. Loughrey, M. B. Bally, and P. R. Cullis, "A non-covalent
method of attaching antibodies to liposomes," Biochim.
Biophys. Acta (BBA)-Biomembranes, vol. 901, no. 1, pp.
157–160, 1987.
R. P. Haugland and W. W. You, "Coupling of monoclonal
antibodies with biotin.," Methods Mol. Biol., vol. 45, pp.
223–233, 1995.
D. Shetty, J. K. Khedkar, K. M. Park, and K. Kim, "Can we
beat the biotin--avidin pair?: cucurbit [7] uril-based
ultrahigh affinity host--guest complexes and their
applications," Chem. Soc. Rev., vol. 44, no. 23, pp. 8747–
8761, 2015.
P. Vilja, K. Krohn, and P. Tuohimaa, "A rapid and sensitive
non-competitive avidin-biotin assay for lactoferrin," J.
Immunol. Methods, vol. 76, no. 1, pp. 73–83, 1985.
V. Singh, D. Joung, L. Zhai, S. Das, S. I. Khondaker, and S.
Seal, "Graphene based materials: Past, present and
future," Prog. Mater. Sci., vol. 56, no. 8, pp. 1178–1271,
2011.
[59] W. S. Hummers and R. E. Offeman, "Preparation of
Graphitic Oxide," J. Am. Chem. Soc., vol. 80, no. 6, p. 1339,
1958.
[60] Marco Brugnara (marco.brugnara at ing.unitn.it), "Contact
Angle Plugin," 2006. [Online]. Available:
https://imagej.nih.gov/ij/plugins/contact-angle.html.
[Accessed: 06-Oct-2017].
R. Richter, A. Mukhopadhyay, and A. Brisson, "Pathways of
lipid vesicle deposition on solid surfaces: a combined QCM-
D and AFM study.," Biophys. J., vol. 85, no. 5, pp. 3035–47,
2003.
T. Smith, "The hydrophilic nature of a clean gold surface,"
J. Colloid Interface Sci., vol. 75, no. 1, pp. 51–55, 1980.
R. Woods and P. Melbourne, "The hydrophilic nature of
gold and platinum," vol. 81, pp. 285–290, 1977.
[61]
[62]
[63]
[64] M. Li, M. Chen, E. Sheepwash, C. L. Brosseau, H. Li, B.
Pettinger, H. Gruler, and J. Lipkowski, "AFM Studies of
Solid-Supported Lipid Bilayers Formed at a Au ( 111 )
[65]
[66]
[67]
[68]
[69]
Electrode Surface Using Vesicle Fusion and a Combination
of Langmuir - Blodgett and Langmuir - Schaefer
Techniques," no. 111, pp. 10313–10323, 2008.
N.-J. Cho, C. W. Frank, B. Kasemo, and F. Höök, "Quartz
crystal microbalance with dissipation monitoring of
supported lipid bilayers on various substrates.," Nat.
Protoc., vol. 5, no. 6, pp. 1096–106, 2010.
I. Czolkos, A. Jesorka, and O. Orwar, "Molecular
phospholipid films on solid supports," Soft Matter, vol. 7,
no. 10, p. 4562, 2011.
N. Cho and C. W. Frank, "Fabrication of a Planar
Zwitterionic Lipid Bilayer on Titanium Oxide," vol. 26, no.
14, pp. 15706–15710, 2010.
A. C.-F. Ip, B. Liu, P.-J. J. Huang, and J. Liu, "Oxidation Level-
Dependent Zwitterionic Liposome Adsorption and Rupture
by Graphene-based Materials and Light-Induced Content
Release," Small, vol. 9, no. 7, pp. 1030–1035, 2013.
P.-J. J. J. J. Huang, F. Wang, and J. Liu, "Liposome/Graphene
Oxide Interaction Studied by Isothermal Titration
Calorimetry," Langmuir, vol. 32, no. 10, pp. 2458–2463,
2016.
[70] W. N, U. A, V. AF, I. M, L. M, H. M, V. A, and S. MS,
[71]
[72]
[73]
[74]
[75]
"Biomimetic Phospholipid Membrane Organization on
Graphene and Graphene Oxide Surfaces: a Molecular
Dynamics Simulation Study," 2017.
A. C.-F. F. Ip, B. Liu, P.-J. J. J. J. Huang, and J. Liu, "Oxidation
Level-Dependent Zwitterionic Liposome Adsorption and
Rupture by Graphene-based Materials and Light-Induced
Content Release," Small, vol. 9, no. 7, pp. 1030–1035,
2013.
S. O. Rivel, Thimothée Yesylevskyy and C. Ramseyer,
"Structures of single , double and triple layers of lipids
adsorbed on graphene : Insights from all-atom molecular
dynamics simulations," vol. 118, pp. 358–369, 2017.
Y. Okamoto, K. Tsuzuki, S. Iwasa, R. Ishikawa, a Sandhu,
and R. Tero, "Fabrication of Supported Lipid Bilayer on
Graphene Oxide," J. Phys. Conf. Ser., vol. 352, p. 12017,
2012.
K. Furukawa and H. Hibino, "Self-spreading of Supported
Lipid Bilayer on SiO2 Surface Bearing Graphene Oxide,"
Chem. Lett., vol. 41, no. 10, pp. 1259–1261, 2012.
R. Frost, S. Svedhem, C. Langhammer, and B. Kasemo,
"Graphene Oxide and Lipid Membranes: Size-Dependent
Interactions," Langmuir, vol. 32, no. 11, pp. 2708–2717,
2016.
[76] M. Hirtz, A. Oikonomou, T. Georgiou, H. Fuchs, and A.
[77]
[78]
Vijayaraghavan, "Multiplexed biomimetic lipid membranes
on graphene by dip-pen nanolithography," no. May, 2013.
O. C. Compton and S. T. Nguyen, "Graphene Oxide, Highly
Reduced Graphene Oxide, and Graphene: Versatile
Building Blocks for Carbon-Based Materials," small, vol. 6,
no. 6, pp. 711–723, 2010.
K. Yamazaki, S. Kunii, and T. Ogino, "Characterization of
interfaces between graphene films and support substrates
by observation of lipid membrane formation," J. Phys.
Chem. C, vol. 117, no. 37, pp. 18913–18918, 2013.
[79] M. Hirtz, A. Oikonomou, T. Georgiou, H. Fuchs, and A.
Vijayaraghavan, "DPN116-Multiplexed biomimetic lipid
membranes on graphene by dip-pen nanolithography.,"
Nat. Commun., vol. 4, no. May, p. 2591, 2013.
J. Rafiee, X. Mi, H. Gullapalli, A. V. Thomas, F. Yavari, Y. Shi,
[80]
[81]
P. M. Ajayan, and N. A. Koratkar, "Wetting transparency of
graphene," Nat. Mater., vol. 11, no. 3, pp. 217–222, 2012.
E. Sackmann, "Supported membranes: Scientific and
practical applications," Science (80-. )., vol. 271, no. 5245,
pp. 43–48, 2007.
[82] M. Eeman and M. Deleu, "From biological membranes to
biomimetic model membranes," Biotechnol. Agron. Soc.
Environ., vol. 14, no. 4, pp. 719–736, 2010.
R. Maget-Dana, The monolayer technique: A potent tool for
studying the interfacial properties of antimicrobial and
membrane-lytic peptides and their interactions with lipid
membranes, vol. 1462, no. 1–2. 1999.
[83]
Adsorption and binding dynamics of graphene-supported phospholipid membranes using the
QCM-D technique
D. A. Melendrez, T. Jowitt, M. Iliut, A.F. Verre, S. Goodwin and A. Vijayaraghavan
Supplementary Information
Experimental
Overview
The successful formation of uniform supported lipid membranes demands following a standardized procedure. Here, we
describe the experimental steps to prepare Small Unilamellar Vesicles (SUVs), to condition the QCM-D system, present them
to selected substrates and acquire the frequency and dissipation responses for further analysis.
Reagents
For Graphene Oxide (GO) preparation: Hydrogen Peroxide (H2O2) 30% (Sigma Aldrich), Sulphuric acid (H2SO4) 98% (Sigma
Aldrich), Sodium Nitrate (NaNO3) 98% (Alfa Aesar), Potassium Permanganate (KMnO4) 98% (Alfa Aesar)
For buffer preparation: Milli-Q water (>18 MΩ), 1,2-dioleoyl-sn- glycero-3-phosphocholine (DOPC) lipid (Avanti Polar
Lipids), Biotinyl Cap dispersed in Chloroform (10 mg/mL, Avanti Polar Lipids), Avidin protein from egg white (Sigma
Aldrich), Analytical grade HEPES buffer (Acros Organics), NaCl (powder, Sigma Aldrich), MgCl2 (powder, Sigma Aldrich),
NaOH (pellets, Fischer Scientific), H2O2 (30% solution), Ammonia (25% solution, Sigma Aldrich),
For cleaning: Sodium dodecyl sulphate (Fischer Scientific), Hellmanex II (Hellma Analytics).
Graphene oxide preparation
Graphene oxide used in this work was prepared according to a modified Hummers method described in ref. (1). Briefly,
graphite flakes of 50 mesh (1g) and NaNO3 (0.9 g) were mixed in concentrated H2SO4 (34 ml) in a round bottom flask and
kept overnight to intercalate. Then the mixture was cooled down in an ice bath and 4.5 g KMnO4 where added slowly under
constant stirring. The resulting mixture was left for 5 days at RT for graphite oxidation. After oxidation process was complete,
the resulting brown slurry was diluted at a slow rate with 100 ml H2SO4 solution of 5% after which 10 ml of H2O2 solution of
30% was added dropwise. Finally, the dispersion was further diluted with 100 ml mixture of H2SO4/H2O2 of 3%/0.5%. The
resulted graphite oxide was purified via centrifugation process by repeated washing with diluted H2SO4 and then DI water
until the pH of the supernatant was close to neutral. The homogenisation and complete exfoliation of graphene oxide was
performed using a vertical stirrer at a low speed for ~1h. The stock solution of GO (8.1 mg/mL) was diluted to a concentration
value of 0.5 mg/mL.
Buffer solution
The buffer solution is prepared diluting 10 mM HEPES, 100 mM NaCl, and 5mM MgCl2 in MilliQ water. The pH is adjusted
to 7.4 with a 1M NaOH solution when necessary. Stir this solution for at least 2 hours to ensure complete dissolution. To
increase the pH, add dropwise the sodium hydroxide solution during gentle stirring until a stable value is reached. Filter the
buffer with the 0.2 µm nylon membranes. Store the buffer in the fridge for up to two weeks.
Cleaning solutions
For cleaning the QCM-D system and quartz crystals, prepare both 2% SDS and Hellmanex II solution in MilliQ water. A
strong cleaning solution for gold crystals (QSX-301) is prepared as a 5:1:1 mixture of MilliQ water, Ammonia (25%) and
Hydrogen Peroxide.
Lipid vesicle preparation (and Biotin caps incorporation)
To obtain DOPC SUVs follow the next procedure.
Thoroughly rinse the inner walls of a 5-mL glass vial with chloroform using chloroform syringes. Dry the vial using a soft
beam of N2. Take 1 mL from DOPC lipid dispersed in chloroform (2.5 mg/mL) and pour it in the clean vial. Note: for Biotin
caps incorporation, take 25 µL of this vitamin dispersed in chloroform and mix it in the same vial. Dry the chloroform with a
soft beam of N2 until complete evaporation. Hydrate the lipid (/vitamin) with 1 mL of HEPES buffer solution. Fill a 1 mL
extruder syringe with the hydrated lipid (/vitamin). Place one 50 nm polycarbonate filtering membrane (Nalgene) at the middle
of the Teflon receptacle of the extruder, add two spacers per side and tightly close the hex nut. Insert another clean and empty
1 mL syringe on the opposite side of the Teflon receptacle. Extrude the dispersed lipid for at least 23 times. Be gentle to avoid
tearing the filtering membrane. It is recommended to use freshly made lipid vesicles to avoid vesicle aggregation.
Initial preparation of Quartz Crystals
These cleaning procedures are based on the protocol provided by QSense (2). The sensors used in this study are QCM with
gold surface (QSX-301) and with Silicon Dioxide (QSX-303)/Silicon Dioxide 300nm (QSX-318). A teflon QCM cleaning
holder (Q-Sense, QCLH 301) was used to prevent scratches on the surface.
SiO2 QCM chips
The following cleaning steps are regarded as mild cleaning and also applies for the gold crystals as a routine cleaning
procedure.
1. UV/ozone treat for 10 minutes.
2.
Immerse the sensor surfaces in the solution of 2% SDS and sonicate them for 15 minutes.
3. Rinse the sensors with abundant MilliQ water.
4.
5.
Immerse the sensors in MilliQ water and sonicate them for 15 minutes.
Immerse the sensors in 99% ethanol and soak them for 10 minutes.
6. Dry the surfaces using a mild beam of nitrogen gas.
7. UV/ozone treat for 25 minutes.
Au QCM chips
- Chemical treatment (Ammonium Peroxide Mix)
This cleaning process should be carried out under a fumes hood, wearing adequate PPE.
1. Using wash bottles, squirt the following solutions over the working electrode: 10% Decon 90, acetone and isopropanol.
Use DI water between each solution to rinse well the surface.
2. Treat the crystals under UV/ozone atmosphere for 10 minutes.
3. Heat the strong cleaning solution for gold to 75 º C.
4. Place the sensor in the heated solution for 5 minutes.
5. Rinse the sensors with MilliQ water. Keep the surfaces wet after ammonium-peroxide immersion until they are rinsed
well with water.
6. Dry with nitrogen gas.
7. UV/ozone treat for 25 minutes.
- Surfactant treatment
Follow the same cleaning steps for silicon dioxide sensors.
Coating QCM-D crystals with GO
In order to coat the Au/SiO2 crystals, the following methodology must be applied after completing the appropriate cleaning
procedure.
1. Configure the following parameters in the spin coating machine (SCM) (Laurell technologies Corp. WS-650MZ-23NPPB)
Speed = 3500 rpm, acceleration = 350 rpm/sec, time = 120 sec.
2. Place a QCM crystal in the vacuum nuzzle of the SCM.
3. Drop cast 70 μl of GO (0.5 mg/ml) on the surface of the Au/SiO2 QCM working electrode. Let the solution settle for 30
seconds.
4. Close the lid of the SCM and start the spinning.
5. Repeat steps 2-4 once for each set of crystals.
Thermal reduction of GO-coated crystals
This procedure requires to preheat the oven before cleaning and coating the desired number of sensors since the heating curve
from each oven may vary. In our case, the oven was set at 180ºC to preheat 1 hour before placing the chips to be reduced.
Follow the next procedure immediately after the spin coating of crystals is completed.
1. Distribute the selected chips to be reduced in petri dishes and label them accordingly.
2. When the over reaches 180 ºC place the petri dishes with the chips inside the chamber and close the door tightening
the screw to ensure good vacuum.
3. Leave the samples for 20 hours.
4. After 20 hours shut down the heater, turn off the vacuum pump and slowly turn the intake valve from the vacuum
oven to the open position. Note: to avoid blowing away the chips inside the oven, turn the valve gently until the
atmospheric pressure fills the chamber.
5. Let the chamber cool down for a few minutes and wearing heat gloves carefully take the QCM chips with Teflon
tweezers holding them from the edges.
6. Store the crystals in order inside holding boxes for a safe transport.
QCM-D measurement procedure
Initial system cleaning and priming
The following steps are intended to be applied on the Q-Sense Omega Auto (Biolin Scientific) system, which consists of 8
sensing ports automatically fed through customized scripts.
- Thorough ports and tubing cleaning
a. Load all ports with clean maintenance sensors. Note: verify the right position of the sensor matching the anchor symbol.
b. All the ports (1-8) must be initially washed by running 2% SDS at a flow rate1 of 25 𝜇L/min for at least 10 min. This step
should remove all remaining lipids and biological material from all tubing and syringes.
c. Rinse with system liquid2 for at least 15 min.
d. Flow 2% Hellmanex through the system for at least 10 min.
e. Finally rinse with system liquid for at least 15 min.
f. Remove the maintenance sensors, rinse the chamber with MilliQ water.
- Sensors & ports priming
Eliminating trapped bubbles is crucial to obtain a stable baseline and a steady response via continuous buffer flow.
a. Load the chamber with the desired number of sensors to be used. Up to four chips can be used for a parallel data acquisition.
Ensure that the electrodes are all dry during placement to avoid any variations during measurements.
b. Set the chamber temperature to 24 ºC to minimize thermal drift.
c. Run system liquid through the loaded ports until a stable baseline is noticeable. Note: despite this step can be programmed
to automatically stop when a stable baseline is reached it is recommended to run it manually to override the baseline criteria
from the system.
d. Vacuum ports and start running buffer solution through the working sensors for at least 5 min before the actual sample
injection and vesicle fusion technique is applied.
1 All flow rates are equal to 25 𝜇L/min unless specified.
2 System liquid refers to ultrapure water.
Formation of supported lipid membranes using the vesicle fusion technique
The aim of these steps is to present lipid vesicles to QCM-D sensors with different working-electrode surfaces by flowing
Small Unilamellar Vesicles (SUVs) dispersed in buffer solution. An initial vesicle-substrate interaction is expected to be
followed by a vesicle-vesicle interaction to obtain specific structures of lipid membranes. This procedure has been successfully
applied to obtain bilayers on clean hydrophilic substrates such as SiO2 and monolayer on modified Au (3). The adsorption and
formation of all the lipid membranes discussed in this work was accomplished by following the same experimental steps.
1. Deposit the extruded lipid vesicles dispersed in buffer in a 1.5 mL vial and place it in the right-hand rack and lock the lid
from the Omega Auto system. Note: in case of carrying out a binding event, also place in the rack a vial containing 100
µL of protein dispersed in chloroform, then dried and finally hydrated with 900 µL of HEPES buffer.
2. Run buffer solution through the desired chips for at least 5 min. Start recording the frequency and dissipation responses
from this step. Note: verify the stability and flatness of the baseline during this time. In case that some harmonics show
jumps and high variation this may indicate the presence of bubbles in the system and/or a bad interface between the
working electrode and the media. To solve this, redo steps c and d from the priming section.
3. After 5 min of stability inject the lipid vesicles (0.1 mg/mL) to the desired chips during 10 minutes. Note: The system
will indicate the quantity of lipid in buffer solution required to complete this step, however 1.0 mL should be enough to
run 4 parallel measurements with the same parameters described here. In case that the vial is not filled to the right level,
the system will automatically stop the execution of the script.
4. Verify the resonant frequency and dissipation values in liquid. Right after lipid injection a frequency shift must occur
showing some mass uptake and an increase in the energy dissipation. Depending on the type of membrane being formed
the frequency shift will stabilize to a specific value. Note: using a control chip is highly recommended to verify the
validity of the experimental procedure. A well-known adsorption kinetics is that for a bilayer on bare SiO2 where the
values from Fig. 2a (main text) are expected, with a tolerance of ∆𝑓 ± 1 Hz and ∆𝐷 ± 0.5 × 10−6.
5. Rinse the lipid layer with buffer for at least 5 min to remove any excess lipid and homogenize the membrane.
6.
If performing the binding measurement, inject the Avidin from the vial (after SLM stabilization) and let it settle under
continuous flow for at least 5 min.
7. Rinse all sensors with buffer to eliminate any excess protein and/or material deposited on the surface and to record
complete values for further analysis.
8. When the main body of the analytical script is completed, the system will run a wash routine. During these steps, all the
material present on the sensors will be removed and the surfaces, syringes and tubing will be washed using the selected
surfactants (Hellmanex and/or SDS) and finally rinsed with system liquid.
9. Upon completion, the door can be opened and the sensors can be taken out of the chamber.
10. It is recommended to leave the chamber clean and dry to be ready to use in subsequent experiments.
Samples characterization
The atomic force microscopy (AFM) of the GO and rGO was performed using a Bruker Dimension FastScan probe microscope
operating in taping mode. The tips used for the surface scanning were aluminium coated silicon FastScan-A tips from Bruker.
For coated crystals characterization, the diluted GO dispersion (0.5 mg/mL) was casted on clean QCM-D substrates and spin
coated as previously described.
The scanning electron microscopy (SEM) was performed on a SEM Zeiss Ultra setup, using an accelerating voltage of 5 kV.
The X-ray photoelectron spectroscopy (XPS) data were collected on a SPECS custom built system composed of a Phobios
150 hemispherical electron analyser with 1D detector. The X-ray source is a microfocus monochromated Al K-alpha
(1486.6eV) source. All spectra were collected with a pass energy of 20eV. Combined ultimate resolution as measured from
Ag 3d is 0.5eV with X-ray source and 20eV pass. The XPS data processing was done using CasaXPS software (version 2.3.16
PR 1.6). The C1s region peak fitting was done using Gaussian/Lorentzian shape components (for sp3 carbon) and asymmetric
shape components (for sp2 carbon) respectively. XPS C1s region was fitted with the synthetic components in the manner which
minimizes the total square error fit and corresponds to the literature reports. In the case of rGO, it was impossible to distinguish
between sp2 and sp3 carbons, therefore the signal was fitted with a single asymmetric component. The GO sample for XPS
was prepared by drop casting the dispersion on a clean Si/SiO2 (300nm) and drying in a vacuum oven to achieve a film
thickness not less than 10 nm. The rGO sample was prepared using the same conditions used for the reduction of GO on QCM
crystals. The GO vas first casted and dried on the Si/SiO2 (290 nm) substrate, followed by the reduction in vacuum at 180 ºC
for 20 hours.
Raman spectrum was taken on a Renishaw Raman system equipped with a Leica microscope and a CCD detector. Raman
spectrum was recorded using 532 nm laser line (Cobolt SambaTM continuous wave diode-pumped solid-state laser, 20 mW),
and the laser power was kept below 10 µW to avoid thermal degradation of the samples. 30 spectra per sample was taken. The
relative intensity ratio (𝑰𝑫 𝑰𝑮⁄ ) was measured from the averaged acquired mappings.
Results and discussion
Contact angle
Fig.S1 Contact angle sheet. a-f) Manual fit of the water droplet using the ImageJ plugin [ref]
The wetting contact angles for the range of QCM crystals is shown in Fig. S1. The manual circle-ellipse fittings were computed
using an ImageJ software plugin developed and published by Marco Brugnara (4) for such specific task. The software works
on pre-captured high contrast images of sessile drops which are processed by first inverting the image upside down, namely,
the water droplet must be pending from the top of the image, then two points are selected for the baseline of the droplet and
finally three edge points that follow the curvature of the droplet are selected. On each case, 5 readings were captured for
statistical effects and the results given by the script are shown in Tables S1 and S2. Table S1 shows the results for the SiO 2
crystal variations (Fig. a-c) while Table S2 shows the results obtained for the Au crystal variations (Fig. d-f). In both cases the
highlighted cells show the final average value for the ellipse fitting from which the standard deviation showed a lower value
than that obtained for the circle fitting.
Crystal type
Theta C
Uncertainty
Theta Left
Theta Right
Theta E
Circle StDev Ellipse StDev
15.8
14
14
13.3
15.1
14.44
0.1
0.1
0.2
0.1
0.2
0.14
20.7
16.6
18.6
14.9
21.8
18.52
18.8
13.5
14
13.4
21
16.14
19.8
15
16.3
14.2
21.4
8.56E-02
8.92E-04
1.16E-01
2.98E-03
2.03E-01
2.27E-03
1.06E-01
2.73E-03
2.35E-01
2.15E-03
17.34
1.49E-01
2.20E-03
SiO2 - bare
Averages
Crystal type
Theta C
Uncertainty
Theta Left
Theta Right
Theta E
Circle StDev Ellipse StDev
SiO2 - GO
25.3
24.8
24.2
24.9
25
0.3
0.3
0.2
0.1
0.2
38.3
35.4
30.8
30
25.3
Averages
24.84
0.22
31.96
34.4
34.7
33
27.7
31.7
32.3
36.4
35
31.9
28.8
28.5
4.31E-01
4.19E-04
5.50E-01
6.91E-04
3.40E-01
9.99E-04
2.38E-01
7.16E-04
3.72E-01
6.09E-04
32.12
3.86E-01
6.87E-04
Crystal type
Theta C
Uncertainty
Theta Left
Theta Right
Theta E
Circle StDev Ellipse StDev
82.6
82.8
78.8
79.5
80.8
80.9
0.6
0.4
0.4
0.3
0.5
89.2
88.5
86.3
87.2
87.9
92.6
89.2
89.2
86.2
88
90.9
88.8
87.8
86.7
88
8.19E-01
1.16E-04
5.10E-01
5.43E-04
4.90E-01
3.60E-04
3.64E-01
5.18E-04
6.14E-01
5.47E-04
0.44
87.82
89.04
88.44
5.59E-01
4.17E-04
SiO2-rGO
Averages
Table S1 Manual fitting results for SiO2 crystal set using the Contact Angle ImageJ plugin. Shadowed cell value is the final angle.
Crystal type
Theta C
Uncertainty
Theta Left
Theta Right
Theta E
Circle StDev Ellipse StDev
34
34.4
34.3
34.2
34
34.18
0.2
0.3
0.3
0.3
0.3
0.28
42.2
41.6
38.7
40.9
40.3
40.74
42.2
39.7
41.2
42.5
43.6
41.84
42.2
40.7
40
41.7
42
3.10E-01
3.86E-04
3.82E-01
4.88E-04
3.99E-01
8.96E-04
4.51E-01
7.15E-04
4.62E-01
4.44E-04
41.32
4.01E-01
5.86E-04
Au - bare
Averages
Crystal type
Theta C
Uncertainty
Theta Left
Theta Right
Theta E
Circle StDev Ellipse StDev
32.2
32.7
32.3
32.2
32
32.28
0.3
0.2
0.2
0.2
0.3
0.24
38.8
40.9
36.8
35
35.6
37.42
40
39.4
38.4
38.1
37.9
38.76
39.4
40.2
37.6
36.6
36.8
3.88E-01
5.83E-04
3.59E-01
7.46E-04
2.61E-01
8.18E-04
3.26E-01
8.88E-04
4.86E-01
4.74E-04
38.12
3.64E-01
7.02E-04
Au - GO
Averages
File Name
Theta C
Uncertainty
Theta Left
Theta Right
Theta E
Circle StDev Ellipse StDev
Au - rGO
87.4
87.2
87.2
88.1
86.7
0.3
0.4
0.4
0.3
0.3
Averages
87.32
0.34
94.6
97.1
95.7
94.9
92.2
94.9
93.9
96.1
94.4
95.8
92.6
94.2
96.6
95.1
95.4
92.4
4.74E-01
5.58E-04
6.32E-01
1.18E-04
5.23E-01
8.16E-05
5.06E-01
2.56E-04
4.11E-01
1.69E-04
94.56
94.74
5.09E-01
2.37E-04
Table S2 Manual fitting results for Au crystal set using the Contact Angle ImageJ plugin. Shadowed cell value is the final angle.
SEM images
Fig.S2 Scanning Electron Microscope (SEM) showing GO and rGO flakes arrangement on: a) on GO-SiO 2 b) rGO-SiO 2, c) GO-Au d)
rGO-Au. Probe voltage 5.00 kV
The SEM images of the SiO2 and Au QCM crystals coated with GO (Fig S2 a) and c)) show full coverage of the substrate with
flakes, with the number of layers (determined from the contrast and further AFM) ranging from single to few layers overlaps,
which is unavoidable when using spin coating deposition technique. The reduction of the GO (Fig S2 b) and d)) doesn't seem
to affect the substrate coverage and the flakes density. However, in case of Au substrate (Fig S2 d)) the rGO flakes present
many small holes (which is not an SEM artifact), unlike rGO present on SiO2. Considering the identical reduction conditions
for both samples, we speculate that the gas evolution during the GO reduction, combined with high temperature (180 ºC) could
have contributed to the Au etching which, in turn, contributed to the holes formation.
AFM images
Fig. S3 AFM mappings of full crystal set. Root-mean-square roughness (RRMS) and height profile values in nm scale are shown for: a) bare
SiO2, b) GO-SiO2, c) rGO-SiO2, d) bare Au, e) GO-Au and f) rGO-Au. Scan area in all images is 3µm2, g) shows the AFM image of GO
reference sample on Si/SiO2 (290 nm) wafer for a 30/30 um surface scan. The AFM height profile shows a thickness of ~1nm for single
layer flakes which increases almost proportionally with the number of flakes.
The topographical characterization of the prepared crystals was performed through Atomic Force Microscopy (AFM) to obtain
a height profile and values for the root-mean-square roughness (RRMS) for each crystal. The influence of the latter parameter
on the response of the QCM has been stressed in different studies, comprising from the variation between a modeled frequency
shift and experimental values of different RMS roughness levels (5), to the effect on the lipid-substrate interaction on the
formation of structurally different areas of the same lipid composition (6). It has been shown that surface roughness affects the
mechanisms of vesicle rupture and, in some cases, the formation of Supported Lipid Bilayers (SLBs) on solid supports (7),
however SLB formation is only slightly affected on the nanometer scale. Therefore, controlling the roughness of a surface has
direct impact on the structure of the membrane formed on top of the selected substrate. In fact, a rough crystal surface may
effectively damp more the response of the frequency shift than a smooth polished crystal.
The AFM images for three individual SiO2 and three Au crystals are presented in Fig S3. Each sample was carefully prepared
by following the same steps and under similar conditions, as described in experimental section. Because the AFM scan has
limited surface scan range, the information about the flakes distribution on the surface and the quality of the coverage are
provided mainly be the SEM images. As the SEM showed, both, SiO2 and Au QCM crystals are fully covered with GO/rGO
with very few small empty spots, and from number of GO/rGO layers ranging mainly from single to 3 layers. The monolayer
character of the original GO is confirmed by the reference sample (Fig S3 g)). However, it is difficult to ascribe in AFM the
exact position or the number of layers present on the SiO2 and Au QCMs substrates because of their high surface roughness
(Fig S3 a) and d)) and the tendency of GO/rGO sheets to flatten on the surface and take its shape (Fig S3 b), c), e) and f)).
Fig. S4 Large scan AFM mappings of full crystal set. a) bare SiO2, b) GO-SiO2, c) rGO-SiO2, d) bare Au, e) GO-Au and f) rGO-Au. Scan
area in all images is 90 µm2.
The only reference of the flakes presence is detected by their crumbles, overlaps and creases formed during the spin coating
and drying process, as can be seen clearly in lower resolution AFM images from Fig. S3.
As it can be seen from Fig S4 a), b) and c) the RRMS value of the surface increases with the addition of GO on the Si. A careful
inspection of the Si-GO, however, shows that the roughness coming from the SiO2 is slightly "smoothened" when the GO is
present. This can be ascribed to the higher thickness of the flakes given by the functional groups and the water molecules
trapped between the substrate and GO, between the GO flakes, and on the surface, due to the hydrophilic nature of GO. The
increased RRMS value is probably given by the contribution of the wrinkles, folds and overlaps of GO flakes to the existing
roughness. In case of Si-rGO substrate, the roughness of the substrate seems very similar to the bare Si. An explanation would
be the reduction in thickness of the GO flakes upon the reduction process accompanied by the dehydration. These, together
with the wrinkled nature of the rGO flakes, will contribute to a higher RRMS value compared to bare Si and GO.
The intrinsic higher roughness of the Au substrate (Fig S4 d)) doesn't change significantly with the addition of GO (Fig S4
e)). Unlike the case of SiO2, in this case the GO coated Au seems to keep the roughness characteristics and the only contribution
to the slightly increased RRMS value is the roughness generated by the flakes, at, however, lower rate than in case of SiO2. This
can be due to the difference in GO – substrate interaction, as well as more hydrophobic nature of Au which leads to a better
dehydration between GO and substrate. After the thermal reduction, the RRMS values for the Au-rGO (Fig S4 f)) are lower than
Au-GO and slightly higher than bare Au. A close look at the AFM scan (Fig. S4 f)) reveals that the deposited Au "islands"
present on Au-rGO have a more flat and uniform character compared to the initial Au substrate. This can be due to a slight Au
etching during the high temperature reduction of GO, which would explain lower roughness compared to Au-GO sample.
XPS on GO and rGO coated substrates
Fig. S5 XPS of coated samples a) wide scan GO, b) deconvoluted C1s of GO, c) wide scan rGO, d) deconvoluted C1s of rGO.
The XPS technique was performed to reveal the nature of chemical bonds in GO and to monitor their evolution after GO
reduction. Fig S5 a), b), c) and d) represent the wide scan and C1s spectra of GO and rGO respectively. The wide scan of GO
reveals a C to O ratio of ~2, in accordance with the literature for GO (8) with small amounts of nitrogen and Sulphur impurities.
After reduction (Fig. S5 c)) the C to O ratio increases significantly to 6. The C1s spectrum of the GO (Fig. S5 b) shows the
presence of different functional groups decorating the basal plane and the edges of GO: hydroxyl (C-OH) and epoxy (C-O-C)
groups between ~285 and 287 eV, carbonyl (C=O) and carboxyl (O-C=O) groups between ~287 and 289 eV, and finally, sp2
and sp3 carbons – around ~284 eV. After reduction (Fig S5 d)) the rGO presents fewer oxygen groups, i.e. single and double
carbon –oxygen groups with a binding energy of ~286 and 287 eV respectively, and an increased intensity sp2 carbon peak.
This proves the reduction of GO to rGO and a significant restoration of sp2 carbons.
Raman mappings
Fig S6. Raman spectra (with fits for G and D peak components) of (a) SiO2 surface with GO coating, (b) SiO2 surface with rGO coating, (c)
Au surface with GO coating and (d) Au surface with rGO coating. Green curves show peak fits and the red curves show the sum of the peak
fits (color online version).
Raman is a powerful technique used for the characterization of the graphitic materials, providing information about number
of layers, lattice defects, doping etc. (9,10). Fig S6 shows the Raman spectra of GO and rGO coated QCM sensors prepared
as described before. Peak fit is shown on the curves in green (online version). One of the spectral features of graphene is
associated with the optical phonon mode, which occurs around ~ 1580 cm − 1 and is called the G band (10). The D peak is
associated with defects in the structure (sp 3 bonding) appears at ~ 1350 cm − 1 (11). The relative intensity of D to G provides
an indicator for determining the in-plane crystallite size or the amount of disorder in the sample, indicating the sp2/sp3 carbon
ratio, ergo, it shows the disorder or the restoration of the graphene lattice (9,12).
Figures S6 a) and b) show the Raman spectra of GO and rGO on SiO 2-QCM-D sensors, respectively. The 𝐼𝐷 𝐼𝐺⁄ value of 0.96
suggests the presence of graphitic domains after the reduction process in SiO 2 (Fig. S6 b)) while the ratio obtained for GO is
equal to 0.93 (Fig. S6 a)). Similarly, Figures S6 c) and d) show the Raman spectra for GO and rGO, respectively, on Au-QCM-
D sensors. The 𝐼𝐷 𝐼𝐺⁄ ratio on Fig. S6 c) and d) remains equal according to our data fit, suggesting equivalent defectiveness
and the absence of any damage due to the reduction process on the scanned regions. Overall, these Raman spectra indicates
the presence of graphene and graphene-like domains on the selected substrates.
Overall frequency and dissipation values (3rd, 5th & 7th harmonics)
Bare substrates
a)
b)
Fig.S7Adsorption of DOPC a) on bare SiO 2 b) on bare Au. Steps: (i) injection of DOPC then (ii) buffer rinse, (iii) SDS wash, (iv) final
buffer rinse.
b)
GO coated substrates
a)
Fig.S8 Adsorption of DOPC a) on GO-SiO 2 b) on GO-Au. Steps: (i) injection of DOPC then (ii) buffer rinse
b)
rGO coated substrates
a)
Fig.S9 Adsorption of DOPC a) on rGO-SiO 2 b) on rGO-Au. Steps: (i) injection of DOPC then (ii) buffer rinse
Binding event on bare substrates
b)
a)
Fig.S10Biotin-Avidin binding event on a) a lipid bilayer on SiO 2 b) intact vesicles on Au. Steps: (i) injection of DOPC then (ii) buffer rinse
(iii) Avidin injection, (iv) final buffer rinse
Binding event on rGO-coated substrates
a)
b)
Fig.S11 Biotin-Avidin binding event on a) a lipid monolayer on rGO-SiO 2 b) a lipid monolayer on rGO-Au. Steps: (i) injection of DOPC
then (ii) buffer rinse (iii) Avidin injection, (iv) final buffer rinse
References
Rourke JP, Pandey PA, Moore JJ, Bates M, Kinloch IA, Young RJ, et al. The real graphene oxide
revealed: Stripping the oxidative debris from the graphene-like sheets. Angew Chemie - Int Ed.
2011;50(14):3173–7.
Scientific B. http://www.biolinscientific.com/q-sense/products/?card=QP1. 2015.
Keller CA, Kasemo B. Surface specific kinetics of lipid vesicle adsorption measured with a quartz
crystal microbalance. Biophys J [Internet]. 1998;75(3):1397–402. Available from:
http://dx.doi.org/10.1016/S0006-3495(98)74057-3
Marco Brugnara (marco.brugnara at ing.unitn.it). Contact Angle Plugin. 2006.
Rechendorff K, Hovgaard MB, Foss M, Besenbacher F. Influence of surface roughness on quartz crystal
microbalance measurements in liquids. J Appl Phys. 2007;101(11).
Yoon T-Y, Jeong C, Lee S-W, Kim JH, Choi MC, Kim S-J, et al. Topographic control of lipid-raft
reconstitution in model membranes. Nat Mater [Internet]. 2006;5(4):281–5. Available from:
http://www.ncbi.nlm.nih.gov/pubmed/16565710
Richter RP, Bérat R, Brisson AR. Formation of solid-supported lipid bilayers: An integrated view.
Langmuir. 2006;22(8):3497–505.
Yu H, Zhang B, Bulin C, Li R, Xing R. High-efficient Synthesis of Graphene Oxide Based on Improved
Hummers Method. Sci Rep [Internet]. 2016;6(1):36143. Available from:
http://www.nature.com/articles/srep36143
Dresselhaus MS, Jorio A, Saito R. Characterizing graphene, graphite, and carbon nanotubes by Raman
spectroscopy. Annu Rev Condens Matter Phys. 2010;1(1):89–108.
Childres I, Jauregui L, Park W, Cao H, Chen Y. Raman Spectroscopy of Graphene and Related
Materials. New Dev Phot Mater Res [Internet]. 2013;1–20. Available from:
https://www.physics.purdue.edu/quantum/files/Raman_Spectroscopy_of_Graphene_NOVA_Childres.p
df
Sobon G, Sotor J, Jagiello J, Kozinski R, Zdrojek M, Holdynski M, et al. Graphene oxide vs. reduced
graphene oxide as saturable absorbers for Er-doped passively mode-locked fiber laser. Opt Express.
2012;20(17):19463–73.
Iliut M, Leordean C, Canpean V, Teodorescu C-M, Astilean S. A new green{,} ascorbic acid-assisted
method for versatile synthesis of Au-graphene hybrids as efficient surface-enhanced Raman scattering
platforms. J Mater Chem C [Internet]. 2013;1(26):4094–104. Available from:
http://dx.doi.org/10.1039/C3TC30177J
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
|
1908.07017 | 1 | 1908 | 2019-07-24T20:15:08 | Optical deep learning nano-profilometry | [
"physics.app-ph",
"physics.optics"
] | Determining the dimensions of nanostructures is critical to ensuring the maximum performance of many geometry-sensitive nanoscale functional devices. However, accurate metrology at the nanoscale is difficult using optics-based methods due to the diffraction limit. In this article, we propose an optical nano-profilometry framework with convolutional neural networks, which can retrieve deep sub-wavelength geometrical profiles of nanostructures from their optical images or scattering spectra. The generality, efficiency, and accuracy of the proposed framework are validated by performing two different measurements on three distinct nanostructures. We believe this work may catalyze more explorations of optics-based nano-metrology with deep learning. | physics.app-ph | physics | OPTICAL DEEP LEARNING NANO-PROFILOMETRY
Jinlong Zhu1, †, Yanan Liu2, †, Sanyogita Purandare1, Jian-Ming Jin2, Shiyuan Liu3, *, and Lynford L. Goddard1, *
1Photonic Systems Laboratory, Department of ECE, UIUC, Urbana, IL, USA
2Center for Computational Electromagnetics, Department of ECE, UIUC, Urbana, IL, USA
3State Key Laboratory of Digital Manufacturing Equipment and Technology, HUST, Wuhan, China
†Equal contribution.
*Corresponding authors: [email protected]; [email protected].
ABSTRACT
Determining the dimensions of nanostructures is critical to ensuring the maximum performance of
many geometry-sensitive nanoscale functional devices. However, accurate metrology at the nanoscale
is difficult using optics-based methods due to the diffraction limit. In this article, we propose an
optical nano-profilometry framework with convolutional neural networks, which can retrieve deep
sub-wavelength geometrical profiles of nanostructures from their optical images or scattering spectra.
The generality, efficiency, and accuracy of the proposed framework are validated by performing two
different measurements on three distinct nanostructures. We believe this work may catalyze more
explorations of optics-based nano-metrology with deep learning.
1 Introduction
With the continuous development of innovative materials and advanced nanoscale fabrication techniques, a wide range
of functional nanostructures can be fabricated today to address various scientific and engineering challenges across
many fields, e.g., integrated photonic devices, metamaterials, and semiconductor transistors, to name a few. More
specific examples include nanoscale scatterers for tuning a whispering gallery-mode micro-toroid cavity to operate at
non-Hermitian spectral degeneracies [1], phase-gradient metasurfaces for controlling the propagation and coupling
of waveguide modes [2], and the sub-10-nm fin-based field-effect transistor and gate all around devices used in the
semiconductor industry [3]. However, as the aforementioned devices shrink in size and become more complicated in
shape, dimensional metrology becomes increasingly important because the high performance of nanoscale devices is
usually accompanied with an ultra-high sensitivity to their geometrical dimensions. For example, the gate-all-around
transistors at the 7-nm node and beyond may not work properly when their critical dimensions deviate by 20% from the
nominal design values. Hence, it is vital to accurately determine the geometrical dimensions of functional nanostructures
for quality control purposes and to ensure their maximum performance.
Transmission electron microscopy (TEM) and scanning electron microscopy (SEM) enable direct imaging of entire
nanostructures with single-digit nanometer resolutions due to the ultra-small de Broglie wavelength of electrons.
However, they suffer from destructive sample preparation and low integratability [4, 5]. Scanning tunneling microscopy
and atomic force microscopy could provide an even higher resolution than SEM, but are inherently inefficient due
to the requirement of serial scanning [6, 7]. Super-resolved fluorescence microscopy offers significant advantages
in particle position measurements by localizing the emitting fields [8, 9], but the artificially introduced fluorescence
dyes are incompatible with solid nanostructures. Recent developments in near-field related super-resolution imaging
techniques have shattered the diffraction limit by up to two orders of magnitude for certain objects [10, 11], but the high
complexity and ultra-small field-of-view make them impractical for many applications. Far-field-based techniques such
as optical profilometry and optical diffractive tomographic microscopy [12] are fast and can be easily integrated into a
fabrication chain, but they have limited transverse resolutions and are usually used to measure height variations only.
Optical scatterometry, which works by matching the experimental and simulated patterns using either a look-up-table or
nonlinear regression [13, 14], enables a nanoscale measurement accuracy, but the intrinsic parameter coupling may
hinder the accurate reconstruction of the critical dimensions in the devices, and yet there is no clue on the table to
conquer this drawback [15 -- 19]. Although there are efforts to improve optical scatterometry by introducing an artificial
neural network [20] and a support vector machine [15], the limitation of dealing with low-dimensional datum in
conventional machine learning makes it more suitable to the identification of nanoscale profile.
With the recent advances in computer vision and machine learning, convolutional neural networks (CNNs) have been
shown to excel at many image processing tasks such as classification [21], recognition [22], and segmentation [23].
More recently, CNNs have been applied to parameter estimation from astrology images [24] and resolution enhancement
and phase retrieval in bio-medical imaging [25 -- 31]. In this study, we propose a CNN-assisted optical nano-profilometry
framework that enables the non-destructive, non-contact, and accurate geometrical dimension reconstruction of
nanostructures with deep subwavelength features. Different from existing deep learning based super-resolution schemes,
we use CNNs to extract geometrical dimensions directly. We demonstrate the effectiveness and generality of the
proposed framework with two drastically different optical modalities, i.e., optical bright-field microscopy (BFM)
and ellipsometry; and three different measurands, i.e., a NIST RM 8820 [32] artifact, a nanoimprinted nanowire
array, and a single dynamic random-access memory (DRAM) transistor at an advanced technology node. These types
of nanostructures are used extensively in many fields including semiconductor industry, integrated photonics,
metamaterials, one-dimensional photonic crystals, biosensors, and neuromorphic chips. The results show that the
proposed framework can achieve dimension reconstruction of up to 1-nm scale accuracy with the field of view at 10,000
µm2 scale. Further, the reconstruction accuracy of individual dimensions can be controlled via tuning the weights in
the loss function of the CNN, offering a degree-of-freedom in investigating the most critical set of parameters that
govern the device performance. Such a feature in the method overcomes the issue of intrinsic parameter coupling
from deterministic reconstruction in conventional pattern-matching-based nanometrology techniques, where the critical
set of parameters may not be accurately reconstructed because of their coupling with influential but less significant
parameters in the inverse reconstruction process [33]. Although BFM and ellipsometry are our focus in this paper, we
expect the proposed framework to be capable of working with other modalities, e.g., dark-field [34], optical coherence
tomography [35, 36], diffractive tomography microscopy [37], and diffractive phase microscopy [38 -- 40], for nanoscale
metrology, provided that the corresponding imaging process can be numerically modeled for training data generation.
2 Methods
2.1 The proposed framework
The proposed CNN metrology framework is shown in Fig. 1a, which consists of four steps: training data generation,
CNN training, optical measurement, and dimension reconstruction. The first step is training data generation through
simulation. The simulation method has to match the measurement system and be able to model the physics of the
optical modality. In our examples, we use Fourier optics-based simulation [41] for the BFM, and rigorous coupled
wave analysis (RCWA) [42] for the ellipsometry. For data generation, we randomly sample the geometric space of the
measurand and use the appropriate simulation method to generate the image (for BFM) or the wavelength-resolved or
angle-resolved spectrum (for ellipsometry) for each data sample. The generated optical data are divided into the training
set for optimizing the CNN parameters, the validation set for model selection, and the test set for model evaluation.
The CNN models the inverse process of the physics-based simulation, where the input is the image or spectrum and
the output is the reconstructed dimensions. After the training is completed, the CNN model can be directly used with
measurement data, which are either raw intensity images or scattering spectra, obtained by a bright-field microscope or
a Mueller matrix ellipsometer (see more details in Sec. 2C), respectively.
2.2 CNN model training
The CNN model can be represented mathematically as a general composite function given by
y
F
,
x
f
1
1
,
f
2
,
2
f
3
...
f
n
n
,
x
...
(1)
where f1, f2, ..., fn are the layers of the network and can be operations such as convolution, pooling, batch normalization,
and dropout. Θ denotes the network parameters to be optimized. Our models follow the VGGNet [21] architecture, with
successive modules of convolution -- batch-normalization -- pooling -- dropout layers followed by fully-connected layers at
the output. An illustration is shown Fig. 1c whereas the details of the models can be found in Table S1 in Supplement 1.
The objective of training is given as
min
1
N
N M
i
1
n
1
yi
( )
n
y
i
( )
n
y
i
.
(2)
2
Figure 1: Schematic of the proposed optical deep learning nano-profilometry. (a) Simplified flowchart. (b) Schematic
of the in-house BFM. (c) Architecture of the CNN model. The inset at the top right corner of (b) presents the spectrum
of the 405-nm LED source with and without the bandpass filter. M, mirror; L, lens; P, polarizer; A, analyzer; PD,
photodetector; OBJ, objective; ID, iris diaphragm; MTS, motorized translation stage; F, filter; CCD, charge-coupled
device. Ports 1 and 2 connect the 405-nm LED and the 532-nm laser sources, respectively. Port 3 is used to capture the
spectrum of the source. The retractable rotating mirror M is utilized to guarantee only one source is selected in the
measurement. Only the 405-nm LED source is used here. The filter in the imaging space significantly filters the tails of
the LED spectrum.
3
n
( )
iy
( )n
iy
is the i-th reconstructed dimension of the n-th data sample and
Here
is the corresponding actual dimension.
N denotes the size of the training data set whereas M is the number of geometrical dimensions to be extracted. We
introduce ωyi here as the weight term associated with the i-th dimension. We call this cost function the weighted
mean squared error (WMSE). The weight term ωyi offers an extra degree of freedom in controlling the mapping
dynamics of the network and allows us to treat each geometrical dimension with its own criticality. Depending on
the measurement system used, the input to the network can be a 1-D, 2-D, or 3-D tensor, while the output is a vector
whose elements correspond to the dimensions to be estimated. For training, the network parameters are randomly
initialized following the method proposed in [43]. We then repeatedly sample a batch (batch size = 1024) of training
data and perform a gradient based optimization on Θ. We use the Adam optimizer [44] with β1 = 0.9, β2 = 0.999, and
the learning rate initially set to 1×10-3 and gradually decaying to 1×10-6. The CNN model and the training are
implemented with the Keras library in Python. The computing platform consists of an Intel Core i7-7700K CPU and
a Nvidia GTX-1080 GPU. We have also made our implementation and pre-trained models open-source on our
GitHub project page [45].
2.3 Measurement systems
We used two types of optical systems, i.e., BFM and ellipsometry, to validate the proposed framework in this paper.
The first instrument is an in-house epi-illumination microscope equipped with a fiber-coupled 405-nm light emitting
diode (LED) source (M405FP1, Thorlabs Inc.) and a CCD camera (C4742-80-12AG, Hamamatsu Inc.). The
magnification of the BFM system is 106.7×. An insertable bandpass filter (FB405-10, Thorlabs Inc.) centered at 405
nm and with a 10-nm full width at half maximum (FWHM) is used in front of the CCD to narrow down the spectrum
for accurate model mapping. A Zeiss plan-apochromat 20×/0.8 numerical aperture (NA) objective lens is used
together with the 5.33×4-f lens system (L4 and L5). Fourier optics is adequately accurate for the generation of
training sets for CNN. The detailed configuration of the BFM is illustrated in Fig. 1b. The second system is an
ellipsometer (ME-L Mueller matrix ellipsometer, Wuhan Eoptics Technology Co., Wuhan, China) that is
conventionally utilized to analyze the optical constants of thin films within the wavelength range of 193 -- 1690 nm. The
instrument can also be operated in incident angle-resolved and azimuthal angle-resolved modes. In the Results and
Discussion section, we will discuss in detail the operation modes we used in the ellipsometry examples.
2.4 Sample preparation and measurement schemes
The samples under investigation include the silicon fins on a NIST RM 8820 artifact [32], a nanowire array fabricated
by nanoimprinting, and a DRAM transistor consisting of three-layer nanostructures [Si3N4, SiO2, and (100)-orientation
Si, from top to bottom] with several features that are smaller than 10 nm. Figure 2d shows the SEM image of the
investigated pattern (pattern G) on the NIST artifact. The geometrical dimensions of the silicon fins and the material
constitutions can be found in [32]. The nanoimprinted wires have a similar shape as that of the NIST artifact, but
with a transverse dimension that is twice as small and a much taller height. See Fig. 3b. The DRAM transistor was
manufactured by a standard 45-nm node process. Because of the difference in etching anisotropy between Si3N4 and
Si, the top and bottom trapezoids have different sidewall angles [see Fig. 4b obtained by a TEM (TE20, TEM.FEI Co.)].
This asymmetry needs to be taken into consideration in the simulation when generating training data for the CNN.
3 Results and Discussion
3.1 BFM-based nanometrology
The first optical modality we used with the proposed framework is BFM. The measurand under consideration is a NIST
RM 8820 artifact, with an array of short lines characterized by three geometrical dimensions, namely the bottom
width (BCD), the height (HGT), and sidewall angle (SWA) (see Fig. 2d for a cross-sectional view). For data
generation, we take 2,400 random samples from BCD~U (1000, 1100) nm, HGT~U (80, 100) nm, and SWA~U (82,
90)°, where U (a, b) refers to the uniform probability distribution on the interval [a, b]. We divide this into groups of
2000, 200, and 200 samples for the training set, validation set, and test set, respectively. We use Fourier optics-
based simulation at λ = 405 nm to obtain the microscopy image of the samples. Note that the spatial resolution of
the simulation must match that of the BFM measurement system, which in our case is 101×101. We take the average
of all the horizontal slices of the experimental image to alleviate perturbations due to measurement noise, dust on the
sample, and imperfect sources, and use the average slice as the input to the CNN.
The CNN is trained for 3000 epochs in around 3 minutes; see Fig. S1a in Supplement 1 for the training curve. After
training, we first test its performance on the simulation data. Figures 2a -- c show the results on the test set, where the x-
and y-axes are the nominal and reconstructed dimensions, respectively. The blue dots correspond to the 200 samples in
the test set. The red lines serve as a guide to the eye for perfect reconstruction. We also inspect the results quantitatively
4
with the average bias B and standard deviation σ, defined as the mean and standard deviation of Dreal - Drecon, where
Dreal and Drecon are the actual and reconstructed dimensions, respectively. The average bias and standard deviation
are at sub-nanometer and sub-degree scales, indicating the good reconstruction accuracy of the proposed method; see
the values in the insets of Figs. 2a -- c. Note that the longitudinal dimensions (HGT and SWA) cannot be directly
determined from a single-shot optical image in conventional BFM. However, we are able to recover these dimensions
here because the intensity distribution of the BFM image is a nonlinear function of all the geometrical dimensions of
the nanostructures. In particular, the BFM image contains the diffraction patterns from light scattering at the edges of
the nanostructures and our CNN method can identify these patterns in the images with sufficient training. We also want
to point out that once trained, the parameter extraction (inference) step is a single feed-forward process in the neural
network computation, which is very fast. It takes less than 10 milliseconds for the 200 samples in the test set.
Figure 2: Reconstruction for the NIST RM 8820 pattern. (a -- c) The extracted geometrical dimensions for BCD, HGT,
and SWA for 200 testing samples (simulation). The average bias (M) and the standard derivation of the bias (σ) are
marked in each plot. (d), (e) SEM and BFM top-view images of the investigated area. The SEM images are adapted
from [32]. (f) Cross-sectional intensity curve across the three device lines within the red cropped window of (e) for
the measured image and the image calculated from the CNN output. The measured curve in (f) averages all of the
101-pixel wide horizontal slices, such as the white dotted line in (e), for different y-positions. The inset in (d) is the
SEM side-view image of a similar sample showing the three dimensions: BCD, HGT, and SWA.
We can now evaluate the performance of our CNN model for retrieving dimensions from experimental BFM images.
The measured BFM image contains three periods of lines, which are chosen in such a way that the sample stage induced
tilting is minimal, as determined by the regularized pseudo-phase imaging method [46]. For the actual measurand
and the measured BFM image, the CNN-reconstruction has an output of BCD = 1.00 µm, HGT = 91.43 nm and
SWA = 90° (see one of the measured images in Fig. 2e as an example). Compared with the NIST [32] measured
dimensions of 1.00 µm, 97.3 nm and 88° (the SWA was measured from a similar silicon nanowire but with a larger
height-width-ratio [32]), our model has a bias of 5.9 nm and 2.0° for HGT and SWA , respectively, and a perfect
reconstruction for BCD. The likely sources for the mismatch are systematic BFM measurement errors, e.g., our inability
to place the sample at the focal plane with an accuracy better than 50 nm, and random BFM measurement noise. Using
the reconstructed dimensions, we simulate the expected image compare it with the BFM image of the actual measurand
(horizontal slice only) in Fig. 2f.
5
3.2 Ellipsometry-based nanometrology
The second modality we worked with is ellipsometry. Ellipsometry measures the polarization-sensitive scattering
spectrum at either multi-wavelengths or multi-angles (incident or azimuthal angle), which is completely different from
BFM in terms of the measurands and operation modes. We use an ellipsometer operating in the specular reflection
mode (see Fig. 3a). The incident angle θ and the azimuthal angle φ can be tuned continuously via a rotation stage and a
single-axis rotation arm (not shown). The source can be either a laser or a broadband source, depending on the detector
in use. We first operate the ellipsometer in angle-resolved mode: we scan both θ and φ and measure the scattering
spectrum versus wavelength for each angle combination. Here, we select five incident angles (in the range 45° -- 65°; 5°
increment) and nineteen azimuthal angles (in the range 0° -- 90°; 5° increment) to obtain the 5× 19× 15 spectra, where
the last dimension of size 15 comes from the 15 elements of the normalized Mueller matrix; see the visualization in Fig.
3c. Data is simultaneously collected at ten wavelengths from 350 nm to 800 nm with a 50-nm increment. The 4-D
data at each wavelength is used to train an individual CNN. Because the datum at different wavelengths are measured
independently by different pixels in the spectrometer of the ellipsometer, the measured data points in the spectrum
are uncorrelated. Thus, we can average the output dimensions of the CNN models at various wavelengths to improve
the overall accuracy. The measurand under investigation is a silicon nanowire, characterized by three geometrical
dimensions, TCD, HGT and SWA (see Fig. 3b). We sample from TCD~U(300, 400) nm, HGT~U(430, 530) nm, and
SWA~U(82, 90)°, to get 3200 samples and use the RCWA algorithm to obtain the corresponding scattering spectra.
The data are divided into 2500 for training, 500 for validation, and 200 for test. For our ten CNN models, each model is
trained for 2500 epochs in around 4 minutes; see Fig. S1b in Supplement 1 for the training curve of a typical model.
The results on the simulation data are presented in Fig. 3d, where we have three sets of box plots showing the relative
percentage errors for the reconstructed TCD, HGT, and SWA on the test set for the ten models trained at different
wavelengths. We are able to achieve less than 0.3% errors with each individual model. Figure 3e shows the experimental
results. We measure the nanowire at the same ten wavelengths and feed the scattering data to the corresponding CNNs.
The dimensions measured by SEM (dotted lines) are also shown within each sub-figure. The dimensions reconstructed
by CNN at each single wavelength are all very close to the SEM measured values of TCD = 350 nm, HGT = 472 nm,
and SWA = 88°, while the average reconstruction over the ten models is TCD = 350.4 nm, HGT = 471.6 nm, and SWA =
87.4°, corresponding to an accuracy of ∆TCD = 0.4 nm (0.11%), ∆HttT = 0.4 nm (0.08%), and ∆SWA = 0.6° (0.68%).
Here, we should mention that during the measurement, there are uncertainties caused by the positioning errors of the
mechanical components (such as the RS and the single-axis rotation arm) and the non-zero spectral bandwidth of the
scattering signal captured by each pixel in the spectrometer (the detector), which could degrade the reconstruction
accuracy of the CNN. We expect the reconstruction accuracy of the proposed method can be further improved by
reducing measurement errors, which can be achieved by using opto-electronic components with better performance and
better system calibration using automatic platforms.
To further validate the generality of the proposed method, we fix the incident and azimuthal angles at 65° and 0°,
respectively, while only capturing the wavelength-resolved scattering spectrum. The source has an operating wavelength
in the range of 200-800 nm with a 10-nm increment, and we use the resulting 61× 15 spectra as the input to the CNN.
We investigate a DRAM transistor, whose geometrical profile consists of three parameters D1, H1, H2, H3, SW A1,
and SW A2; see the schematic and the TEM measured cross-section in Figs. 4a and 4b, respectively. We sample from
D1~U(50, 100) nm, H1~U(110, 160) nm, H2~U(3, 23) nm, H3~U(110, 160) nm, SWA1~U(82, 90)°, and SWA2 U (82,
90)°, to get 3,200 samples and use the RCWA algorithm to obtain the scattering spectra. Among all the data, 2500
are used for training, 500 are used for model selection, and 200 for test.
The CNN is trained for 2,500 epochs in around 4 minutes; see Fig. S1c in Supplement 1 for the training curve. Figures
4c -- e show the performance of the CNN on the test set of the simulation data for D1, H2, and SWA1, which are the most
critical dimensions governing the performance of the DRAM transistor. We quantify the accuracy of our predictions by
calculating the intervals containing 68% and 95% of the reconstructed dimensions from their nominal values, indicated
by the orange and black lines in these plots. Similar to Figs. 2a -- c, the horizontal and vertical axes correspond to the
nominal and reconstructed dimensions, respectively, whereas each dot represents a data point in the test set and the red
lines serve as a guide to the eye for perfect reconstruction. We can see the high reconstruction accuracy of the CNN
model.
We can further improve the reconstruction accuracy for an individual dimension by adjusting its associated weight in
the WMSE loss function. Here we demonstrate this process on the thickness of the SiO2 layer denoted as H2. There
are two major reasons for this adjustment. First, without any special treatment, the prediction accuracy is worst for
H2 among all six dimensions. This is because the scattering signature is very insensitive to H2. Second, H2 is the
thickness of the central layer and governs the leakage current of the DRAM unit and thus has a significant effect on the
device performance. Therefore, accurately determining its value is critical. The optimal value of ωH2 is determined
6
Figure 3: Dimension reconstruction for a nanowire array with CNN-based ellipsometry operating in the angle-resolved
mode. (a) The simplified schematic of the ellipsometer operating in the specular reflection mode. (b) The SEM
cross-section image of the nanowire. (c) The 4D plot of a representative 5×19×15 spectrum. (d) Box plots of the
relative percentage error on the test set of the simulation data. (e) The reconstructed dimensions of TCD, HGT, and SWA
of the silicon nanowire. S, source; D, detector; RS, rotation stage; PSG, polarization state generator; PSA, polarization
state analyzer. Because the SEM measurement does not depend on wavelength, we use the dashed line in (e) to indicate
that they are obtained from a local position of the nanowire.
by performing hyper-parameter tuning on the validation set. The improved accuracy can be observed by comparing
Figs. 4d and 4g, where the relative error reduces from 4.38% to 1.66%. Note that to achieve this improvement, the
reconstruction accuracy for other dimensions dropped slightly, as shown in Figs. 4f and 4h. Thus, there is a trade-off to
make.
We next consider the reconstruction with measurement data. We select four dies on a 12-inch wafer to take the die-to-die
variations into account. Each die is measured thirty times and we use the averaged spectrum to mitigate random
perturbations in the measurement. As can be seen in Figs. 5b -- g, we get an improved reconstruction for H2 and H3
when ωH2 is larger, whereas the accuracy for the other dimensions slightly degrades. Nonetheless, all the reconstructed
dimensions are adequately accurate for good predictions of DRAM performance. The difference in the reconstructed
values obtained for different dies is partly due to the inevitable fabrication errors (e.g., patterning uniformity, line-edge
roughness, and line-width roughness) and measurement errors (e.g., random measurement noise as well as positioning
error when translating the wafer horizontally). We expect that the reconstruction accuracy through the proposed method
can be improved by further reducing the measurement errors.
4 Conclusion
In this paper, we introduced a CNN-assisted nano-metrology framework for the reconstruction of deep-subwavelength
geometrical profiles of nanostructures. The proposed framework works with diffraction-limited optical modalities such
7
Figure 4: Dimension reconstruction of a DRAM array with CNN-based ellipsometry operating in wavelength-resolved
mode. (a), (b) Geometrical model and TEM image of the DRAM array, showing that the critical dimension (marked in
red) is smaller than 10 nm. (c -- e) The reconstructed dimensions of D1, H2, and SWA1 on the test data when all ωi = 1.
The orange and black lines denote the 68% and 95% confidence intervals, respectively. (f -- h) The reconstruction results
on the test data when ωH2 = 10. The thickness of the 95% confidence interval bands (normalized by the nominal
dimensions) are reported in each sub-figure. The red arrows in (f -- h) denote the change in the confidence interval bands
with respect to those from (c -- e) due to the increase in ωH2.
Figure 5: Experimental reconstruction of four DRAM transistors on a 12" wafer. (a) Schematic of the 12" wafer and
the investigated four dies. (b -- g) The averaged reconstruction dimensions of D1, H1, H2, H3, SWA1, and SWA2 for 30
repeated measurements within each die. The TEM measured values are represented by the dark cyan dash-dotted line.
Because the TEM measurement is conducted only on a single DRAM transistor, we use the dash-dotted line to connect
the four values in (b -- g) to indicate that they are obtained from the same TEM measurement.
8
as the bright-field microscopy and the optical ellipsometry, and is non-contact, non-destructive and fast in run-time.
We demonstrated the effectiveness and generality of the proposed method by the successful reconstruction of profiles
(including sub-10-nm dimensions) of various nanostructures, which are widely seen in fields like semiconductor
industry, integrated photonics, metamaterials, one-dimensional photonic crystals, biosensors, and neuromorphic chips,
at a sub-nanometer scale accuracy using single-shot microscopy images or scattering spectra. Moreover, we showed
that the reconstruction accuracy with respect to individual dimensions is adjustable, enabling us to investigate the most
critical dimensions that govern the performance of nanoscale devices. The proposed nanometrology framework is built
upon the power of neural networks in representing the highly nonlinear mapping between the scattering information and
the geometrical dimensions of the measurands and is expected to work with other diffraction-limited optical modalities.
Acknowledgments
We acknowledge Cisco for access to its Arcetri cluster. J. Zhu, S. Purandare, and L. L. Goddard acknowledge
funding from Cisco, UIUC, and ZJU-UIUC; Y. Liu and J.-M. Jin acknowledge funding from Cisco; S. Liu
acknowledges funding from NSFC.
References
[1] Weijian Chen, S¸ ahin Kaya Özdemir, Guangming Zhao, Jan Wiersig, and Lan Yang. Exceptional points enhance
sensing in an optical microcavity. Nature, 548(7666):192, 2017.
[2] Zhaoyi Li, Myoung-Hwan Kim, Cheng Wang, Zhaohong Han, Sajan Shrestha, Adam Christopher Overvig, Ming
Lu, Aaron Stein, Anuradha Murthy Agarwal, Marko Loncar, et al. Controlling propagation and coupling of
waveguide modes using phase-gradient metasurfaces. Nature Nanotech., 12(7):675, 2017.
[3] Ndubuisi G Orji, Mustafa Badaroglu, Bryan M Barnes, Carlos Beitia, Benjamin D Bunday, Umberto Celano,
Regis J Kline, Mark Neisser, Y Obeng, and AE Vladar. Metrology for the next generation of semiconductor
devices. Nat. Electron., 1(10):532, 2018.
[4] Kazuo Furuya. Nanofabrication by advanced electron microscopy using intense and focused beam. Sci. and Tech.
of Adv. Mater., 9(1):014110, 2008.
[5] MJ Williamson, RM Tromp, PM Vereecken, R Hull, and FM Ross. Dynamic microscopy of nanoscale cluster
growth at the solid -- liquid interface. Nature Mater., 2(8):532, 2003.
[6] MJ Rost, L Crama, P Schakel, E Van Tol, GBEM van Velzen-Williams, CF Overgauw, H Ter Horst, H Dekker,
B Okhuijsen, M Seynen, et al. Scanning probe microscopes go video rate and beyond. Review of Scientific Instru.,
76(5):053710, 2005.
[7] Rostislav V Lapshin. Feature-oriented scanning methodology for probe microscopy and nanotechnology. Nan-
otechnology, 15(9):1135, 2004.
[8] Alistair N Boettiger, Bogdan Bintu, Jeffrey R Moffitt, Siyuan Wang, Brian J Beliveau, Geoffrey Fudenberg,
Maxim Imakaev, Leonid A Mirny, Chao-Ting Wu, and Xiaowei Zhuang. Super-resolution imaging reveals distinct
chromatin folding for different epigenetic states. Nature, 529(7586):418, 2016.
[9] Michael J Rust, Mark Bates, and Xiaowei Zhuang. Stochastic optical reconstruction microscopy (STORM)
provides sub-diffraction-limit image resolution. Nature Methods, 3(10):793, 2006.
[10] Nicholas Fang, Hyesog Lee, Cheng Sun, and Xiang Zhang. Sub -- diffraction-limited optical imaging with a silver
superlens. Science, 308(5721):534 -- 537, 2005.
[11] Satoshi Kawata, Atsushi Ono, and Prabhat Verma. Subwavelength colour imaging with a metallic nanolens.
Nature Photon., 2(7):438, 2008.
[12] Slimane Arhab, Gabriel Soriano, Yi Ruan, Guillaume Maire, Anne Talneau, Daniel Sentenac, PC Chaumet, Kamal
Belkebir, and Hugues Giovannini. Nanometric resolution with far-field optical profilometry. Physical Rev. Lett.,
111(5):053902, 2013.
[13] Jinlong Zhu, Shiyuan Liu, Hao Jiang, Chuanwei Zhang, and Xiuguo Chen. Improved deep-etched multilayer
grating reconstruction by considering etching anisotropy and abnormal errors in optical scatterometry. Optics
Lett., 40(4):471 -- 474, 2015.
[14] Jinlong Zhu, Shiyuan Liu, Xiuguo Chen, Chuanwei Zhang, and Hao Jiang. Robust solution to the inverse problem
in optical scatterometry. Optics Express, 22(18):22031 -- 22042, 2014.
9
[15] Jinlong Zhu, Shiyuan Liu, Chuanwei Zhang, Xiuguo Chen, and Zhengqiong Dong. Identification and recon-
struction of diffraction structures in optical scatterometry using support vector machine method. Journal of
Micro/Nanolithography, MEMS, and MOEMS, 12(1):013004, 2013.
[16] Jinlong Zhu, Yating Shi, Shiyuan Liu, and Lynford L Goddard. Generalized measurement configuration optimiza-
tion for accurate reconstruction of periodic nanostructures using optical scatterometry. In Metrology, Inspection,
and Process Control for Microlithography, volume 9778, page 977823. International Society for Optics and
Photonics, 2016.
[17] Jinlong Zhu, Yating Shi, Lynford L Goddard, and Shiyuan Liu. Application of measurement configuration opti-
mization for accurate metrology of sub-wavelength dimensions in multilayer gratings using optical scatterometry.
Applied Opt., 55(25):6844 -- 6849, 2016.
[18] Jinlong Zhu, Hao Jiang, Yating Shi, Chuanwei Zhang, Xiuguo Chen, and Shiyuan Liu. Fast and accurate solution
of inverse problem in optical scatterometry using heuristic search and robust correction. Journal of Vacuum
Science & Tech. B,, 33(3):031807, 2015.
[19] Zhengqiong Dong, Shiyuan Liu, Xiuguo Chen, and Chuanwei Zhang. Determination of an optimal measurement
configuration in optical scatterometry using global sensitivity analysis. Thin Solid Films, 562:16 -- 23, 2014.
[20] Issam Gereige, Stéphane Robert, Sylvie Thiria, Fouad Badran, Gérard Granet, and Jean Jacques Rousseau.
Recognition of diffraction-grating profile using a neural network classifier in optical scatterometry. J. Opt. Soc.
Am. A, 25(7):1661 -- 1667, 2008.
[21] Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition.
arXiv preprint arXiv:1409.1556, 2014.
[22] Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster R-CNN: Towards real-time object detection with
region proposal networks. In Advances in Neural Information Processing Systems, pages 91 -- 99, 2015.
[23] Jonathan Long, Evan Shelhamer, and Trevor Darrell. Fully convolutional networks for semantic segmentation. In
Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3431 -- 3440, 2015.
[24] Yashar D Hezaveh, Laurence Perreault Levasseur, and Philip J Marshall. Fast automated analysis of strong
gravitational lenses with convolutional neural networks. Nature, 548(7669):555, 2017.
[25] Yair Rivenson, Zoltán Göröcs, Harun Günaydın, Yibo Zhang, Hongda Wang, and Aydogan Ozcan. Deep learning
microscopy. Optica, 4(11):1437 -- 1443, 2017.
[26] Yair Rivenson, Hatice Ceylan Koydemir, Hongda Wang, Zhensong Wei, Zhengshuang Ren, Harun Günaydın,
Yibo Zhang, Zoltán Göröcs, Kyle Liang, Derek Tseng, et al. Deep learning enhanced mobile-phone microscopy.
ACS Photon., 5(6):2354 -- 2364, 2018.
[27] Yair Rivenson, Yibo Zhang, Harun Günaydın, Da Teng, and Aydogan Ozcan. Phase recovery and holographic
image reconstruction using deep learning in neural networks. Lig.: Science & Applications, 7(2):17141, 2018.
[28] Ayan Sinha, Justin Lee, Shuai Li, and George Barbastathis. Lensless computational imaging through deep learning.
Optica, 4(9):1117 -- 1125, 2017.
[29] Elias Nehme, Lucien E Weiss, Tomer Michaeli, and Yoav Shechtman. Deep-STORM: super-resolution single-
molecule microscopy by deep learning. Optica, 5(4):458 -- 464, 2018.
[30] Thanh Nguyen, Yujia Xue, Yunzhe Li, Lei Tian, and George Nehmetallah. Deep learning approach for fourier
ptychography microscopy. Optics Express, 26(20):26470 -- 26484, 2018.
[31] Yujia Xue, Shiyi Cheng, Yunzhe Li, and Lei Tian. Reliable deep-learning-based phase imaging with uncertainty
quantification. Optica, 6(5):618 -- 629, 2019.
[32] Michael T Postek, Andras E Vladar, William Keery, Michael Bishop, Benjamin Bunday, and John Allgair.
Reference material (RM) 8820: a versatile new NIST standard for nanometrology. In Metrology, Inspection, and
Process Control for Microlithography XXIV, volume 7638, page 76381B. International Society for Optics and
Photonics, 2010.
[33] Jinlong Zhu, Hao Jiang, Chuanwei Zhang, Xiuguo Chen, and Shiyuan Liu. Data refinement for robust solution to
the inverse problem in optical scatterometry. In Metrology, Inspection, and Process Control for Microlithography
XXIX, volume 9424, page 94240Y. International Society for Optics and Photonics, 2015.
[34] David P Biss, Kathleen S Youngworth, and Thomas G Brown. Dark-field imaging with cylindrical-vector beams.
Applied Opt., 45(3):470 -- 479, 2006.
[35] Eric A Swanson, Joseph A Izatt, Michael R Hee, David Huang, CP Lin, JS Schuman, CA Puliafito, and James G
Fujimoto. In vivo retinal imaging by optical coherence tomography. Optics Lett., 18(21):1864 -- 1866, 1993.
10
[36] Wolfgang Drexler, Uwe Morgner, Ravi K Ghanta, Franz X Kärtner, Joel S Schuman, and James G Fujimoto.
Ultrahigh-resolution ophthalmic optical coherence tomography. Nature Medic., 7(4):502, 2001.
[37] Guillaume Maire, Yi Ruan, Ting Zhang, Patrick C Chaumet, Hugues Giovannini, Daniel Sentenac, Anne
Talneau, Kamal Belkebir, and Anne Sentenac. High-resolution tomographic diffractive microscopy in reflection
configuration. J. Opt. Soc. Am. A, 30(10):2133 -- 2139, 2013.
[38] Basanta Bhaduri, Chris Edwards, Hoa Pham, Renjie Zhou, Tan H Nguyen, Lynford L Goddard, and Gabriel
Popescu. Diffraction phase microscopy: principles and applications in materials and life sciences. Advances in
Optics and Photon., 6(1):57 -- 119, 2014.
[39] Chris Edwards, Basanta Bhaduri, Tan Nguyen, Benjamin G Griffin, Hoa Pham, Taewoo Kim, Gabriel Popescu,
and Lynford L Goddard. Effects of spatial coherence in diffraction phase microscopy. Optics Express, 22(5):5133 --
5146, 2014.
[40] Renjie Zhou, Chris Edwards, Amir Arbabi, Gabriel Popescu, and Lynford L Goddard. Detecting 20 nm wide
defects in large area nanopatterns using optical interferometric microscopy. Nano Lett., 13(8):3716 -- 3721, 2013.
[41] Joseph W Goodman. Introduction to Fourier optics. Roberts and Company Publishers, 2005.
[42] MG Moharam, Eric B Grann, Drew A Pommet, and TK Gaylord. Formulation for stable and efficient im-
plementation of the rigorous coupled-wave analysis of binary gratings. J. Opt. Soc. Am. A, 12(5):1068 -- 1076,
1995.
[43] Xavier Glorot and Yoshua Bengio. Understanding the difficulty of training deep feedforward neural networks. In
Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, pages 249 -- 256,
2010.
[44] Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980,
2014.
[45] Yanan Liu and Jinlong Zhu. Github repository for Optical Deep Learning Nano-Profilometry. https://doi.
org/10.5281/zenodo.2830806, May 2019.
[46] Jinlong Zhu, Renjie Zhou, Lenan Zhang, Baoliang Ge, Chongxin Luo, and Lynford L Goddard. Regularized
pseudo-phase imaging for inspecting and sensing nanoscale features. Optics Express, 27(5):6719 -- 6733, 2019.
11
Supplementary Material
CNN Model Architecture
Our CNN models are adapted from the VGGNet architecture and are modified for critical dimension extraction with
the weighted mean-square-error (WMSE) regression loss. We use the same basic model but with different
parameters for our three examples, due to the difference in input size. For the NIST 8820 artifact example, we use
bright-field microscopy (BFM) to get a 2-D image of the measurand. We crop the image to a size of 101×101. We
then take the average of the 101 horizontal slices of the image, resulting in an input of size 101×1. For the nanowire
example, we use angle-resolved ellipsometry to get the scattering spectra of the measurand, resulting in an input size
of 5×19×15, for the five incident angles, nineteen azimuthal angles, and fifteen elements of the normalized Muller
matrix. In the DRAM case, we use wavelength-resolved ellipsometry with sixty-one wavelengths, resulting in an
input of size 61×15 for the sixty-one wavelengths and fifteen elmenets of the normalized Muller matrix. We use
convolution of kernel size 2 and max pooling of size 2 along each input dimension. We normalize the input and
output to be within range [0, 1], and use the rectified linear unit (ReLU) activation function defined as y(x) = max(0,
x) with convolution layers and the first fully-connected layer, and the sigmoid activation with the last fully-
connected layer. Table S1 lists the details of the CNN models.
Table S1: Architectures of the CNN models used in this paper.
DRAM
15
×
×
2, ReLU)
15
Input: 61
1, ReLU) Conv: (100, 2
Input: 5
Conv: (100, 2, ReLU) Conv: (100, 2
NIST 8820
×
Input: 101
1
1, ReLU)
Batch Norm
Conv: (50, 2
×
2, ReLU)
Conv: (50, 2
Max Pooling (2
× ×
2)
Nanowire
19
2
× ×
× ×
Batch Norm
× ×
Batch Norm
2
Drouput (0.9)
FC (50, ReLU)
2
Batch Norm
Drouput (0.9)
FC (3, sigmoid)
Batch Norm
Conv: (50, 2, ReLU)
Batch Norm
Max Pooling (2)
Drouput (0.9)
FC (50, ReLU)
Batch Norm
Drouput (0.9)
FC (3, sigmoid)
Batch Norm
Max Pooling (2
×
Drouput (0.9)
FC (50, ReLU)
2)
Batch Norm
Drouput (0.9)
FC (6, sigmoid)
CNN Training
For training, we use the Xavier initialization and the Adam optimizer, with the batch size chosen to be 1024, and β1
= 0.9, β2 = 0.999. The learning rate is set initially to 1×10−3, and decays by a factor of 5 whenever the validation loss
hits a plateau, until it reaches 1×10−6. We also checkpoint the network parameters every 10 epochs and use early
stopping to terminate the training once the validation loss is not decreasing for over 500 epochs. The training curves
for the three examples can be found in Fig. S1. The training is done on a single desktop computer with an Intel Core
i7-7700K CPU and NVIDIA GTX-1080 GPU.
12
E
S
M
d
e
t
h
g
e
W
i
E
S
M
d
e
t
h
g
e
W
i
E
S
M
d
e
t
h
g
e
W
i
100
10-1
10-2
10-3
10-4
0
10-5
100
10-1
10-2
10-3
10-4
0
10-5
100
10-1
10-2
10-3
0
Training loss
Validation loss
500
1000
1500
2000
2500
3000
Epoch
(a) The NIST 8820 artifact
Training loss
Validation loss
500
1000
1500
2000
2500
Epoch
(b) The nanowire array.
Training loss
Validation loss
500
1000
1500
2000
2500
Epoch
(c) The DRAM transistor.
Figure S1: Training curves for the CNNs in three examples of this paper.
13
|
1905.09901 | 1 | 1905 | 2019-05-23T20:08:23 | Electromigration Response of Microjoints in 3DIC Packaging Systems | [
"physics.app-ph",
"physics.comp-ph"
] | In multilevel 3D integrated packaging, three major microstructures are viable due to the application of low volume of solders in different sizes, and/or processing conditions. Thermodynamics and kinetics of binary compounds in Cu/Sn/Cu low volume interconnection is taken into account. We show that current crowding effects can induce a driving force to cause excess vacancies saturate and ultimately cluster in the form of microvoids. A kinetic analysis is performed for electromigration mediated intermetallic growth using multi-phase-field approach. Faster growth of intermetallic compounds (IMCs) in anode layer in the expense of shrinkage of cathode IMC layer in shown. This work paves the road for computationally study the ductile failure due to formation of microvoids in low volume solder interconnects in 3DICs. | physics.app-ph | physics | Electromigration Response of Microjoints in 3DIC Packaging Systems
a Department of Materials Science and Engineering, Texas A&M University, College Station, TX, USA
b Department of Mechanical Engineering, Texas A&M University, College Station, TX, USA
Vahid Attari a, Thien Duong a, Raymundo Arroyave a,b
email: [email protected]
Abstract
integrated packaging,
In multilevel 3D
three major
microstructures are viable due to the application of low volume
of solders in different sizes, and/or processing conditions.
Thermodynamics and kinetics of binary compounds in
Cu/Sn/Cu low volume interconnection is taken into account.
We show that current crowding effects can induce a driving
force to cause excess vacancies saturate and ultimately cluster
in the form of microvoids. A kinetic analysis is performed for
electromigration mediated intermetallic growth using multi-
phase-field approach. Faster growth of
intermetallic
compounds (IMCs) in anode layer in the expense of shrinkage
of cathode IMC layer in shown. This work paves the road for
computationally study the ductile failure due to formation of
microvoids in low volume solder interconnects in 3DICs.
Introduction
Solid-Liquid Interdiffusion (SLID1) bonding process [1,2] in
Three Dimensional Integrated Circuits (3DICs) plays a crucial
role in forming ultimate shape and morphology of microjoints
located in between Through Silicon Vias (TSVs); and multi-
phase-field modeling has proven to be a promising approach to
study this process in Cu/Sn/Cu interconnections [3 -- 5]. It is
experimentally shown that electromigration (EM) induced
electron wind force and the subsequent mass flux affects the
diffusion in the form of atomic advection and enhances the
reaction kinetics in the direction of electric field, resulting in a
new dominant kinetic mechanism during operation of these
trend
in
(IMC) growth
devices [3,6 -- 11]. In other words, EM changes the intermetallic
compound
the multi-layered
interconnection systems. We now know that with higher current
densities, the IMCs in the anode side grow faster than the IMCs
in the cathode side [3,6,9,12], and the path to failure can be
amplified further by current-induced temperature localization.
In this respect, the study by Hu et al. [13] shows the extensive
dissolution of Cu in the cathode layer and failure of the flip-
chip bump only after 95 minutes, under 2.5×108 𝐴𝑚−2, while
Joule heating effect also causes a temperature increase up to
157℃ at the backside of the chip.
3DIC interconnect architectures are also subject to considerable
performance issues due to EM and much work remains to be
done to further improving the technology. Physics-based
predictive and quantitative models have the potential to greatly
enable the development of 3DIC low-volume interconnect
(LVI) technologies. This study addresses the need by putting
forward a framework (Refer to Figure 1) that has its main
objective as the development of a high-fidelity multi-physics
microstructure-sensitive model to simulate mass, energy and
charge
complex hierarchical
architectures. The model is capable of predicting the onset of
damage, nucleation of microvoids and therefore loss of electric
conductivity due to formation and coalescence of EM-induced
microvoids. The major result of this work is a fully-coupled
multi-physics software suite capable of simulating not only the
formation of LVI joints but their evolution under current-
stressing conditions. The proposed software tool will in turn be
used to develop process and damage maps that can assist in the
further development of LVI 3DIC technologies.
transport
across
these
Figure 1 The evolution of coupled multi-phase-field model over the past three years [3,5].
1 SLID is also known as Transient Liquid Phase Bonding (TLPB).
1
Microstructure Generation(Multi-Phase field Method)Electro-migration(Due to external Electric Field)Vacancy Transport(Due to external Electric Field)MicrovoidformationResidual Stress Distribution(Due to inhomogeneity & lattice misfit)
Methods
Multi-phase-field Model
A multi-phase-field formalism is integrated to study the
evolution of microstructure at isobaric and isothermal states.
The components of the material system evolve based on the
variational principles of total free energy of the material. A set
of non-conserved (𝜙) and conserved variables (c) describe the
components of the microstructure, where the non-conserved
variables define the spatial fraction of available phases over the
domain (Ω) and the conserved variables define the phase
compositions. We start from a general model of total free
energy of a chemically heterogeneous system that involves
interfacial, bulk and electrical, and elastic interactions [14]:
𝑭𝒕𝒐𝒕𝒂𝒍 = ∫ 𝒇𝒊𝒏𝒕 + 𝒇𝒃𝒖𝒍𝒌 + 𝒇𝒆𝒍𝒆𝒄 + 𝒇𝒆𝒍𝒂𝒔
Ω
(1)
where the four contributing factors are respectively formulated
as:
𝒇𝒊𝒏𝒕 = ∑ ∑
𝜶
𝜷>𝜶
𝟒𝝈𝜶𝜷
𝒘𝜶𝜷
[−
𝟐
𝒘𝜶𝜷
𝝅𝟐 𝛁𝝓𝜶. 𝛁𝝓𝜷 + 𝝓𝜶𝝓𝜷]
𝒇𝒃𝒖𝒍𝒌 = ∑ 𝝓𝜶𝒇𝜶
𝟎(𝒄𝜶)
𝜶
𝒇𝒆𝒍𝒆𝒄 = ∑ 𝝓𝜶𝒇𝜶
𝒆𝒍𝒆𝒄(𝒄𝜶)
𝜶
𝒇𝒆𝒍𝒂𝒔 =
𝟏
𝟐
{∑ 𝝓𝜶(𝜺𝜶
𝒊𝒋 − 𝜺𝜶
∗𝒊𝒋)𝑪𝜶
𝒊𝒋𝒌𝒍(𝜺𝜶
𝒌𝒍 − 𝜺𝜶
∗𝒌𝒍)
}
𝜶
(2)
(3)
(4)
(5)
where 𝜎𝛼𝛽 is the interface energy, and 𝑤𝛼𝛽 is the interface
width. The parabolic double-obstacle potential 𝜙𝛼𝜙𝛽 is
0
defined in the interfacial region only, where 0 < 𝜙𝛼/𝛽 < 1. 𝑓𝛼
is the homogenous free energy of phase "𝛼" with composition
𝑒𝑙𝑒𝑐 is the energy in
𝑐𝛼 where c is the molar fraction of Sn. 𝑓𝛼
𝑖𝑗𝑘𝑙
𝑖𝑗
phase 𝛼 due to the imposed electric field. 𝜀𝛼
are the elastic strain, eigenstrain, and elastic modulus of phase
𝛼, respectively. The chemical free energies are obtained based
on the study of Shim et al. [15]. The evolution of the field
parameter and composition field is governed by:
, and 𝐶𝛼
∗𝑖𝑗
, 𝜀𝛼
𝜕𝜙𝛼
𝜕𝑡
= − ∑
𝛼≠𝛽
𝑀𝛼𝛽
𝑁𝑝
[
−
𝛿𝐹𝑡𝑜𝑡𝑎𝑙
𝛿𝜙𝛽
]
𝛿𝐹𝑡𝑜𝑡𝑎𝑙
𝛿𝜙𝛼
𝐷(𝜙⃗ )
𝐾𝐵𝑇
𝜕𝑐
𝜕𝑡
= ∇. [𝐷(𝜙⃗ ) ∑ 𝜙𝛼
∇𝑐𝛼 −
𝛼
∑ 𝑐𝛼𝜙𝛼. 𝑒𝑍∗(𝜙𝛼)𝜓
𝛼
]
∇. [𝜅(𝑐, 𝜙⃗ ). ∇𝜓] = 0
𝜕𝑐𝑣
𝜕𝑡
= ∇. [𝐷𝑣
1(𝜙⃗ )∇𝑐𝑣 −
𝑒𝑙𝑎𝑠
𝜕𝜎𝑖𝑗
𝜕𝑟𝑗
2(𝜙⃗ )
𝐷𝑣
𝐾𝐵𝑇
= 0
∑ 𝑐𝛼𝜙𝛼. 𝑒𝑍∗(𝜙𝑖)𝜓
𝛼
] −
𝑣
𝑛
(𝑐𝑣
𝛼
− ∑ 𝜙𝛼𝑐𝑒𝑞
𝛼
)𝑒𝑥𝑝(
𝛼
− ∑ 𝜙𝛼𝐸𝑀
𝛼
𝐾𝐵𝑇
)
(6)
(7)
(8)
(9)
(10)
the elastic energy
to be quasi-static and
where 𝑀𝛼𝛽 is the interfacial mobility, and 𝑁𝑝 serves as the
number of coexisting phases in the neighboring grid points. D
is the interdiffusivity parameter as a function of the field order
parameter. The applied electric field and the solid media are
assumed
isotropic, respectively.
Consequently, the distribution of the electrostatic potential over
the domain is calculated assuming the quasi-stationary
conduction process. Hence, Ohm's equation (eqn. (8)) is used
to obtain the electrostatic potential over the domain. 𝜅 is the
conductivity, and 𝜓 is the electrostatic potential. Equation of
mechanical equilibrium (eqn. (9)) is solved iteratively to obtain
the static equilibrium,
is calculated
𝑒𝑙𝑎𝑠 is the elastic stress and 𝑟𝑗 is the jth
subsequently. 𝜎𝑖𝑗
component of position vector, r. Higher order solutions for the
equation of mechanical equilibrium is derived using lower
order approximations till satisfying the tolerance of 10−8 in
displacements. The elastic strain (nth order approximation) is
obtained using Fourier spectral approach [16,17]. Equation
(10) is the vacancy diffusion equation taking into account the
fluxes due to the difference in intrinsic diffusivity of elements
(Cu and Sn) in different features of the microstructure, EM-
mediated flux of vacancies in the reverse direction of atomic
𝛼 is the equilibrium
flux, and annihilation of vacancies. 𝑐𝑒𝑞
𝛼 is the
concentration of vacancies in the phases, and 𝐸𝑀
migration energy of vacancies.
The composition of coexisting phases at a given point is
constrained by the condition of equality of chemical potential
and mass conservation as:
𝟏 [𝒄𝟏(𝒙, 𝒕)] = 𝒇𝒄𝟐
𝒇𝒄𝟏
𝟐 [𝒄𝟐(𝒙, 𝒕)] = ⋯ = 𝒇𝒄𝑵
𝑵 [𝒄𝑵(𝒙, 𝒕)] ≡ 𝒇(𝒙, 𝒕)
𝑵
𝑪 = ∑ 𝝓𝜶𝒄𝜶
𝜶=𝟏
(11)
(12)
where 𝑐𝑁 subscripts under the free energy densities denote the
derivative of the free energy term with respect to concentration
of the phase.
Nucleation of Secondary Phases
Classical nucleation theory is combined with a stochastic
probabilistic framework to account for nucleation of IMC
grains during the isothermal reactions between the Cu substrate
and the Sn solder. The heterogeneous nucleation behavior of
IMCs [18] is modeled by considering that nucleation is an
stochastic event that can be modeled by means of a Poison
distribution. In this respect, the nuclei of Cu6Sn5 and Cu3Sn
IMCs are explicitly seeded in the microstructure by evaluating
the probability of a nucleation event at any (untransformed)
region where nucleation is likely to happen -- -- i.e. at interfaces --
-- at a specific time. The stochastic nucleation probability is
approximated by the Poisson distribution:
𝑷𝒏 = 𝟏 − 𝐞𝐱𝐩 [−(𝑱. 𝒗. 𝚫𝒕)]
(13)
Where 𝐽 is the nucleation rate, 𝑣 is volume of nucleus, and Δ𝑡
is the time interval of nucleation. Consequently, the nucleation
rate can be obtained for an undercooled liquid systems [19]:
2
𝑱 = 𝑱𝟎𝐞𝐱𝐩 [−
𝟏𝟔𝝅𝝈𝟑
𝟑𝒌𝑩𝑻(𝚫𝑮𝒗)𝟐 (
𝑪𝒐𝒔𝟑𝜽 − 𝟑𝑪𝒐𝒔𝜽 + 𝟐
𝟒
)]
(14)
where 𝐽0 is the nucleation rate constant utilized as a frequency
factor. T is the temperature, 𝑘𝐵 is the Boltzmann constant, 𝜎 is
the interface energy, Δ𝐺𝑣 is the driving force for nucleation and
𝜃 is the contact angle between nuclei and substrate. In the
calculation of the stochastic probability of nucleation, we
accounted for the variation in nucleation rate constant and the
contact angle of the embryo with the substrate. The interval of
nucleation, the time of seeding the nucleus into the domain, and
other parameters in equation (13) and equation (14) are
obtained by fitting the nucleation theory to the material
conditions and relevant experimental data [5,19,20].
Results and Discussion
in
Microstructure generation
Multi-phase-field modeling is used to study the formation of a
Cu/Sn/Cu microjoint
the 3DIC packaging systems.
Experimental studies show three different microstructural
patterns for Cu/Sn/Cu microjoint depending on the processing
conditions. These microstructural patterns are composed of Cu
and Sn as initial materials, and Cu3Sn and Cu6Sn5 IMCs as the
consequence of reaction between Cu and Sn. These patterns are
reported in
Table 1. Accordingly, we first use multi-phase-field to study the
formation of these patterns and then analyze the response of the
microstructures under the accelerated EM conditions. This first
step forms the necessary initial microstructures during virtual
thermal treatment conditions, and helps to realize and
understand
thermodynamic and kinetic
conditions that are dominant locally during chemical reactions.
Table 1 Microstructure case studies
the appropriate
change, and Cu3Sn IMCs consume Cu6Sn5 IMCs at the expense
of Sn and after a while Cu3Sn phases cover the entire interlayer.
Another sequence of nucleation of Cu3Sn takes place at this
state. Hence, the number density of Cu3Sn IMCs is higher at
this stage. Such a microstructure is depicted in Figure 2(c).
These three sets of microstructures are the consequence of
dimensional limitations of microbumps in 3DIC systems, and
also different/successive thermal treatment conditions that are
usually utilized during reflowing process.
Figure 2 a) Microstructure case
I [3]: Sn-retained
microstructure for the joint treated at low durations/low
temperatures. b) Microstructure case II [3]: Sn-exhausted
microstructure based on the second pattern in Table 1. c)
Microstructure case III: Cu3Sn intermetallics covering the
entire interlayer of the microstructure in LVIs for longer
treatment durations/high temperatures.
Figure 3 shows an example of the calculated polycrystalline
microstructure with Cu/Cu3Sn/Cu pattern in comparison with
the real microstructure. The microstructure is in close
agreement with the real case [21] which demonstrates the
prosperous integration of 2 𝜇m TSV and Cu-Sn SLID
interconnection between stacked chips.
Case
Initial state
Case I
Case II
Case III
Microstructure pattern
Cu/Sn/Cu
Cu/Cu3Sn/Cu6Sn5/Sn/Cu6Sn5/Cu3Sn/Cu
Cu/Cu3Sn/Cu6Sn5/Cu3Sn/Cu
Cu/Cu3Sn/Cu
The calculated microstructures are shown in Figure 2. In
general, low temperature and low duration thermal treatments
yield a microstructure where the solder material is still present
in the interlayer (Refer to Figure 2(a)). This is also valid for
larger soldering bumps such as C4 and BGAs. In this stage, the
multi-phase-field model takes care of the solid and liquid state
reactions at different locations of the simulation cell. Liquid
state reaction proceeds until consumption of Sn layer entirely.
Increasing thermal treatment temperature and/or duration make
all solder material to be consumed by the Cu6Sn5 IMC layers.
Hence, we may have microstructures that contain both Cu3Sn
and Cu6Sn5 IMCs as shown in Figure 2(b). Furthermore, the
growth regime changes in favor of Cu3Sn after all of the pure
Sn phase is consumed. Accordingly, the kinetics of the reaction
Figure 3 The calculated microstructure (Case III) using multi-
phase-field method versus experimental microstructure [21]
Electromigration response
We have studied the EM response of these micro-joints under
accelerated EM conditions by applying current densities in the
range of 4× 107 at 180℃. Here we used equations 10, and 11
in a coupled way with the phase-field equations to perform
elastic-electrical-chemical multi-phase-field calculations to
3
a)b)c)Cu3SnCuCu
Figure 4 Normalized current density distribution at 180℃ under 4×107 Am-2 in the microstructures. a) case I b) case II, and c) case III.
Less current divergences are present in microstructure case III.
study the evolution of diffusion-driven phenomena in the
sandwich interconnection systems.
The calculations show that the wind force and subsequent
current stressing effect enhance diffusion and alter reaction
kinetics in the direction of electric field, resulting in a new
dominant kinetic mechanism. In other words, EM changes the
IMC growth trend in the multi-layered interconnection systems
as observed in Refs. [6,12]. Our previous study [3] consistently
showed this effect under 4×107 Am-2 or higher current densities
for microstructure case I, and under 4×108 Am-2 or higher
current densities for microstructure case II.
This study shows that microstructure case III has better
response due to presence of a lot of vertical or semi-vertical fast
diffusive channels in Cu3Sn. In addition, since resistivity of
Cu3Sn (9.83) is relatively closer to the resistivity of Cu (1.70),
less EM flux divergences happens in the microstructure. This
has a direct impact on the convective flux of atoms and allows
atoms to migrate easily without inducing further blocking. The
snapshots of distribution of current densities
the
microstructures under 4× 107 Am-2 at 180℃ are shown in
Figure 4.
Residual stresses due to lattice mismatches between the
phases
Microelasticity is the regime in which a material undergoes a
phase transition along with elastic interactions due to misfit
strains
properties.
Crystallographic misfits result in a volume change when a
phase nucleation or transition happens. The stress-free strains
that would result from unconstrained crystallographic misfits
∗ . The accommodation of these
are eigenstrains, denoted 𝜀𝑖𝑗
strains in elastically heterogeneous materials, e.g. multiphase
and/or polycrystalline material systems determine
the
𝑒𝑙 fields. By taking
development of stress 𝜎𝑖𝑗
into account
the appropriate mismatches and elastic
inhomogeneities in elastic modulus of the phases, the elastic
strain and stress distribution are calculated. Accordingly, the
respective elastic driving force for evolution of the multi-phase
microstructure is obtained as it is proposed in Ref. [22].
The snapshots of elastic strain and stress distribution in the
simulation cell for microstructure case I are depicted in Figure
𝑒𝑙 and elastic strain 𝜀𝑖𝑗
inhomogeneity
in
and
in
elastic
5. The elastic strain and stress, and the subsequent calculated
elastic driving force determines whether a phase transition with
expansion or contraction can be favored or hindered. In these
calculations, Cu6Sn5 IMC layers are in direct contact with solid
state Sn. The calculations show that Sn is under tension due to
the growth of Cu6Sn5 from top and bottom side. Both Cu3Sn
and Cu6Sn5 IMCs are under compression. Anisotropy in the GB
energy, the larger the misorientation angle between grains the
faster the shrinkage rate and thicker the grain boundary.
Figure 5 Snapshots of elastic strain and stress distribution in
the simulation cell for microstructure case I. Cu6Sn5 IMC layers
are in direct contact with solid state Sn.
Supersaturation of non-equilibrium vacancies
The vacancy evolution trend is calculated by coupling equation
(10) with the phase-field equations and the result is shown in
Figure 6. Vacancy flux is governed by the differences in
intrinsic diffusivities of elements (Cu and Sn) in each feature of
the microstructure, and the induced convective vacancy flux
due to applied unidirectional electric field. The simulations
show that vacancies evolve in the opposite direction of atomic
flux. The results agree with literature where an increase in the
vacancy amount in the Cu3Sn layer at the cathode side is
expected (red region in Fig. 6). On the other hand, the vacancy
amount in the Cu3Sn layer at the anode side decreases (blue
region in the same figure). The change on the order of EM-
4
a)b)c)
mediated mass flux from TSV to the Sn phase (interlayer) is
more than fourteen orders of magnitude in the Sn-retained
structure. However, the flux only changes about eight orders of
magnitude in the Sn-exhausted microstructure.
shown that IMC covered microstructure has higher resistivity
to EM induced severe mass fluxes.
Acknowledgments
to acknowledge
like
The authors would
the Terra
supercomputing facility of Texas A&M University for
providing computing resources useful in conducting the
research reported in this paper. This research was supported by
the National Science Foundation under NSF Grant No. CMMI-
1462255.
References
[1] H. Liu, K. Wang, K.E. Aasmundtveit, N. Hoivik,
Intermetallic Compound Formation Mechanisms for Cu-Sn
Solid -- Liquid Interdiffusion Bonding, J. Electron. Mater. 41
(2012) 2453 -- 2462.
[2] J. Chang, R. Cheng, K. Kao, T. Chang, T. Chuang, Reliable
Microjoints Formed by Solid -- Liquid Interdiffusion (SLID)
Bonding Within a Chip-Stacking Architecture, IEEE Trans.
Compon. Packag. Manuf. Technol. 2 (2012) 979 -- 984.
[3] V. Attari, S. Ghosh, T. Duong, R. Arroyave, On the
Interfacial Phase Growth and Vacancy Evolution during
Accelerated Electromigration in Cu/Sn/Cu Microjoints, Acta
Mater. (2018).
[4] M.S. Park, R. Arróyave, Multiphase Field Simulations of
Intermetallic Compound Growth During Lead-Free Soldering,
J. Electron. Mater. 38 (2009) 2525 -- 2533.
[5] V. Attari, R. Arroyave, Phase Field Modeling of Joint
Formation During Isothermal Solidification in 3DIC Micro
Packaging, J. Phase Equilibria Diffus. 37 (2016) 469 -- 480.
[6] H. Gan, K.N. Tu, Polarity effect of electromigration on
kinetics of intermetallic compound formation in Pb-free solder
V-groove samples, J. Appl. Phys. 97 (2005) 63514.
[7] B. Chao, S.-H. Chae, X. Zhang, K.-H. Lu, M. Ding, J. Im,
P.S. Ho, Electromigration enhanced intermetallic growth and
void formation in Pb-free solder joints, J. Appl. Phys. 100
(2006)
[8] K.N. Tu, Recent advances on electromigration in very-
large-scale-integration of interconnects, J. Appl. Phys. 94
(2003) 5451 -- 5473.
[9] H. Ceric, S. Selberherr, Electromigration in submicron
interconnect features of integrated circuits, Mater. Sci. Eng. R
Rep. 71 (2011) 53 -- 86.
[10] T.L. Yang, T. Aoki, K. Matsumoto, K. Toriyama, A.
Horibe, H. Mori, Y. Orii, J.Y. Wu, C.R. Kao, Full intermetallic
joints for chip stacking by using thermal gradient bonding, Acta
Mater. 113 (2016) 90 -- 97.
[11] M.O. Alam, B.Y. Wu, Y.C. Chan, K.N. Tu, High electric
current density-induced interfacial reactions in micro ball grid
array (μBGA) solder joints, Acta Mater. 54 (2006) 613 -- 621.
[12] J. Feng, C. Hang, Y. Tian, B. Liu, C. Wang, Growth
kinetics of Cu 6 Sn 5 intermetallic compound in Cu-liquid Sn
interfacial reaction enhanced by electric current, Sci. Rep. 8
(2018) 1775.
[13] Y.C. Hu, Y.H. Lin, C.R. Kao, K.N. Tu, Electromigration
failure in flip chip solder joints due to rapid dissolution of
5
Figure 6 Non-equilibrium vacancy evolution over time along with faster growth
of IMCs at anode layer (bottom) under 4×107 Am-2 at 180℃.
The experimental observations show that the voids form after
certain amount of time in the Cu3Sn IMCs of the cathode layer
[7]. In this respect, the calculations suggest that the migration
of atoms from cathode side to the anode side will leave
vacancies behind. It is important to consider the rate of such a
migration. Obviously, the rate is not consistent in the entire path
and changes from one phase to the other one. In addition, the
interfaces play different roles in this path, as they act as
generators or annihilators of vacancies. While the experiments
suggest that the short-circuit channels and their quantity play
important role as vacancy sink locations in prevention of the
formation of voids, the distinct role of each of these channels
are rarely addressed in the literature. This study illustrates that
the horizontal Cu6Sn5/Cu3Sn interface act as the vacancy
generation source in the cathode layer, while the same
interfaces in the anode side acts as a vacancy annihilation sinks.
On the other hand, other vertical interfaces and the GBs serve
as the rapid pathways for the migration of atoms along the
direction of the applied current.
It is important to develop, and combine the existing model, with
failure prediction models [23 -- 26] to study the microstructural
effects upon ductile failure in the LVI solder joints to
effectively retain good strength and ductility during accelerated
EM conditions in 3DIC systems.
Conclusions
The multi-physics multi-phase-field approach is integrated and
utilized to study EM induced mass transport and the subsequent
impact on IMC growth. IMC growth is enhanced by electric
current at the anode in the expense of IMCs at the cathode. It is
JvacancyJEMEvolution time
copper, J. Mater. Res. 18 (2003) 2544 -- 2548.
[14] N. Moelans, B. Blanpain, P. Wollants, An introduction to
phase-field modeling of microstructure evolution, Calphad. 32
(2008) 268 -- 294.
[15] C.-S.O. J. -H. Shim, Thermodynamic assessment of the
Cu-Sn system, Z. Fuer Met. Res. Adv. Tech. 87 (1996) 205 --
212.
[16] H. Moulinec, P. Suquet, A numerical method for
computing the overall response of nonlinear composites with
complex microstructure, Comput. Methods Appl. Mech. Eng.
157 (1998) 69 -- 94.
[17] J.C. Michel, H. Moulinec, P. Suquet, Effective properties
of composite materials with periodic microstructure: a
computational approach, Comput. Methods Appl. Mech. Eng.
172 (1999) 109 -- 143.
[18] C.-C. Pan, C.-H. Yu, K.-L. Lin, The amorphous origin and
the nucleation of intermetallic compounds formed at the
interface during the soldering of Sn -- 3.0Ag -- 0.5Cu on a Cu
substrate, Appl. Phys. Lett. 93 (2008) 61912.
[19] J.P. Simmons, Y. Wen, C. Shen, Y.Z. Wang,
Microstructural development involving nucleation and growth
phenomena simulated with the Phase Field method, Mater. Sci.
Eng. A. 365 (2004) 136 -- 143.
[20] R. Gagliano, M.E. Fine, Growth of η phase scallops and
whiskers in liquid tin-solid copper reaction couples, JOM. 53
(2001) 33 -- 38.
[21] C.-T. Ko, K.-N. Chen, Low temperature bonding
technology for 3D integration, Microelectron. Reliab. 52 (2012)
302 -- 311.
[22] I. Steinbach, M. Apel, Multi phase field model for solid
state transformation with elastic strain, Phys. Nonlinear
Phenom. 217 (2006) 153 -- 160.
[23] M. Ambati, T. Gerasimov, L. De Lorenzis, Phase-field
modeling of ductile fracture, Comput. Mech. 55 (2015) 1017 --
1040.
[24] B. Haghgouyan, N. Shafaghi, C.C. Aydıner, G. Anlas,
Experimental and computational investigation of the effect of
phase transformation on fracture parameters of an SMA, Smart
Mater. Struct. 25 (2016) 75010.
[25] C. Miehe, M. Hofacker, L.-M. Schänzel, F. Aldakheel,
Phase field modeling of fracture in multi-physics problems. Part
II. Coupled brittle-to-ductile failure criteria and crack
propagation in thermo-elastic -- plastic solids, Comput. Methods
Appl. Mech. Eng. 294 (2015) 486 -- 522.
[26] P. Eisenlohr, M. Diehl, R.A. Lebensohn, F. Roters, A
spectral method solution to crystal elasto-viscoplasticity at
finite strains, Int. J. Plast. 46 (2013) 37 -- 53.
6
|
1905.07480 | 1 | 1905 | 2019-05-17T21:04:01 | Elastic Weyl points and surface arc states in 3D structures | [
"physics.app-ph"
] | The study of Weyl points in electronic systems has inspired many recent researches in classical systems such as photonic and acoustic lattices. Here we show how the Weyl physics can also inspire the design of novel elastic structures. We construct a single-phase 3D structure, an analogue of the AA-stacked honeycomb lattice, and predict the existence of Weyl points with opposite topological charges (${\pm 1}$), elastic Fermi arcs, and the associated gapless topologically protected surface states. We employ full-scale numerical simulations on the elastic 3D structure, and present a clear visualization of topological surface states that are directional and robust. Such designed lattices can pave the way for novel vibration control and energy harvesting on structures that are ubiquitous in many engineering applications. | physics.app-ph | physics |
Elastic Weyl points and surface arc states in 3D structures
Xiaotian Shi,1 Rajesh Chaunsali,1, 2 Feng Li,3 and Jinkyu Yang1, ∗
1Aeronautics and Astronautics, University of Washington, Seattle, WA 98195, USA
2LAUM, CNRS, Le Mans Universit´e, Avenue Olivier Messiaen, 72085 Le Mans, France
3South China University of Technology, School of Physics and
Optoelectronic Technology, Guangzhou, Guangdong, 5140640, China
The study of Weyl points in electronic systems has inspired many recent researches in classical
systems such as photonic and acoustic lattices. Here we show how the Weyl physics can also inspire
the design of novel elastic structures. We construct a single-phase 3D structure, an analogue of the
AA-stacked honeycomb lattice, and predict the existence of Weyl points with opposite topological
charges (±1), elastic Fermi arcs, and the associated gapless topologically protected surface states.
We employ full-scale numerical simulations on the elastic 3D structure, and present a clear visual-
ization of topological surface states that are directional and robust. Such designed lattices can pave
the way for novel vibration control and energy harvesting on structures that are ubiquitous in many
engineering applications.
INTRODUCTION
Phononic crystals and metamaterials have shown new
and exciting ways to control the flow of wave propaga-
tion in the medium [1 -- 4]. Recently, the topology of band
structures has emerged as a new design tool in this con-
text. The essential idea is to characterize the bulk dis-
persion topologically and predict its implications on the
edges/surfaces of the system. A nonzero topological in-
variant of the bulk usually implies the existence of edge
or surface waves with nontrivial properties, such as direc-
tionality and robustness [5 -- 7]. Several interesting strate-
gies to manipulate elastic waves have thus been shown
[8, 9]. However, the studies so far focused mainly on 1D
and 2D systems. It is not clear how a 3D elastic structure
could be designed to support topological surface states.
What special characteristics those surface state would
have and how they could be harnessed in engineering set-
tings are some key questions. In this study, we attempt
to address these questions by taking inspiration from the
Weyl physics.
Weyl semimetals [10 -- 14] have recently attracted sig-
In Weyl
nificant attention for their exotic features.
semimetals, the Weyl point refers to the degeneracy point
of two bands having linear dispersion in all directions in
the 3D reciprocal space. The effective Weyl Hamilto-
nian is, in general, given by H(k) = f (k)σ0 + vxkxσx +
vykyσy + vzkzσz, where f (k) is an arbitrary real func-
tion, vi, ki, and σi represent group velocity, momentum,
and Pauli matrix, respectively. Weyl points behave as
the sources or the sinks of the Berry curvature in the re-
ciprocal space. By integrating the Berry flux on a sphere
surrounding a Weyl point, we can get the non-vanishing
topological charge (or Chern number) associated with it
[15]. The Weyl point is robust against small perturba-
tions and cannot be easily gapped unless it is annihilated
∗[email protected]
with another Weyl point with the opposite topological
charge [16]. For electronic systems, previous researches
have shown many unusual phenomena associated with
Weyl points, such as robust surface states [10] and chiral
anomaly [17]. Later on, the Weyl physics has shown to
be useful in the classical systems of photonic [16, 18 -- 20]
and acoustic lattices [21 -- 27].
The implementation of the Weyl physics in elastic
structures, however, has been challenging so far. Re-
cently, a self-assembled double gyroid structure that con-
tains Weyl points for both electromagnetic and elastic
waves was proposed [28]. Later, a design consisting of
a thin plate and beams, which carries both Weyl and
double-Weyl points, was also proposed [29]. In spite of
these, the experimental demonstrations of elastic Weyl
points remain elusive to date. Furthermore, there has
not been even a study reporting full-scale numerical sim-
ulations in the elastic setting, by which the Weyl physics
can be directly visualized and appreciated. This is due
to the fact that such structures are extremely intricate
to fabricate. At the same time, it is very demanding to
computationally simulate their full-scale wave dynamics,
because it involves with several types of elastic modes.
In this research, we design a 3D elastic lattice made
entirely of beams, which allow both translational and ro-
tational degrees of freedom along their length directions.
We employ the finite element analysis (FEA) to obtain a
dispersion diagram for the unit cell and discuss its topo-
logical features in relation to the Weyl physics. Inspired
by widely used 3D hollow structures in engineering, e.g.,
fuselage, we construct a full-scale hollow structure and
show the existence of topological surface states in it. We
also elucidate the relation of their directionality with the
elastic Fermi arcs in the reciprocal space. We perform
a transient simulation of the structure to numerically
demonstrate the propagation of nontrivial surface waves.
This study therefore paves the way for future research on
the design and fabrication of Weyl physics-based struc-
tures for engineering applications, such as vibration con-
trol and energy harvesting.
2
FIG. 1: [Color online] (a) AA-stacked hexagonal lattice (blue)
with chiral inter-layer hopping (orange). (b) Top (left panel)
and slanted (right) view of the unit cell of the 3D elastic
structure. (c) Illustration of the first Brillouin zone and Weyl
points with opposite topological charges indicated by the yel-
low and purple spheres.
DESIGN OF WEYL STRUCTURE
Previous studies in acoustics demonstrated the exis-
tence of Weyl points in a AA-stacked honeycomb lat-
tice with chiral interlayer hopping [21]. A schematic of
its nearest-neighbor tight-binding model is illustrated in
Fig. 1(a). To make an equivalent mechanical system, we
would need to use masses and springs that are connected
by hinge joints. However, for the more realistic design,
we deviate from spring-mass description and propose a
unit cell made of space beams as shown in Fig. 1(b). We
take beam length L = 20 mm and height P = 10 mm.
All in-plane beams (parallel to xy plane) have square
cross section of width 3.0 mm, while out-of-plane beams
have square cross section of width 0.7 mm to reduce the
inter-layer stiffness. Note that we can still calculate the
effective tight-binding Hamiltonian to analyze topologi-
cal properties of our elastic structure (Appendix A). In
Fig. 1(c), we show the first Brillouin zone with marked
Weyl points at the high symmetric points. These are of
two opposite charges (Appendix A) and will appear in
the simulation results shown in the next section.
To conduct the numerical simulation, we use the com-
mercial FEA software ABAQUS. We model the space
beams using the Timoshenko beam elements. We fol-
low the method used in [30] to apply periodic boundary
conditions and calculate frequency band diagrams. We
FIG. 2:
[Color online] Dispersion diagrams on the reduced
reciprocal kxky plane with fixed (a) kz = 0, (b) kz = π/(2P ),
and (c) kz = π/P . The yellow (purple) sphere refers to the
Weyl point located at the K (H) point with topological charge
−1 (+1). The grey area in (b) represents a complete band
gap. (d) Dispersion diagram along the KH line. The red and
black curves are obtained from the two-band Hamiltonian and
the FEA simulations, respectively.
use stainless steel 316L as the structural material with
elastic modulus E = 180 GPa, density ρ = 7900 kg/m3,
and Poisson's ratio ν = 0.3, which could be used for cur-
rent 3D metal printing [31]. We neglect any material
dissipation. The out-of-plane beams produce an effective
synthetic gauge flux and break the effective time-reversal
symmetry at a fixed kz. Therefore, the system can be
treated as an elastic realization of the topological Hal-
dane model [32].
RESULT
Weyl points in unit-cell dispersion
We show the band structure of the unit cell along the
irreducible Brillouin zone at kz = 0 in Fig. 2(a). We ob-
serve that the 13th and the 14th bands, predominately
with out-of-plane polarization (see Appendix B), are de-
generate at the K point around 6.12 kHz [see the rectan-
gular box and its close-up inset in Fig. 2(a)]. This is a
Weyl point in the system, and it is the same as the yellow
spheres in Fig. 1(c). We calculate the Weyl charge by
fitting the dispersion diagram of a two-band Hamiltonian
around the Weyl point (see the red curves in the inset).
As a result, we obtain the Weyl charge of −1 from this
Weyl point (see Appendix A).
When we plot the dispersion diagram for kz = π/(2P ),
the degeneracy of the bands is lifted, and there emerges
a band gap between 13th and 14th bands, as shown by
the grey region in Fig. 2(b). We increase kz further
to plot the dispersion curves at kz = π/P [Fig. 2(c)].
𝐿𝐾𝐴𝐻𝑀ഥ𝐾തΓഥ𝑀Γ𝑘𝑥𝑘𝑧𝑘𝑦cabLPPLFrequency (kHz)64bcad6.06.16.25.25.35.4ΓΚΜΑΗLΚΗK�Γ�M�5+1−1738 3
FIG. 3: [Color online] (a) Schematic of supercells with zigzag (green box) and armchair (red box) type boundaries. (b)-(c) Band
structure with fixed kz = π/(2P ) for the zigzag and armchair type supercells, respectively. The red and blue curves represent
the topological surface modes at the two opposite ends of the supercell. (d)-(e) Mode shapes of the surface modes at 5.4 kHz
corresponding to the solid and hollow stars in (b) and (c). Color intensity represents the magnitude of total displacements.
We observe that the band gap closes again, and the two
bands establish a degeneracy at 5.34 kHz, but at the H
point of the Brillouin zone instead of the K point. This is
the second Weyl point in the system, and it corresponds
to the purple markers in Fig. 1(c). As shown in the inset,
we again use the two-band Hamiltonian to represent the
dispersion characteristics around this Weyl point and find
that the Weyl point has +1 topological charge.
In Fig. 1(d), we plot the dispersion diagram along
the KH direction to clearly visualize the Weyl degener-
acy at the K and H points and the existence of a band
gap between the 13th and the 14th bands when the kz
value lies somewhere in between. We note that the two-
band Hamiltonian (red curves) captures this evolution of
the bands obtained through the FEA simulations (black
curves) reasonably well. We use this effective Hamitonian
to numerically calculate the Chern numbers of the two
bands above and below the band gap for a fixed kz. For
positive kz, these are −1 and +1 for the upper and the
lower bands, respectively, as marked in Fig. 2(b). This
indicates that the band gap is topologically nontrivial.
Directional surface states in supercell
Based on the bulk-edge correspondence of topology, we
expect topologically protected boundary modes arising at
finite boundaries. To this end, we construct two types of
supercells, consisting of 15 unit cells each, having both
armchair and zigzag types of finite boundaries. Figure
3(a) shows a schematic of how we choose the two types of
the supercells. For the zigzag supercell (see the slanted
green box), we apply periodic boundary conditions in
both x- and z-directions. We use free boundary condi-
tions at the top and bottom ends. We fix kz = π/2P
and plot dispersion diagram in Fig. 3(b). We observe
two modes inside the band gap, which are localized on
the top (red) and the bottom (blue) ends of the supercell
[see Fig. 3(d)]. Based on their slope we can conclude
that the top (bottom) end mode would have a negative
(positive) group velocity in the x-direction [see the green
arrows in Fig. 3(a)].
Similarly, we study another supercell with the arm-
chair type of boundaries [see the horizontal red box in
Fig. 3(a)]. We show that it supports localized modes
on the left and the right ends [see Figs. 3(c) and 3(e)].
These left and right end modes exhibit negative and pos-
itive group velocities, respectively [see the red arrows in
Fig. 3(a)]. Therefore, it is straightforward to deduce
that a wave packet injected at 5.4 kHz (shown as star) on
the surface of the full-scale lattice, having simultaneous
zigzag and armchair boundaries, would travel counter-
clockwise for kz = π/(2P ). In the same vein, we expect
to obtain a traveling surface wave in the clockwise direc-
tion for kz = −π/(2P ).
Elastic Fermi arcs in full-scale model
We now demonstrate the existence of surface states in
a full-scale 3D structure. We choose a hollow structure
for two reasons: (1) such structures are ubiquitous in ap-
plications; and (2) they require a reduced amount of com-
putational time for numerical simulations, compared to
solid ones. Without fixing kz, we first excite our system
at 5.4 kHz in the z-direction. In Figs. 4(a) and 4(b), we
show the z-component of steady-state wave displacement
when the excitation is placed at the centers of positive
xz (denoted as PXZ, see the red star mark in the inset)
and negative yz (NYZ) planes, respectively. As we can
see, the surface states propagate in particular directions
and do not spread all across the whole plane. Especially
in Fig. 4(b), the wave propagates in the y-direction pre-
dominantly reflecting the collimation effect [25].
To investigate further, we perform the fast Fourier
transformation (FFT) on the displacement field in the
two spatial directions and plot the spectrum in Figs. 4(c)
and 4(d). We observe the arc-like pattern of the peak
spectral density (in yellow). These are called "Fermi
arcs", also seen as the counterpart representation of the
1357Frequency (kHz)-110𝑘𝑥Τ𝜋3𝐿bc𝑘𝑦Τ𝜋3𝐿-110deZigzagArmchaira3𝐿3𝐿⋮⋮⋮⋮⋮⋮⋯⋯⋯⋯⋯⋯1357Frequency (kHz)-110𝑘𝑥Τ𝜋3𝐿bc𝑘𝑦Τ𝜋3𝐿-110deZigzagArmchaira3𝐿3𝐿⋮⋮⋮⋮⋮⋮⋯⋯⋯⋯⋯⋯1357Frequency (kHz)-110𝑘𝑥Τ𝜋3𝐿bc𝑘𝑦Τ𝜋3𝐿-110deZigzagArmchaira3𝐿3𝐿⋮⋮⋮⋮⋮⋮⋯⋯⋯⋯⋯⋯ 4
FIG. 4: [Color online] Surface states and elastic Fermi arcs in
a full-scale 3D hollow structure. (a)-(b) Surface states under
the harmonic excitation at 5.4 kHz on the xz and yz planes,
respectively. The point sources of excitation are placed in
the center of each plane, as shown in the insets. The color
intensity represents the nodal displacement in the z-direction
(uz).
(c)-(d) Spatial Fourier transforms of the field distri-
butions of the surface states on the xz and yz planes, re-
spectively. Spectral density shows the elastic Fermi arcs that
connect the projections of the Weyl points with the opposite
topological charges in the reduced 2D Brillouin zone. The
red solid curves represent the simulated elastic Fermi arcs on
the corresponding surfaces through supercell analysis, while
the dashed curves indicate the Fermi arcs on the opposite sur-
faces. The projected bulk bands are shown as the dotted grey
curves.
surface states in the reciprocal space. Since the normal
vector to the Fermi arcs would determine the direction of
the wave's group velocity, we can deduce from Fig. 4(d)
that the wave will propagate in the ky-direction predom-
inantly given the straight posture of the arc. This thus
confirms the aforementioned collimation effect [Fig. 4(b)]
in the wavevector space.
These Fermi arcs could also be obtained from the equi-
frequency contour (FEC) analysis on the supercell (see
details in Appendix C). To achieve this, we calculate
dispersion characteristics of the supercells, as shown in
Figs. 3(b) and 3(c), for all values of kz inside the first
Brillouin zone. We then extract the wave numbers for
5.4 kHz to obtain the red curves in Figs. 4(c) and 4(d).
Here, solid curves correspond to the the solutions on the
plane of excitation, while the dashed curves represent
those on the opposite surface. Evidently, the red solid
curves closely match the spectral density arcs obtained
from the full-scale simulation. The Fermi arc generally
connects the Weyl points of the opposite charges [23],
but here we see that they connect the two Weyl points
FIG. 5: [Color online] Robust one-way surface state propagat-
ing in a 3D hollow structure. (a) Star highlights the location
of multi-point phased excitation. The exact locations of ex-
ternal loads are shown in the inset. (b) -- (d) The magnitude
of total displacement at time t = 4 ms, t = 6.5 ms, and t = 9
ms, respectively. Inset in (c) shows the top view when the
wave turns at the corner.
(in purple and yellow) roughly but not exactly. This is
because the system supports the two Weyl points at dif-
ferent frequencies [see the frequency offset of the yellow
and purple points in Fig. 2(d)].
Robust one-way propagation
We now proceed to the full-scale, transient numerical
study performed at 5.4 kHz, but for a fixed kz. Figure
5(a) shows the entire structure. We fix all the degrees
of freedom of the nodes on the top and bottom layers.
To ensure the excitation with the desired kz, we apply
four-point excitation in the z-direction on the 10th to the
13th layers on the YZ plane (marked with the red arrows
in Fig. 5(a) inset). We use a Gaussian-modulated sinu-
soidal pulse with the center frequency of 5.4 kHz, and we
fix kz = −π/(2P ) by increasing the phase of the input
signal by π/2 from the 10th to the 13th layer. From the
discussions above, we expect that the wave packet would
propagate clockwise when looking from the top (z-axis).
In Figs. 5(b) -- 5(d), we plot the displacement amplitude
of the system at time t = 4 ms, t = 6.5 ms, and t = 9 ms,
respectively. We observe that the elastic wave remains
on the surface of the structure and only travels upward
in the clockwise direction (viewing from the top) with-
out obvious reflection at the corners (see Fig. 5(c) and
its inset, also Supplementary Movie 1). This, therefore,
demonstrates a robust one-way propagation of surface
elastic wave in our Weyl structure.
acdb𝑘𝑥Τ2𝜋3𝐿𝑥(𝑚)𝑘𝑦2Τ𝜋3𝐿y(𝑚)z(𝑚)𝑘𝑧Τ𝜋𝑃0100.800.41−100.5−0.51−100MaxMinMaxMinNYZNXZ𝑧𝑥𝑦PXZPYZNYZNXZ𝑧𝑥𝑦PXZPYZabcdMaxMin CONCLUSION
We design a 3D mechanical structure -- analogous to
the AA-stacked honeycomb lattice -- by using slender
beams. We show that this relatively simple design car-
ries Weyl points at the vertices of the Brillouin zone. We
use a two-band Hamiltonian model to describe the dy-
namics around the Weyl points and calculate their topo-
logical charges. We show the finite boundaries of this
structure, both zigzag and armchair types, host local-
ized states at a fixed kz. Using numerical simulations on
a full-scale 3D structure, we show the existence of the
Fermi arcs and compare them with the results obtained
from equi-frequency contour analysis. We highlight two
unique wave phenomena in our structure: (i) collimation
of the propagating elastic waves and (ii) robust one-way
transport of elastic energy around the corners. Our de-
sign could be easily scaled up or down, and can be rele-
vant to applications, such as sensing, energy harvesting,
and vibration control on 3D elastic structures. Studies on
the experimental verification of the elastic surface states
are expected in the authors' future publications.
ACKNOWLEDGMENTS
We thank Dr. Hyunryung Kim, Dr. Ying Wu, and
Shuaifeng Li for fruitful discussions. X. S., R. C., and J.
Y. are grateful for the financial support from the U.S. Na-
tional Science Foundation (CAREER1553202 and EFRI-
1741685).
APPENDIX A: TIGHT-BINDING MODEL OF
AA-STACKED GRAPHENE
We consider the tight-binding model of a AA-stacked
graphene with chiral interlayer coupling, as depicted in
Fig. 1(a). The unit cell has an in-plane lattice constant
L and out-of-plane lattice constant P in the z-direction
[see Fig. 1(b)]. Let the intralayer (interlayer) coupling
be tn (tc). Therefore, we write the Bloch Hamiltonian
given by [16, 21, 27]
(cid:18) ε + tcf (kzP )
(cid:19)
e−ikyL + 2eikyL/2 cos(cid:0)√
3kxL/2(cid:1),
3kxL − kzP(cid:1) +
2 cos(cid:0)√
3kxL/2 + kzP(cid:1).
where
β
and
applying
the k · p method[16], we can expand the Hamiltonian
√
near the K point [kx = 4
3π/(9L), ky = 0, kz = 0] and
obtain the effective Hamiltonian
ε
=
f (kzP )
4 cos(3kyL/2) cos(cid:0)√
tnβ
ε + tcf (−kzP )
denotes
the
on-site
potential,
By
H(k) =
∗
(tnβ)
,
=
H(∆k) =(ε − 3tc)σ0 − 3
2
Ltn(∆kxσx − ∆kyσy)
√
− 3
3Ltc∆kzσz,
5
where ∆k = (∆kx, ∆ky, ∆kz) is a small k-vector deviat-
ing from the K point, σ0 is the 2 × 2 unit matrix, and
σx, σy, σz are the Pauli matrices.
We use ε = 5.73 kHz, tn = 0.875 kHz, and tc = −0.131
kHz for fitting the two-band dispersion with the curves
obtained from the FEA results seen in Fig. 2. The effec-
tive Hamiltonian describes a Weyl point at the K point,
whose topological charge is given by C = sgn(vxvyvz),
where Dirac velocities vx = −3Ltn/2, vy = 3Ltn/2,
√
and vz = −3
3Ltc. Therefore, C = −1 in this case.
√
Similarly, by expanding the Hamiltonian at the H point
[kx = 4
3π/(9L), ky = 0, kz = π/P ], we verify that there
is another Weyl point with the topological charge of +1
located at the H point.
APPENDIX B: UNIT-CELL DISPERSION AND
MODE POLARIZATION
In this appendix, we show the modes that are degen-
erate at the Weyl points are an out-of-plane type with
a predominant z component. To this end, we define
a polarization factor Pz = Uz2/(Ux2 + Uy2 + Uz2)
to distinguish bands with different polarization compo-
nents, where Ux, Uy, and Uz are the x-, y-, and z- com-
ponents of the eigenvectors. Therefore, the out-of-plane
modes, with predominately Uz component, would have
the polarization factor close to a unity, while the po-
larization factors of the in-plane modes would approach
zero. We plot the bulk bands of the unit cell -- colored
with the information about the polarization factors -- on
the 2D reciprocal plane at various kz values in Fig. 6.
We can clearly see that the Weyl points are formed by
the degeneracy of the two bands containing out-of-plane
modes (in red).
FIG. 6: [Color online] Dispersion diagram of the unit cell in
the kxky plane at (a) kz = 0, (b) kz = π/(4P ), (c) kz =
π/(2P ), (d) kz = 3π/(4P ), and (e) kz = π/P , respectively.
Colormap represents the polarization factor Pz. The grey area
in (b) represents a partial frequency band gap (noted as PBG)
while the grey zones in (c) and (d) refer to complete band gaps
(noted as CBG). The flat branches mainly represent the cases
when the interlayer beams are locally resonant.
aFrequency (kHz)6420bcdeΓΚΜK�Γ�M�K�Γ�M�K�Γ�M�ΑΗL01CBGCBGPBG APPENDIX C: EQUI-FREQUENCY CONTOUR
ANALYSIS
6
√
√
Equi-frequency contour on the the xz plane, as shown
in Fig. 4(c), can be obtained by calculating the band
structure of the supercell with the zigzag type bound-
ary. At f = 5.4 kHz, we vary kx from −4π/(3
3L)
3L) and kz from −π/P to π/P , to obtain
to 4π/(3
wavevector plots in green as shown in Fig. 7(a). Note
that we plot another equi-frequency plot at slightly
higher frequency (f = 5.5 kHz) in blue to get a sense
of the normal vector, i.e., the group velocity shown by
red arrows at kz = ±π/2P . Similarly, result of the yz
plane, from the supercell with the armchair type bound-
ary, is shown in Fig. 7(b). This equi-frequency contour
confirms the collimation effect in the ky direction.
FIG. 7: [Color online] (a) Equi-frequency contour of the su-
percell with the zigzag type boundaries at f = 5.4 kHz (green)
and f = 5.5 kHz (blue). Red arrows show the normal vectors
of the dispersion curves. Solid and dashed lines represents the
mode in the forward and the opposite planes as indicated in
Fig. 4. The grey regions refers to the projections of the bulk
bands. (b) Similar results of the supercell with armchair type
boundaries.
[1] Z. Liu, X. Zhang, Y. Mao, Y. Y. Zhu, Z. Yang, C. T.
Chan, and P. Sheng, Locally Resonant Sonic Materials,
Science 289, 1734 (2000).
[2] M. I. Hussein, M. J. Leamy, and M. Ruzzene, Dynamics
of Phononic Materials and Structures: Historical Origins,
Recent Progress, and Future Outlook, Appl. Mech. Rev.
66, 040802 (2014).
[3] M. Kadic, T. Bckmann, R. Schittny, and M. We-
gener, Metamaterials beyond electromagnetism, Rep.
Prog. Phys. 76, 126501 (2013).
[4] K. Bertoldi, V. Vitelli, J. Christensen, and M. van Hecke,
Flexible mechanical metamaterials, Nature Reviews Ma-
terials 2, 17066 (2017).
[5] M. Z. Hasan and C. L. Kane, Colloquium: Topological
insulators, Rev. Mod. Phys. 82, 3045 (2010).
[6] X.-L. Qi and S.-C. Zhang, Topological insulators and su-
perconductors, Rev. Mod. Phys. 83, 1057 (2011).
[7] E. Prodan and H. Schulz-Baldes, Bulk and Boundary In-
variants for Complex Topological Insulators (Springer In-
ternational Publishing, Cham, 2016) arXiv:1510.08744.
[8] S. D. Huber, Topological mechanics, Nature Physics 12,
621 (2016).
[9] G. Ma, M. Xiao, and C. T. Chan, Topological phases
in acoustic and mechanical systems, Nature Reviews
Physics 1 (2019).
[10] X. Wan, A. M. Turner, A. Vishwanath, and S. Y.
Savrasov, Topological semimetal and Fermi-arc surface
states in the electronic structure of pyrochlore iridates,
Phys. Rev. B 83, 205101 (2011).
[11] C. Fang, M. J. Gilbert, X. Dai, and B. A. Bernevig,
Multi-Weyl Topological Semimetals Stabilized by Point
Group Symmetry, Phys. Rev. Lett. 108, 266802 (2012).
[12] S.-Y. Xu, I. Belopolski, N. Alidoust, M. Neupane, G.
Bian, C. Zhang, R. Sankar, G. Chang, Z. Yuan, C.-C.
Lee, S.-M. Huang, H. Zheng, J. Ma, D. S. Sanchez, B.
Wang, A. Bansil, F. Chou, P. P. Shibayev, H. Lin, S. Jia,
and M. Z. Hasan, Discovery of a Weyl fermion semimetal
and topological Fermi arcs, Science 349, 613 (2015).
[13] A. A. Soluyanov, D. Gresch, Z. Wang, Q. Wu, M. Troyer,
X. Dai, and B. A. Bernevig, Type-II Weyl semimetals,
Nature 527, 495 (2015).
[14] N. P. Armitage, E. J. Mele, and A. Vishwanath, Weyl
and Dirac semimetals in three-dimensional solids, Rev.
Mod. Phys. 90, 015001 (2018).
[15] Z. Fang, N. Nagaosa, K. S. Takahashi, A. Asamitsu, R.
Mathieu, T. Ogasawara, H. Yamada, M. Kawasaki, Y.
Tokura, and K. Terakura, The Anomalous Hall Effect
and Magnetic Monopoles in Momentum Space, Science
302, 92 (2003).
[16] W.-J. Chen, M. Xiao, and C. T. Chan, Photonic crys-
tals possessing multiple Weyl points and the experimen-
tal observation of robust surface states, Nature Commu-
nications 7, 13038 (2016).
[17] H. B. Nielsen and M. Ninomiya, The Adler-Bell-Jackiw
anomaly and Weyl fermions in a crystal, Physics Letters
B 130, 389 (1983).
[18] L. Lu, L. Fu, J. D. Joannopoulos, and M. Soljai, Weyl
points and line nodes in gyroid photonic crystals, Nature
Photonics 7, 294 (2013).
[19] L. Lu, Z. Wang, D. Ye, L. Ran, L. Fu, J. D. Joannopoulos,
and M. Soljai, Experimental observation of Weyl points,
Science 349, 622 (2015).
[20] M.-L. Chang, M. Xiao, W.-J. Chen, and C. T. Chan,
Multiple Weyl points and the sign change of their topo-
logical charges in woodpile photonic crystals, Phys. Rev.
B 95, 125136 (2017).
[21] M. Xiao, W.-J. Chen, W.-Y. He, and C. T. Chan, Syn-
thetic gauge flux and Weyl points in acoustic systems,
Nature Physics 11, 920 (2015).
[22] Z. Yang and B. Zhang, Acoustic Type-II Weyl Nodes
from Stacking Dimerized Chains, Phys. Rev. Lett. 117,
224301 (2016).
[23] F. Li, X. Huang, J. Lu, J. Ma, and Z. Liu, Weyl points
and Fermi arcs in a chiral phononic crystal, Nature
Physics 14, 30 (2018).
[24] T. Liu, S. Zheng, H. Dai, D. Yu, and B. Xia, Acous-
tic semimetal with Weyl points and surface states,
arXiv:1803.04284 [Cond-Mat] (2018).
𝑘𝑥Τ2𝜋3𝐿𝑘𝑦2Τ𝜋3𝐿00.5−0.51−10𝑘𝑧Τ𝜋𝑃1−10ab [25] H. Ge, X. Ni, Y. Tian, S. K. Gupta, M.-H. Lu, X. Lin,
W.-D. Huang, C. T. Chan, and Y.-F. Chen, Experimen-
tal Observation of Acoustic Weyl Points and Topological
Surface States. Phys. Rev. Applied 10, 014017 (2018).
[26] H. He, C. Qiu, L. Ye, X. Cai, X. Fan, M. Ke, F. Zhang,
and Z. Liu, Topological negative refraction of surface
acoustic waves in a Weyl phononic crystal, Nature 560,
61 (2018).
[27] X. Zhang, M. Xiao, Y. Cheng, M.-H. Lu, and J. Chris-
tensen, Communications Physics 1, 97 (2018).
[28] M. Fruchart, S.-Y. Jeon, K. Hur, V. Cheianov, U. Wies-
ner, and V. Vitelli, Soft self-assembly of Weyl materials
for light and sound, PNAS 115, E3655 (2018).
[29] Y.-T. Wang and Y.-W. Tsai, Multiple Weyl and double-
7
Weyl points in an elastic chiral lattice, New J. Phys. 20,
083031 (2018).
[30] P. Wang, F. Casadei, S. H. Kang, and K. Bertoldi, Lo-
cally resonant band gaps in periodic beam lattices by
tuning connectivity, Phys. Rev. B 91, 020103 (2015).
[31] C. S. Lefky, B. Zucker, D. Wright, A. R. Nassar, T. W.
Simpson, and O. J. Hildreth, Dissolvable Supports in
Powder Bed Fusion-Printed Stainless Steel,3D Printing
and Additive Manufacturing 4, 3 (2017).
[32] F. D. M. Haldane, Model for a quantum Hall effect with-
out Landau levels: Condensed-matter realization of the
"Parity Anomaly", Phys. Rev. Lett. 61, 2015 (1988).
|
1910.08619 | 1 | 1910 | 2019-10-18T20:47:39 | Atomically Controlled Tunable Doping in High Performance WSe2 Devices | [
"physics.app-ph",
"cond-mat.mes-hall"
] | Two-dimensional transitional metal dichalcogenide (TMD) field-effect transistors (FETs) are promising candidates for future electronic applications, owing to their excellent transport properties and potential for ultimate device scaling. However, it is widely acknowledged that substantial contact resistance associated with the contact-TMD interface has impeded device performance to a large extent. It has been discovered that O2 plasma treatment can convert WSe2 into WO3-x and substantially improve contact resistances of p-type WSe2 devices by strong doping induced thinner depletion width. In this paper, we carefully study the temperature dependence of this conversion, demonstrating an oxidation process with a precise monolayer control at room temperature and multilayer conversion at elevated temperatures. Furthermore, the lateral oxidation of WSe2 under the contact revealed by HR-STEM leads to potential unpinning of the metal Fermi level and Schottky barrier lowering, resulting in lower contact resistances. The p-doping effect is attributed to the high electron affinity of the formed WO3-x layer on top of the remaining WSe2 channel, and the doping level is found to be dependent on the WO3-x thickness that is controlled by the temperature. Comprehensive materials and electrical characterizations are presented, with a low contact resistance of ~528 ohm-um and record high on-state current of 320 uA/um at -1V bias being reported. | physics.app-ph | physics | Atomically Controlled Tunable Doping in High Performance WSe2 Devices
Chin-Sheng Pang, Terry Y.T. Hung, Ava Khosravi, Rafik Addou, Qingxiao Wang, Moon J.
Kim, Robert M. Wallace, and Zhihong Chen*
C.-S. Pang, T.Y.T. Hung, Prof. Z. H. Chen
Birck Nanotechnology Center
Department of Electrical and Computer Engineering
Purdue University
1205 W State St, West Lafayette, IN 47907, USA
Email: [email protected]
A. Khosravi, Dr. R. Addou, Q. Wang, Prof. M. J. Kim, Prof. R. M. Wallace
Department of Materials Science and Engineering
University of Texas at Dallas
800 West Campbell Road, Richardson, TX 75080, USA
Dr. R. Addou
School of Chemical, Biological, and Environmental Engineering
Oregon State University
Corvallis, OR 93771, USA
Abstract:
Two-dimensional transitional metal dichalcogenide (TMD) field-effect transistors
(FETs) are promising candidates for future electronic applications, owing to their excellent
transport properties and potential for ultimate device scaling. However, it is widely
acknowledged that substantial contact resistance associated with the contact-TMD interface has
impeded device performance to a large extent. It has been discovered that O2 plasma treatment
can convert WSe2 into WO3-x and substantially improve contact resistances of p-type WSe2
devices by strong doping induced thinner depletion width. In this paper, we carefully study the
temperature dependence of this conversion, demonstrating an oxidation process with a precise
monolayer control at room temperature and multilayer conversion at elevated temperatures.
Furthermore, the lateral oxidation of WSe2 under the contact revealed by HR-STEM leads to
potential unpinning of the metal Fermi level and Schottky barrier lowering, resulting in lower
contact resistances. The p-doping effect is attributed to the high electron affinity of the formed
WO3-x layer on top of the remaining WSe2 channel, and the doping level is found to be
1
dependent on the WO3-x thickness that is controlled by the temperature. Comprehensive
materials and electrical characterizations are presented, with a low contact resistance of ~528
m and record high on-state current of 320 A/m at -1V bias being reported.
1. Introduction
Two-dimensional (2D) transitional metal dichalcogenides (TMDs) have attracted wide
attention, owing to their excellent material properties and potential applications in post-
CMOS,1 -- 8 neuromorphic computing,9 -- 11 as well as flexible electronics.12 -- 15 Studying
semiconducting TMDs (e.g. MoS2, WSe2) as the channel material for field-effect transistors
(FETs) is one of the most vital research areas, due to their superior transport properties and
ultra-thin body thickness for ultimate device scaling.16,17 However, how to make good contacts
remains a big challenge for TMD FETs, while it is widely acknowledged that minimizing
contact resistance (RC) is essential to obtain high performance devices and reveal intrinsic TMD
properties.18 -- 22
In general, there are two strategies to optimize current injection at a metal/semiconductor
(MS) interface. One method is to select a metal contact with the preferred work function for
electron or hole injection, given no strong Fermi level pinning at the contact interface, which is
typically not the case for TMDs.23 -- 25 The other method is to dope the semiconductor
degenerately to reduce the depletion width of the MS junction.26 -- 30 We will show in our paper
that by controlling the temperature at which a multi-layer WSe2 FET device is exposed to direct
O2 plasma, the number of top WSe2 layers gets converted into WO3-x (0 < x < 1) can be precisely
controlled, which in turn determines the p-type doping level in the device. Moreover, this
conversion is found to extend into the contact area by tens of nanometers, which can possibly
unpin the Fermi level of the metal contact and dope a small segment of WSe2 underneath the
contact, resulting in reduced contact resistance of the device. The use of controlled oxidation at
2
the Sc/WSe2 interface has also been recently reported to produce an optimized Schottky
junction which can be controlled to exhibit n- or p-type transport.31
In our unique O2 plasma treatment, the top few WSe2 layers are converted into WO3-x that
behaves as a p-type doping layer for the underlying WSe2 due to its high electron affinity.
Different from the previously reported self-limiting oxidation of only the topmost WSe2 layer
in a remote plasma environment,32 a direct O2 plasma is employed in our process. Interestingly,
we found that the doping level can be tuned from non-degenerate to degenerate by increasing
the treatment temperature which directly controls the number of WSe2 layers that get converted
into WO3-x. We further demonstrate low contact resistance of 528 m and a record high hole
current in a scaled WSe2 FET, shedding light on a promising path in the quest for high-
performance electronics. In addition, the achieved p-type doping on WSe2 has excellent air
stability, precise doping level control, and is an industry compatible process.
2. Results and Discussions
Two types of Schottky barrier (SB) device structures (2 or 4-probe) were implemented,
as shown schematically in Figure 1a, b. A scanning electron microscope (SEM) of one of our
4-probe devices is shown in Figure 1c with dimension being labelled. The critical fabrication
processes are described in the Experimental Section. High resolution scanning transmission
electron microscope (HR-STEM) image and electron energy loss spectroscopy (EELS) line
scan across a multi-layer WSe2 flake underneath of a Ti/Pd contact are shown in Figure 1d.
Band diagrams and working principles of SB devices are illustrated in the insets of Figure 1e.
It is known that the Fermi level of metal contact is pinned close to the mid-gap of WSe2.33,34
Therefore, band movements modulated by scanning the back-gated voltage (VB) lead to either
hole injection from the source to the valence band at negative VB or electron injection from the
drain to the conduction band at positive VB. Consequently, transfer characteristics of a pristine
WSe2 device exhibit a typical ambipolar behavior, as shown in Figure 1e.
3
2.1. Material Analyses
Different samples were exposed to a direct O2 plasma at various conditions. Simply
comparing color contrast of the samples before and after exposure under an optical microscope,
it was rather clear that process temperature is more effective than plasma power or exposure
time in controlling the WSe2 oxidation process. Optical images and details are presented in
Section I (Supporting Information). Raman spectra of three sets of pristine bi-layer and tri-layer
CVD WSe2 samples are presented in the top panels of Figure 2a. Consistent with previous
reports,35,36 the 1B2g Raman mode at 310 cm-1 only appears in multi-layers and bulk WSe2 but
does not show up in monolayers. In the bottom left panel of Figure 2a, the 1B2g peak vanished
after the room temperature (RT) O2 plasma treatment on all three bi-layer samples, indicating
that only the bottom WSe2 layer was left while the top layer was oxidized. In contrast, the 310
cm-1 peak remained in the spectra for all three tri-layer samples (right bottom panel), suggesting
that only the top layer was converted to oxide while the bottom two layers were intact after the
RT treatment. Interestingly, when the temperature was raised to 150 oC, no Raman peaks were
observed in bilayer samples after the treatment. We conclude that both layers were converted
into WO3-x. Therefore, different from the RT treatment, more than one layer of WSe2 can be
converted by O2 plasma at elevated temperatures.
The surface chemistry alteration of CVD WSe2 flakes upon O2 plasma exposure at
different temperatures was investigated by x-ray photoelectron spectroscopy (XPS). Detailed
information can be found in the Experimental Section. Figure 2b, c show W 4f and Se 3d core
level spectra of a pristine WSe2 flake and following O2 plasma exposure for 60s at RT, 90 °C,
150 °C, and 250 °C. The dominant XPS signal originated from the substrate (SiO2/Si) due to
the lateral size and thickness of WSe2 flakes is shown in Section II (Supporting Information).
The XPS scan of pristine WSe2 flake did not reveal any additional chemical states such as W-
O or Se-O. In Figure 2b, after O2 plasma exposure at RT, an additional chemical state
4
corresponding to W-O chemical bond was detected at 36.2 eV in W 4f core level. No additional
chemical states were detected in Se 3d core level, indicating that oxygen did not react with
selenium. Similar to the O2 plasma treatment at RT, the W-O chemical state was detected in W
4f chemical state following the treatment at 90 °C, 150 °C and 250 °C with increasing intensity.
Throughout the treatment at different temperatures, the oxide species were below the XPS
detection limit in Se 3d core level spectra. Figure 2d presents an increase in the percentage of
oxidized W atoms on the top few WSe2 layers after O2 plasma treatment at different
temperatures. In Figure 2e, the selenium to tungsten ratio of the WSe2 flakes was calculated
using the integrated intensity of XPS core levels and corresponding sensitivity factor. It shows
that the Se/W ratio decreased gradually from 2.1 in pristine flakes to 0.6 after the treatment at
250 °C, suggesting that the density of W-O chemical states depends critically on the treatment
temperature. The W 4f and Se 3d core level binding energies after each treatment are shown in
Figure 2f. The red shift of binding energy was detected in both W4f and Se 3d core level spectra
in all treatment temperatures, indicating different levels of p-doping. Lower binding energy
suggests that more prominent doping effect can be achieved at higher temperature
treatment.37,38
The doping effect in WSe2 was also examined by Raman spectroscopy. Figure 2g
compares the Raman spectra of WSe2 flakes before and after the O2 plasma treatment at 250
°C. The degenerate E1
2g/A1g vibrational mode at 250.8 cm-1 and the higher wave number peak
2LA(M) at 259.0 cm-1 as a Raman fingerprint of WSe2 were detected. A clear blue shift of
E1
2g/A1g and 2LA(M) was observed after the 250 °C treatment. According to previous
studies37,39, the blue shift (~ 1.3 cm-1) in the E1
2g/A1g peak is correlated to p-type doping effect
in WSe2 flakes, which is consistent with our XPS analysis and electrical characterizations
discussed in the later paragraph. Therefore, we have confirmed that both atomically precise
layer control and doping level modulation can be achieved through different treatment
temperatures.
5
HR-STEM measurements were performed to directly quantify the number of WSe2
layers being converted to oxide by the 250 °C O2 plasma treatment. Detailed information of
HR-STEM can be found in the Experimental Section. Figure 3a shows a cross-section view of
a WSe2 device with a channel length (LG) of 65nm. The observed bending curvature was caused
by the carbon layer deposition induced stress during the TEM sample preparation using focused
ion beam (FIB). From the magnified image of the channel region presented in Figure 3b, nearly
three layers of WSe2 were converted to WO3-x by the 250 °C treatment, leaving two WSe2 layers
remaining underneath. Interestingly, it was observed that WO3-x penetrated laterally into the
contact at the scale of ~12nm, as revealed in Figure 3c. We believe this phenomenon contributes
significantly to unpinning of the metal Fermi level and potentially lowering of the SB height
for easier hole injection. Furthermore, WO3-x induced heavy doping in the remaining WSe2
under the contact can effectively reduce the depletion width to produce a transparent barrier for
carrier injection into the channel. Both mechanisms contribute to a very low contact resistance
of 528 m and correspondingly record high on-state performance reported in the later
paragraph. More detailed information regarding the EELS mapping can be found in Section III
(Supporting Information).
2.2. Electrical Characterizations
We now focus on electrical characterization of devices that have undergone O2 plasma
treatment at different temperatures, as shown in Figure 4a. One could immediately observe
significant differences in the magnitude of the threshold voltage (VTH) shift from the pristine
(black) to after treatment (red) characteristics. It is clear that VTH shift, an indication of the
doping level, increases with increasing temperature, consistent with the shift of binding energy
shown in Figure 2f. We believe that the higher doping level achieved at higher temperature can
be attributed to a larger number of WSe2 layers being converted into a thicker WO3-x layer.
Except for the device treated at 250 °C, VTH of the other two devices treated at RT and 150 °C
6
is within the voltage window to reveal the off-state performance. The preserved on/off ratios of
~107 indicate that WO3-x serves as an effective doping layer rather than a conductive layer
shunting between the S/D electrodes.40 Based on the VTH shift extracted from the device
characteristics, we can calculate the amount of charges induced by doping (Q = CoxVTH_Shift,
Cox = 38.5 nF/cm2) and estimate the dopant concentration to be ~ 2.2x1012 cm-2 for devices
undergone the RT treatment and ~ 8.3 1012 cm-2 for those gone through the treatment at 150 oC.
More devices data set can be found in Section IV (Supporting Information)
In addition, device contact resistances (RC) were significantly reduced after the plasma
treatment. 4-probe configuration shown in Figure 1b was used to extract RC. Figure 4b, c show
RC as a function of back gate voltage for pristine devices and the same set of devices after the
O2 plasma treatment at 150 °C and 250 °C, respectively. Due to the positive VTH shift, RC is less
gate voltage dependent and reduced drastically. RC of a device gone through the 250 oC
treatment was extracted to be as low as 528 m, at VBG = -50V. As explained earlier, we
believe that the O2 plasma treatment not only affects the channel doping but also lowers the SB
height and barrier width at the contact interface to allow for higher current injection, which is
now attributed to the lateral penetration of WO3-x as observed from our HR-STEM analysis. A
table of extracted RC values from different devices with or without treatment can be found in
Section V (Supporting Information).
Utilizing the demonstrated doping and low RC, we fabricated devices with scaled
channel length (LG) and achieved the outstanding on-state performance in WSe2 FETs. Twelve
SB-devices with different LG ranging from ~70 nm to ~1050 nm were fabricated on exfoliated
multilayer (5-10 layers) WSe2. 250 oC O2 plasma treatment was performed to all devices. Total
device resistance (Rtotal) and current density (IDS) extracted at VB = -50V and VDS = -0.9V are
shown in Figure 5a. Although these devices were not fabricated on the same flake to guarantee
an accurate extraction of RC from the transmission line method (TLM), we still performed the
extraction to get a rough estimate. RC ~ 1.1 k m was extracted from the linear fitting of the
7
Rtotal vs LG plot, which agrees with the values obtained from the 4-probe measurements,
presented in Table S1 (Supporting Information). The output characteristics of our best
performing device with LG ~ 70nm is shown in Figure 5b with IDS = 320 A/m being achieved
at VDS = -1V. Finally, these O2 plasma treated devices were placed in the laboratory ambient
environment without any passivation layer for 7 days before re-measurements. Negligible
changes in characteristics were observed as shown in Section VI (Supporting Information),
suggesting a robust p-doping scheme for high performance WSe2 devices.
Finally, we compare our result to other reported contact resistance for hole injection in
WSe2-based devices,22,26 -- 28,41 -- 44 and summarize in Table 1. Our O2 plasma treatment offers a
comparable RC while requiring a simpler fabrication process compared to a 2D/2D contact.22
3. Conclusion
We have achieved tunable p-type doping on WSe2 through O2-plasma treatment at
different temperatures, with supporting evidences from XPS, Raman, HR-STEM and electrical
characteristics. We conclude that the doping level is determined by how many WSe2 layers
being converted into WO3-x by the O2-plasma treatment. Larger number of layers are converted
at higher temperatures, resulting in thicker WO3-x for higher doping. The penetration of WO3-x
into the contact region is believed to contribute to the unpinning of the Fermi level and thinning
of the barrier width for higher current injection. Low RC ~528 m was measured from 4-
probe measurements after 250 oC O2 plasma treatment, leading to a record-high hole current in
WSe2 devices. This air-stable, efficient p-doping strategy can enable high performance WSe2-
based electronics or be applied to other material of interests for tunable doping effect by
transferring WSe2 on top followed by self-limiting oxidation under specific temperature
treatment.
4. Methods
8
CVD Flakes Transferring Process: Mono/bi/tri-layer CVD WSe2 flakes were
purchased from 2D Layer (https://2dlayer.com/). To transfer onto a desired substrate,
polystrene (PS) was used as the supporting film to peel off the WSe2 flakes from the growth
substrate. 9 g of PS (Molecular weight ∼192 000 g/mol) was dispersed in 50 mL toluene. Then
this solution was spin-coated on the growth substrate at a speed of 4000 rpm for 40 secs,
followed by baking at 90 oC for 5 mins. In order to allow water to penetrate to the interface
between the WSe2 film and the SiO2/Si substrate to detach the WSe2 flakes, a diamond scribe
was used to make some scratches at the edges of the PS film. Next, the PS film attached to the
WSe2 flakes was gradually peeled off from the growth substrate in water and transferred to the
target substrate. Finally, the PS film was removed by soaking in toluene, acetone and IPA.
2-probe and 4-probe Device Fabrication: Mono/bi-layer CVD WSe2 flakes were
transferred or multilayer layers (5-10) WSe2 was exfoliated from a bulk crystal onto a 90nm
SiO2 capped p++ doped Si substrate as a global back-gated scheme. E-beam lithography was
employed to define source/drain (S/D) regions (for 2-probe devices) and two additional voltage
probes (for 4-probe devices) followed by e-beam evaporated Ti (1nm) / Pd (30nm) (at pressure
~ 1E-6 torr) as electrodes and a PMMA lift-off process.
Electrical Characterization: The electrical measurements were performed by HP 4156B
precision semiconductor parameter analyzer with Lake Shore probe station under vacuum at
room temperature.
X-ray Photoelectron Spectroscopy: XPS scans were carried out in an Ultra High
Vacuum (UHV) cluster tool using an Omicron EA125 hemispherical 7 channels analyzer with
a monochromatic Al Kα source (hν= 1486.7 eV).45 XPS peaks were deconvoluted and analyzed
using Aanalyzer software.46 Quantitative analysis of the elemental concentration of the samples
is acquired based on integrated photoelectron intensity and sensitivity factor for a given core
level.
9
Quantitative analysis of relative elemental concentration can be determined from XPS
measurement. The number of photoelectrons per second for specific elemental core level (I) is
directly proportional to the number of atoms of the elements per centimeter cubic of the sample
surface (n), while it is indirectly proportional to atomic sensitivity factor (S). Atomic sensitivity
factor (S) of the elements core level is developed from a specific spectrometer. Thus, the
elemental concentration (Cx) is described as: Cx =
nx
ni
=
Ix
⁄
Sx
∑Ii
⁄
Si
.
Raman Spectroscopy: Raman spectra were taken using a 532 nm laser focused through
a 100X objective lens at room temperature and under ambient condition.
TEM Analysis: TEM analysis is carried out using an aberration-corrected (probe Cs-
corrector) JEM-ARM200F (JEOL. USA. Inc.) electron microscope operated at 200 kV. The
high angle annular dark field (HAADF) and annular bright field (ABF) images are performed
to study the cross-sectional morphology of the device. The convergence semi-angle of the
electron probe is set to 25 mrad, and the collection semi-angle is 70-250 mrad for HAADF and
12-24 mrad for ABF, respectively. The elemental characterization of the device is performed
using the electron energy loss spectroscopy (EELS) line scan using a Gatan Enfina spectrometer
with the collection semi-angle for EELS for 30 mrad.
10
References
1.
Jena, D. Tunneling transistors based on graphene and 2-D Crystals. Proc. IEEE 101,
1585 (2013).
2.
Seabaugh, B. A. C. &Zhang, Q. Low-Voltage Tunnel Transistors for Beyond CMOS
Logic. Proc. IEEE 98, 2095 (2010).
3. Müller, M. R. et al. Gate-Controlled WSe2 Transistors Using a Buried Triple-Gate
Structure. Nanoscale Res. Lett. 11, 512 (2016).
4.
Pang, C.-S., Ilatikhameneh, H. &Chen, Z. Gate tunable 2D WSe2 Esaki diode by SiNx
doping. in Device Research Conference - Conference Digest, DRC 1 -- 2 (2017).
doi:10.1109/DRC.2017.7999450
5.
Sarkar, D. et al. A subthermionic tunnel field-effect transistor with an atomically thin
channel. Nature 526, 91 (2015).
6.
Pang, C.-S. et al. WSe 2 Homojunction Devices : Electrostatically Configurable as
Diodes , MOSFETs , and Tunnel FETs for Reconfigurable Computing. Small 1902770
(2019). doi:10.1002/smll.201902770
7.
Pang, C.-S., Thakuria, N., Gupta, S. K. &Chen, Z. First Demonstration of WSe 2 Based
CMOS-SRAM. in IEEE Int. Electron Devices Meet. 22.2.1-22.2.4 (IEEE, 2018).
doi:10.4231/D3ZC7RV9X
8.
Pang, C.-S. &Chen, Z. First Demonstration of WSe 2 CMOS Inverter with Modulable
Noise Margin by Electrostatic Doping. in Device Research Conference - Conference
Digest, DRC 1 -- 2 (IEEE, 2018). doi:10.1109/DRC.2018.8442258
9.
Paul, T., Ahmed, T., Tiwari, K. K. &Thakur, C. S. A high-performance MoS 2 synaptic
device with floating gate engineering for neuromorphic computing A high-performance
MoS 2 synaptic device with floating gate engineering for neuromorphic computing. 2D
Mater. 6, 045008 (2019).
11
10.
Seo, S. et al. Artificial optic-neural synapse for colored and color-mixed pattern
recognition. Nat. Commun. 9, 5106 (2018).
11. Xie, D., Hu, W. &Jiang, J. Bidirectionally-trigged 2D MoS 2 synapse through
coplanar-gate electric- double-layer polymer coupling for neuromorphic
complementary spatiotemporal learning. Org. Electron. 63, 120 (2018).
12. Gao, L. Flexible Device Applications of 2D Semiconductors. Small 13, 1603994
(2017).
13. Lin, P. et al. Piezo-Phototronic Effect for Enhanced Flexible MoS 2 / WSe 2 van der
Waals Photodiodes. Adv. Funct. Mater. 28, 1802849 (2018).
14. Yu, X., Prévot, M. S., Guijarro, N. &Sivula, K. Self-assembled 2D WSe 2 thin films
for photoelectrochemical hydrogen production. Nat. Commun. 6, 7596 (2015).
15. Akinwande, D., Petrone, N. &Hone, J. Two-dimensional flexible nanoelectronics. Nat.
Commun. 5, 5678 (2015).
16. Desai, S. B. et al. MoS 2 transistors with 1-nanometer gate lengths. Science (80-. ).
354, 99 (2016).
17. Nourbakhsh, A. et al. MoS 2 Field-Effect Transistor with Sub-10 nm Channel Length.
Nano Lett. 16, 7798 (2016).
18. Cui, X. et al. Multi-terminal transport measurements of MoS2 using a van der Waals
heterostructure device platform. Nat. Nanotechnol. 10, 534 -- 540 (2015).
19. Leong, W. S. et al. Low Resistance Metal Contacts to MoS2 Devices with Nickel-
Etched-Graphene Electrodes. ACS Nano 869 -- 877 (2015). doi:10.1021/nn506567r
20.
Smithe, K. K. H., English, C. D., Suryavanshi, S.V &Pop, E. Intrinsic electrical
transport and performance projections of synthetic monolayer MoS 2 devices Intrinsic
electrical transport and performance projections of synthetic monolayer MoS 2 devices.
2D Mater. 4, 011009 (2017).
12
21. English, C. D., Shine, G., Dorgan, V. E., Saraswat, K. C. &Pop, E. Improved Contacts
to MoS 2 Transistors by Ultra-High Vacuum Metal Deposition. Nano Lett. 16, 3824
(2016).
22. Chuang, H. J. et al. Low-Resistance 2D/2D Ohmic Contacts: A Universal Approach to
High-Performance WSe 2 , MoS 2 , and MoSe 2 Transistors. Nano Lett. 16, 1896 -- 1902
(2016).
23. Kim, C. et al. Fermi Level Pinning at Electrical Metal Contacts of Monolayer
Molybdenum Dichalcogenides. ACS Nano 11, 1588 (2017).
24. Gong, C., Colombo, L., Wallace, R. M. &Cho, K. The Unusual Mechanism of Partial
Fermi Level Pinning at Metal − MoS2 Interfaces. Nano Lett. 14, 1714 (2014).
25. Bampoulis, P. et al. Defect Dominated Charge Transport and Fermi Level Pinning in
MoS2 / Metal Contacts. ACS Appl. Mater. Interfaces 9, 19278 (2017).
26. Yamamoto, M., Nakaharai, S., Ueno, K. &Tsukagoshi, K. Self-Limiting Oxides on
WSe 2 as Controlled Surface Acceptors and Low-Resistance Hole Contacts. Nano Lett.
16, 2720 -- 2727 (2016).
27. Cai, L. et al. Rapid flame synthesis of atomically thin MoO3 down to monolayer
thickness for effective hole doping of WSe2. Nano Lett. 17, 3854 (2017).
28.
Fang, H. et al. High-Performance Single Layered WSe 2 p-FETs with Chemically
Doped Contacts. Nano Lett. 12, 3788 (2012).
29. Chen, K. et al. Air stable n-doping of WSe2 by silicon nitride thin films with tunable
fixed charge density. Apl Mater. 2, 092504 (2014).
30. Hung, T. Y. T., Pang, C.-S., Liu, X., Zemlyanov, D. &Chen, Z. Atomically Thin p-
doping Layer and Record High Hole Current on WSe2. in Device Research Conference
- Conference Digest, DRC 1 -- 2 (2019).
13
31.
Smyth, C. M. et al. Engineering the interface chemistry for scandium electron contacts
in WSe 2 transistors and diodes Engineering the interface chemistry for scandium
electron contacts in WSe 2 transistors and diodes. 2D Mater. 6, 045020 (2019).
32.
Pudasaini, P. R. et al. High-performance multilayer WSe2field-effect transistors with
carrier type control. Nano Res. 11, 722 (2018).
33.
Smyth, C. M. et al. Engineering the Palladium − WSe 2 Interface Chemistry for Field
E ff ect Transistors with High-Performance Hole Contacts. ACS Appl. Nano Mater. 2,
75 (2019).
34. Das, S. &Appenzeller, J. WSe2 field effect transistors with enhanced ambipolar
characteristics. Appl. Phys. Lett. 103, 103501 (2013).
35. Luo, X. et al. Effects of lower symmetry and dimensionality on Raman spectra in two-
dimensional WSe 2. Phys. Rev. B 88, 195313 (2013).
36. Li, Z. et al. Layer Control of WSe2 via Selective Surface Layer Oxidation. ACS Nano
10, 6836 -- 6842 (2016).
37. Khosravi, A. et al. Covalent nitrogen doping in molecular beam epitaxy-grown and
bulk WSe2. APL Mater. 6, 026603 (2018).
38. Addou, R. et al. One dimensional metallic edges in atomically thin WSe 2 induced by
air exposure. 2D Mater. 5, 025017 (2018).
39. Desai, S. B., Seol, G., Kang, J. S., Fang, H. &Battaglia, C. Strain-Induced Indirect to
Direct Bandgap Transition in Multilayer. Nano Lett. 14, 4592 (2014).
40. Liu, B. et al. High-Performance WSe 2 Field-E ff ect Transistors via Controlled
Formation of In- Plane Heterojunctions. ACS Nano 10, 5153 (2016).
41.
Jung, Y. et al. Transferred via contacts as a platform for ideal two-dimensional
transistors. Nat. Electron. 2, 187 -- 194 (2019).
14
42. Chien, P. Y. et al. Reliable doping technique for WSe2 by W:Ta co-sputtering process.
in 2016 IEEE Silicon Nanoelectronics Workshop, SNW 2016 58 -- 59 (2016).
doi:10.1109/SNW.2016.7577984
43. Zhao, P. et al. Air stable p-doping of WSe2 by covalent functionalization. ACS Nano 8,
10808 -- 10814 (2014).
44. Zhang, R., Drysdale, D., Koutsos, V. &Cheung, R. Controlled Layer Thinning and p-
Type Doping of WSe2 by Vapor XeF2. Adv. Funct. Mater. 27, 1702455 (2017).
45. R.M.Wallace. In-Situ Studies of Interfacial Bonding of High-N Dielectrics for CMOS
Beyond 22nm. ECS Trans. 16, 255 (2008).
46. Herrera-Go´mez, A., Hegedus, A. &Meissner, P. L. Chemical depth profile of ultrathin
nitrided films. Appl. Phys. Lett. 81, 1014 (2002).
Acknowledgements
C.-S. P. And T.Y.T. H. contributed equally to this work.
C.-S. P., T.Y.T. H., A. K., R. A., R. M. W., and Z. C. acknowledge financial support from
NEWLIMITS, a center in nCORE, a Semiconductor Research Corporation (SRC) program
sponsored by NIST through award number 70NANB17H041. M. J. K. was supported in part
by the Louis Beecherl, Jr. Endowment Funds, Global Research and Development Center
Program (2018K1A4A3A01064272) and Brain Pool Program (2019H1D3A2A01061938)
through the National Research Foundation of Korea (NRF) funded by the Ministry of Science
and ICT.
Author Contributions
Z. C. conceived and managed the project. C.-S. P and T. H. fabricated the devices, perform the
O2-plasma treatment, and conducted the electrical and Raman measurements. A. K. and R. A.
15
performed the XPS measurement and analysis. Q. W. performed the FIB on samples for
HRTEM analysis. All authors wrote and revised the manuscript.
Competing Interests Statement
The authors declare no competing interests.
16
(a) (b) (c)
(d)
(e)
Figure 1. Schematics illustrations of a) 2-probe and b) 4-probe (type 2) WSe2 devices. c)
SEM image of a 4-probe device with dimensions labelled. d) HR-STEM and EELS line scan
across the SiO2/WSe2/contact region. e) Transfer characteristics of a WSe2 SB device showing
ambipolar behaviors with gate-dependent electron/hole injection.
17
(a)
(b) (c)
(d)
(e)
18
(f)
(g)
Figure 2. a) Comparison of Raman spectra between pristine bi-layer/tri-layer WSe2 and after
RT O2 plasma treatment. b) W 4f and c) Se 3d core level spectra of pristine WSe2 and after O2
plasma treatment at RT, 90 oC, 150 oC, or 250 oC. d) Percentage of oxidized W atoms and e) Se
to W ratio on the top few layers in pristine WSe2 and after O2 plasma treatment at various
temperatures. f) The red shift of the binding energy calculated from b, c after O2 plasma
treatment at different temperatures. g) Raman spectra of WSe2 before and after treatment at 250
oC. The blue shift of E1
2g/A1g and 2LA(M) indicate p-doping effect in WSe2.
19
(a)
(b)
(c)
Figure 3. a) Cross-section view of a WSe2 device with LG = 65nm. b) The zoom-in observation
of the channel region from a, indicating ~ 3 layers of WSe2 is converted into WO3-x. c) The
lateral penetration of WO3-x underneath the edge of contact after the O2-plasma treatment.
20
(a)
(b)
(c)
Figure 4. a) Transfer characteristics comparison of pristine WSe2 FETs and after O2 plasma
treatment at RT, 150 oC, and 250 oC. The comparison of RC between pristine and after O2 plasma
treatment at b) 150 oC and c) 250 oC using 4-probe measurements.
21
(a)
(b)
Figure 5. a) Measured Rtotal (black data) and current density (red data) at VB = -50V and VDS =
-0.9V for 12 devices with LG ranging from ~70nm to ~1050nm. RC ~ 1.14 k m is extracted
from a linear fitting curve. b) Output characteristics of a device with record-high hole current
density of 320 A/um at VDS = -1V after 250 oC O2 plasma treatment.
22
Table 1. RC comparison for WSe2 hole injection at room temperature.
Layer Number [ref]
RC (kΩ*μm)
Special Treatment
10L 22
6L 27
4L 26
10L 28
2L 41
20L 42
13L 41
1L 43
1L 41
7L 44
~8L (this work)
0.3
0.9
1.1
1.3
4.0
11.4
12.5
38.0
50.0
100
0.5
2D/2D Contact
MoO3 Passivation
Ozone Treatment
NO2 Treatment
Transferred Contact
W:Ta Co-sputtering
Transferred Contact
NO2 Treatment
Transferred Contact
XeF2 Thinning
O2-plasma at 250 oC
23
|
1711.08444 | 1 | 1711 | 2017-10-24T08:29:45 | THz electrical writing speed in an antiferromagnetic memory | [
"physics.app-ph"
] | The speed of writing of state-of-the-art ferromagnetic memories is physically limited by an intrinsic GHz threshold. Recently, an alternative research direction has been initiated by realizing memory devices based on antiferromagnets in which spin directions periodically alternate from one atomic lattice site to the next. In our work we experimentally demonstrate at room temperature that the speed of reversible electrical writing in a memory device can be scaled up to THz using an antiferromagnet. Efficient current-induced spin-torque mechanism is responsible for the switching in our memory devices throughout the twelve orders of magnitude range of writing speeds from Hz to THz. Our work opens the path towards the development of memory-logic technology reaching the elusive THz band. | physics.app-ph | physics |
THz electrical writing speed in an antiferromagnetic
memory
K. Olejn´ık,1,# T. Seifert,2 Z. Kaspar,1,3 V. Nov´ak,1 P. Wadley,4 R. P. Campion,4
M. Baumgartner,5 P. Gambardella,5 P. Nemec,3 J. Wunderlich,1,6
J. Sinova,1,7 M. Muller,2 T. Kampfrath,2 T. Jungwirth1,4
1Institute of Physics, Academy of Sciences of the Czech Republic,
Cukrovarnick´a 10, 162 00 Praha 6, Czech Republic
1Fritz Haber Institute of the Max Planck Society, 14195 Berlin, Germany
3Faculty of Mathematics and Physics, Charles University,
Ke Karlovu 3, 121 16 Prague 2, Czech Republic
4School of Physics and Astronomy, University of Nottingham,
Nottingham NG7 2RD, United Kingdom
5Department of Materials, ETH Zurich, Honggerbergring 64,
CH-8093 Zurich, Switzerland
6Hitachi Cambridge Laboratory, J. J. Thomson Avenue,
Cambridge CB3 0HE, United Kingdom
7Institut fur Physik, Johannes Gutenberg Universitat Mainz,
55128 Mainz, Germany
#To whom correspondence should be addressed; E-mail: [email protected].
November 23, 2017
The speed of writing of state-of-the-art ferromagnetic memories is physically
limited by an intrinsic GHz threshold. Recently, an alternative research direc-
tion has been initiated by realizing memory devices based on antiferromagnets
in which spin directions periodically alternate from one atomic lattice site to
1
the next. In our work we experimentally demonstrate at room temperature
that the speed of reversible electrical writing in a memory device can be scaled
up to THz using an antiferromagnet. Efficient current-induced spin-torque
mechanism is responsible for the switching in our memory devices through-
out the twelve orders of magnitude range of writing speeds from Hz to THz.
Our work opens the path towards the development of memory-logic technol-
ogy reaching the elusive THz band.
Magnetic random access memories[1, 2, 3] represent the most advanced example of spin-
tronic devices that are foreseen to become the leading alternatives to CMOS for the "beyond
Moore's law" information technologies[4]. Among the unresolved fundamental problems in
spintronics is the electrical writing speed. Whether using Oersted fields or advanced spin-
torque mechanisms[5, 6, 7, 8, 2, 9, 10], the writing speed has a physical limit in ferromagnetic
memories in the GHz range beyond which it becomes prohibitively energy-costly[10, 11]. The
interest in antiferromagnetic memories is driven by the vision of ultra-fast operation far ex-
ceeding the GHz range [12]. Recently, THz writing speed has become a realistic prospect with
the experimental discovery[13] of the electrical switching in the CuMnAs antiferromagnet by
a staggered spin-torque field[14] at ambient conditions. This was followed by upscaling of the
writing speed to the GHz range, and by demonstrating a fabrication compatibility with III-V
semiconductors or Si, and a device compatibility with common microelectronic circuitry[15].
These initial experiments verified several unique features of antiferromagnetic bit cells, in-
cluding their magnetic field hardness, absence of fringing stray fields, and neuron-like multi-
level memory-logic functionality[13, 15, 16]. The results have been already replicated in an-
other suitable antiferromagnet Mn2Au[17, 18]. However, the envisaged THz electrical writing
speed[14, 16, 19, 12] in antiferromagnetic memories has not been experimentally demonstrated
prior to this work.
2
Switching mechanism and measurement setup. The principle of the current-driven anti-
ferromagnetic switching is illustrated in Fig. 1A which shows the crystal and magnetic struc-
ture of CuMnAs, a prototypical antiferromagnetic compound with a high N´eel temperature[20].
The Mn spin-sublattices with opposite magnetic moments (thick red/purple arrows in Fig. 1A)
occupy non-centrosymmetric crystal sites that are inversion partners. The local symmetry prop-
erties of the lattice together with the spin-orbit coupling imply that a global electrical cur-
rent (black arrow) driven through the crystal generates local, oppositely-oriented carrier spin-
polarizations (thin red/purple arrows) at the inversion partner sites, with the polarization axis
perpendicular to the applied current[14, 13, 21]. This alternating non-equilibrium polarization
acts as an effective staggered magnetic field on the antiferromagnetic moments. The strength of
the staggered field is proportional to the current-induced polarization and to the exchange cou-
pling between the carrier spins and the antiferromagnetic moments[14, 13, 21]. The physics is
analogous to the highly efficient spin-orbit torque switching mechanism in ferromagnets[6, 7, 8]
whose writing speed is, however, limited by the GHz threshold[9, 10].
To establish the feasibility in antiferromagnets to extend the writing speeds to the THz
band, we compare our ultra-short writing pulse experiments to the results obtained with longer
writing pulses in the same device structures. We first recall the previously established technique
that has enabled the scaling of the writing pulse time τp down to 250 ps, corresponding to the
writing speed 1/τp of up to 4 GHz[20, 15]. In this scheme, the current pulses are delivered via
wire-bonded contacts, and here τp ∼ 100 ps is at the limit achievable with common current-
pulse setups. The reversible switching is realized in cross-shape bit cells by alternating current
pulses along one of the two orthogonal directions, as illustrated by white dashed lines on the
electron microscopy image of the cell in Fig. 1B. White double-arrows depict the corresponding
orthogonal N´eel vector orientations preferably set by the two current directions. Apart from
the reversible switching controlled by alternating the two orthogonal writing currents, earlier
3
studies have also shown that multiple-pulses can be applied successively along one writing path,
revealing a neuron-like multi-level bit-cell characteristics typical of antiferromagnets[13, 15,
22, 23]. This has been associated with multi-domain reconfigurations[24] and we will exploit
the feature also in our ps-pulse experiments described below.
For our experiments with 1/τp in the THz range, we use the same cross-shape bit cell and
an analogous experimental setup. However, as illustrated in Fig. 1D, we employ a non-contact
technique for generating the ultra-short current pulses in the memory cell to overcome the above
limit of common current-pulse setups. To explore reversible writing with pulses of τp ≈ 1 ps,
we apply THz electro-magnetic transients whose linear polarization can be chosen along two
orthogonal directions, as sketched in Fig. 1D. The real wave-form of the electric-field transient
is plotted in Fig. 1E [25]. As in the contact writing scheme, white double-arrows depict the
corresponding orthogonal N´eel vector orientations preferably set by the two orthogonal polar-
izations.
In both contact and non-contact setups, the electrical readout is performed by detecting the
antiferromagnetic transverse anisotropic magnetoresistance (AMR), as illustrated in Fig. 2[14,
13, 22]. Here the readout current direction is depicted by the white dashed line. The transverse
AMR has an opposite sign for the two orthogonal N´eel vector directions. To confirm the AMR
symmetry, we employed two detection geometries in which we interchanged the readout current
and transverse-voltage axes (see Fig. 2). For the AMR, the readout signal flips sign between the
two geometries[13]. Note that the AMR readout signal does not depend on the polarity of the
switching current. In the present experiments we applied a bipolar wave-form in the contact,
larger pulse-time experiments to explicitly highlight the correspondence to the non-contact, ps-
pulse measurements (cf. Figs. 1C,E). All experiments are performed at room temperature.
Experimental results.
In Fig. 2A we first present measured data for µs writing pulses
delivered by the contact method in a CuMnAs/GaAs bit cell. The pulse-train of one current-
4
direction is turned on for 30 s, then the train is turned off for 30 s followed by turning on for
30 s the pulse-train with the orthogonal current direction. The data show the phenomenology
attributed in the earlier studies to the multi-level switching of the antiferromagnet by the current-
induced staggered spin-orbit field[13, 15]. The readout signal increases as the successive pulses
within a train arrive at the bit-cell. When the pulse-train is turned off, the readout signal of
the memory cell partially relaxes. (Note that the retention stability can be broadly varied by
changing the antiferromagnetic structure parameters[15].) The sign of the signal is switched
when applying the pulse-train with the orthogonal current direction and the overall sign of
these reversible switching traces flips between the two readout geometries, consistent with the
AMR symmetry. Data in Fig. 2A were obtained for an applied writing current density j =
3 × 107 Acm−2 and a writing pulse repetition rate of 1 Hz. Earlier systematic studies of the
multi-level switching in these antiferromagnetic memory devices showed that the magnitude of
the signal depends, apart from the number of pulses, on the magnitude of the writing current
density and the pulse time[13, 15].
Remarkably, analogous reversible switching traces, with an initial steep increase of the
AMR signal followed by a tendency to saturate, can be written in the same CuMnAs/GaAs
memory cell structure by the ps-pulses, as shown in Fig. 2B. Here the writing pulse repetition
rate was set to 1 kHz and the current density, recalculated from the applied intensity of the
THz electric-field E = 1.1 × 105 Vcm−1, was increased to j ≈ 2.7 × 109 Acm−2 for these
ultra-short pulses. (The E to j conversion is discussed in detail below.) The correspondence
between the measured data in Figs. 2A,B indicates that for the ps-pulses the antiferromagnetic
switching is also due to the current-induced staggered spin-orbit field. Note that this switching
mechanism allows us to employ the electric-field transient and that we do not rely on the weak
magnetic-field component of the radiation,[26] or on non-linear orbital-transition effects[27].
To resolve the change in the AMR readout signal induced by a single ps-pulse, we increased
5
the current density to j ≈ 2.9× 109 Acm−2 and reduced the repetition rate of the writing pulses
to 125 Hz and matched it closely to the readout repetition rate (100 Hz). The measured data
plotted as a function of the pulse number are shown in Fig. 3A. We observe that the initial ps-
pulse accounts for a sizable portion of the total signal generated by the pulse-train. (Note that
the scatter in the measured data is likely of an instrumental origin due to the electrical noise
from the laser setup.)
In Fig. 3B we show corresponding measurements with the µs-pulses
which again highlight the analogous phenomenology of the THz-speed writing and the slower
writing in the multi-level antiferromagnetic bit cell.
The electrical current density generated in our cells for a given amplitude of the ps electric-
field pulse (see Fig. 1E) could not be directly measured. In the simplest approach, one can,
according to the Ohm's law, divide the THz field by the CuMnAs resistivity. This would result
in a device-geometry independent current density of j = 8 × 107 Acm−2 per peak incident
THz field of 105 Vcm−1. (Here we accounted for the factor of 10 reduction of the THz field in
the material due to reflection.) However, when plotting in Fig. 4A the THz-induced switching
signal as a function of the incident THz peak field, we find different dependencies in three
devices with the width of the central cross of 1, 2, and 3 µm, and with otherwise the same
geometry of the Au-contact pads and approximately the same resistance of the three devices.
We ascribe this observation to the Au electrodes which strongly modify the incident THz
field in the cross region. As shown previously,[28, 29] a THz field polarized parallel to the
electrode drives currents inside the metal, thereby inducing charges of opposite sign on opposite
electrode apexes. These charges, and the resulting voltage across the inner device, govern
the current in our CuMnAs crosses. At a given incident THz peak field, the current flowing
through the three crosses is comparable and, correspondingly, the current density scales up with
decreasing width of the crosses. This is consistent with our observation of increasing AMR
signal with decreasing cross size (see main plot of Fig. 4A). Indeed, when we accordingly
6
rescale the data of the main plot of Fig. 4A by the cross sizes, as shown in the inset of Fig. 4A,
the three curves fall on top of each other, in agreement with the expected phenomenology for
the current-induced spin-torque switching.
We point out that charges at the electrode apexes are induced also when a voltage is applied
between opposite electrodes in the contact experiments. Therefore, similar field distributions in
the cross region are expected for the contact (Fig.1B) and non-contact (Fig. 1C) field applica-
tions, differing only by a global scaling factor. We take advantage of this notion to quantitatively
calibrate the THz current. We determine in the contact experiments the critical absorbed Joule
energy density leading to device damage for 1/τp up to 4 GHz. Since the energy already satu-
rates in the GHz range, extrapolation to the THz writing speed is straightforward and enables
determination of the scaling factor between the incident peak field and the resulting peak current
density in the CuMnAs cross.
The calibration procedure allows us to plot the characteristic Joule energy density required
to obtain a reference switching signal of, e.g., 1 mΩ up to the THz range, as shown in the inset of
Fig. 4B. We find a steeply decreasing with increasing 1/τp below ∼MHz to a saturated value
of in the GHz range (see also main plot of Fig. 4B) that extends to the THz writing speed. This
demonstrates that current-induced switching at the THz writing speed is as energy-efficient as
at the GHz speed.
The Joule energy leading to device damage and the energy required for the reference 1 mΩ
switching signal are separated by approximately a factor of 2 in both the GHz and THz writing-
speed range. This breakdown margin is sufficiently large, allowing us to demonstrate tens of
thousands of reversible write-read cycles without any notable device wear-out[15].
In Fig. 4C we finally demonstrate that one antiferromagnetic multi-level bit-cell can be
simultaneously addressed by the non-contact THz-speed writing and the contact lower-speed
writing. For the illustration we choose a CuMnAs/Si bit cell and plot the readout signal when
7
applying 200 ms pulses of j = 1×107 Acm−2 by the contact method with alternating orthogonal
current-path directions and a delay between pulses of 5 s. After 15 switchings we added the
non-contact writing in the form of a 90 s long train of ps-pulses with a kHz repetition rate and a
current density j ≈ 1.6×109 Acm−2. The non-contact ps-pulses induce an additional switching
of the multi-level cell with the superimposed smaller switching signals controlled by the contact
pulses.
Discussion. We will now discuss our results in the broader context of writing of magnetic
memories. Data presented in Figs. 2-4 illustrate that we have pushed the electrical spin-torque
switching, that drives current research and development of magnetic memories, to the THz
writing-speed range. At the GHz writing speed, the antiferromagnetic CuMnAs bit cells are
written by current densities j ∼ 108 Acm−2[15]. These are the same values as reported in fer-
romagnetic spin-orbit torque MRAMs[10, 11]. For the THz writing speeds, the applied current
density in our antiferromagnetic devices increases only to a value of the order of j ∼ 109 Acm−2
and the writing energy remains comparable to that of the GHz writing speed. This makes an-
tiferromagnets realistic candidate materials for electrically controlled memory devices in the
THz band.
The result is in striking contrast to ferromagnets where the projected current and energy
of the writing would increase by three orders of magnitude from the GHz to the THz writing
speed. For 1/τp up to a GHz, the current density required for switching in spin-orbit torque
memories fabricated in the common ferromagnetic transition-metal structures is ∼ 108 Acm−2
and only weakly deviates from the steady-current limit, owing to heat-assisted magnetization
reversal[10]. This current density corresponds to an effective field Hef f ∼ 10 − 100 mT re-
quired for the switching of ferromagnetic moments over the magnetic energy barrier. However,
when 1/τp is above the GHz threshold, the steady-current limit Hef f is no longer sufficient
for switching because τp becomes comparable or smaller than the limiting magnetization reori-
8
entation time-scale. This is given by 1/f where f = γ
2π Hef f is the ferromagnetic resonance
frequency. To keep f in scale with 1/τp, the effective writing field, and therefore j, have to in-
crease linearly with 1/τp above the GHz threshold[10]. In particular, f reaches 1 THz at about
30 T and the switching current density linearly extrapolated to the THz writing speed would
be ∼ 1011 Acm−2 in the ferromagnetic spin-orbit torque devices. This also means that while
the writing energy ∼ j2τp initially drops down with increasing 1/τp, in ferromagnets it starts to
increase linearly with 1/τp above the GHz threshold. A THz writing speed would then require
three orders of magnitude higher energy than the GHz writing speed used in present MRAM
devices. With these basic physical limitations, the current-induced spin-orbit torque switching
has not been pushed in ferromagnetic MRAMs to 1/τp far exceeding 5 GHz[10, 11].
On the other hand, the antiferromagnetic resonance frequency is exchange-enhanced and
scales as ∼ √
exchange field and HA stands for the anisotropy field (in the absence of externally applied
HEHA, which puts it in the THz range[30]. Here HE is the inter-spin-sublattice
fields). This enhancement of the antiferromagnetic resonance frequency and, correspondingly,
of the threshold writing speed is due to canting of the antiferromagnetic spin-sublattices when
brought out of equilibrium.
A uniform magnetic field required for switching is also exchange-enhanced in antiferro-
magnets and scales as ∼ √
HEHA which typically takes it into the tens of Tesla range, inac-
cessible in microelectronics. However, we exploited the current-induced staggered field that is
commensurate with the staggered N´eel order in the antiferromagnet, for which the exchange-
enhancement factor of the switching field amplitude is absent[14]. This principle allows anti-
ferromagnets to reach THz writing speeds at accessible fields and energies, as confirmed by our
experiments and predicted in earlier numerical simulations[14, 19].
We note that in our present experiments, electrical readout is performed at macroscopic
time-delay scales compared to the ps-scale of the writing pulses, taking advantage of the mem-
9
ory effect in our antiferromagnetic CuMnAs bit cells. Among our envisaged future research
directions is to combine the ps-scale writing with a comparable readout scale. This will pro-
vide a time-resolved physical picture of the switching mechanism and will open the prospect
of an ultra-fast complete write-read cycle in antiferromagnetic memories. In an earlier work by
several of us[31] we have already demonstrated an initial step in this direction by detecting the
N´eel vector direction in CuMnAs in a fs-laser setup by magnetic linear dichroism, which is the
optical counterpart of the AMR.
As a concluding remark we point out that a THz-speed memory, whether realized in anti-
ferromagnets or another alternative system that may be discovered in the future, is only one of
the many components that need to be developed to make true THz electronics and information
technologies a realistic prospect. In the meantime, however, the ultra-fast writing of antiferro-
magnetic bit cells can be potentially exploitable without separate THz-speed processors. The
multi-level neuron-like characteristics allows for integrating memory and logic within the anti-
ferromagnetic bit cell, as illustrated in earlier measurements on the example of a pulse-counter
functionality[15]. Future low-noise experiments with THz writing pulses and repetition rates
spanning a broad range up to THz will establish the feasibility and versatility of this autonomous
THz memory-logic concept built within the antiferromagnetic bit-cells, that requires no separate
processor to perform the THz-speed logic operation.
Acknowledgements
We acknowledge support from the Ministry of Education of the Czech Republic Grants LM2015087
and LNSM-LNSpin, the Grant Agency of the Czech Republic Grant No. 14-37427, the Uni-
versity of Nottingham EPSRC Impact Acceleration Account grant No. EP/K503800/1, the
Alexander von Humboldt Foundation, the ERC Synergy Grant SC2 (No. 610115), the Transre-
gional Collaborative Research Center (SFB/TRR) 173 SPIN+X, the ERC Consolidator Grant
TERAMAG (No. 681917), and the Swiss National Science Foundation Grant (No. 21-63864).
10
References and Notes
[1] C. Chappert, A. Fert, F. N. Van Dau, Nature Materials 6, 813 (2007).
[2] A. Brataas, A. D. Kent, H. Ohno, Nature Materials 11, 372 (2012).
[3] A. D. Kent, D. C. Worledge, Nature Nanotechnology 10, 187 (2015).
[4] M. M. Waldrop, Nature 530, 144 (2016).
[5] D. Ralph, M. D. Stiles, Journal of Magnetism and Magnetic Materials 320, 1190 (2008).
[6] A. Chernyshov, et al., Nature Physics 5, 656 (2009).
[7] I. M. Miron, et al., Nature 476, 189 (2011).
[8] L. Liu, et al., Science 336, 555 (2012).
[9] D. Bedau, et al., Applied Physics Letters 97, 262502 (2010).
[10] K. Garello, et al., Applied Physics Letters 105, 212402 (2014).
[11] G. Prenat, et al., IEEE Transactions on Multi-Scale Computing Systems 2, 49 (2016).
[12] C. H. Marrows, Science 351, 558 (2016).
[13] P. Wadley, et al., Science 351, 587 (2016).
[14] J. Zelezn´y, et al., Physical Review Letters 113, 157201 (2014).
[15] K. Olejn´ık, et al., Nature Communications 8, 15434 (2017).
[16] T. Jungwirth, X. Marti, P. Wadley, J. Wunderlich, Nature Nanotechnology 11, 231 (2016).
[17] S. Y. Bodnar, et al. (2017).
11
[18] M. Meinert, D. Graulich, T. Matalla-Wagner (2017).
[19] P. Roy, R. M. Otxoa, J. Wunderlich, Physical Review B 94, 014439 (2016).
[20] P. Wadley, et al., Nature Communications 4, 2322 (2013).
[21] J. Zelezny, et al., Phys. Rev. B 95, 014403 (2017).
[22] D. Kriegner, et al., Nature Communications 7, 11623 (2016).
[23] S. Fukami, C. Zhang, S. DuttaGupta, A. Kurenkov, H. Ohno, Nature Materials 15, 535
(2016).
[24] M. J. Grzybowski, et al., Physical Review Letters 118, 057701 (2017).
[25] M. Sajadi, M. Wolf, T. Kampfrath, Optics Express 23, 28985 (2015).
[26] T. Kampfrath, et al., Nature Photonics 5, 31 (2011).
[27] S. Baierl, et al., Nature Photonics 10, 715 (2016).
[28] A. Novitsky, et al., Journal of Applied Physics 12, 074318 (2012).
[29] J. M. Mcmahon, S. K. Gray, G. C. Schatz, Phys. Rev. B 83, 115428 (2011).
[30] C. Kittel, Physical Review 82, 565 (1951).
[31] V. Saidl, et al., Nature Photonics 11, 91 (2017).
12
Figure 1: (A) Schematics of the crystal and magnetic structure of the CuMnAs antiferromag-
net. An electrical current (black arrow) generates a staggered spin-orbit field (thin red/purple
arrows) that switches the antiferromagnetic moments (thick red/purple arrows). (B) Electron
microscopy image of the cross-shape bit cell and schematics of the reversible writing by elec-
trical pulses of two orthogonal current directions delivered via wire-bonded contacts. White
dashed lines illustrate electrical current paths and white double-arrows the corresponding pre-
ferred N´eel vector orientations. (C) Wave-form of the applied µs electrical pulses. (D) Schemat-
ics of the reversible writing by THz electric-field transients whose linear polarization can be
chosen along two orthogonal directions. (E) Wave-form of the applied ps radiation pulses.
13
(B)CuMnAs(A)(C)(cid:1)(cid:1)(E)(D)2μm Figure 2: (A) Reversible multi-level switching by 30 s trains of µs electrical pulses with a
Hz pulse-repetition rate, delivered via wire-bonded contacts along two orthogonal directions.
The applied writing current density in the 3.5 µm-size CuMnAs/GaAs cell is 3 × 107 Acm−2.
Intervals with the pulse trains turned on are highlighted in grey and the two orthogonal current-
directions of the trains are alternating from one interval to the next. Electrical readout is per-
formed at a 1 Hz rate. Right insets show schematics of the transverse AMR readout. White
dashed lines depict readout current paths.
(B) Same as (A) for ps-pules with a kHz pulse-
repetition rate. The writing current density in the 2 µm-size CuMnAs/GaAs bit cell recalculated
from the amplitude of the applied THz electric-field transient is 2.7 × 109 Acm−2. Electrical
readout is performed at a 8 Hz rate.
14
VVVV(A)(B)(cid:1)(cid:1) Figure 3: (A) The multi-level memory signal as a function of the number of applied ps-pulses.
The writing current density in the 2 µm-size CuMnAs/GaAs bit cell recalculated from the ap-
plied THz field amplitude is 2.9× 109 Acm−2. (B) Same as (A) for the µs-pulses and an applied
writing current density of 3 × 107 Acm−2 in the 3.5 µm-size CuMnAs/GaAs cell.
15
(A)(B)(cid:1) Figure 4: (A) Magnitude of the switching signal as a function of the THz field amplitude (main
panel) and of the converted current density (inset) for 1, 2 and 3 µm size devices (blue, black,
and red dots). Top insets show electron microscopy images of the 1 and 3 µm size devices.
Light regions are Au-contact pads, grey regions are etched down to the GaAs substrate, and
black regions are CuMnAs. (B) Writing energy density (black, red, and green dots), = ρj2τp,
required to obtain a 1 mΩ switching signal as a function of the writing speed 1/τp in the lin-
ear scale (main plot) and in the log-log scale (inset). All data points, except for the point at
1/τp=1 THz in the inset are obtained from the contact measurements. The point at 1/τp=1 THz
in the inset is from the non-contact measurement using the E to j conversion based on the
breakdown energy (see text). Black dots in the main plot correspond to 2 µm, red to 3 µm, and
green to 4 µm size CuMnAs/GaAs bit cells. Black star-symbols and dashed line represent the
limiting breakdown energy density. (C) Contact writing by 200 ms pulses of current density
1 × 107 Acm−2 (white intervals) and the contact writing superimposed on the non-contact writ-
ing by a train of ps-pulses with a kHz repetition rate and a THz field amplitude corresponding
to an additional current density of approximately 1.6 × 109 Acm−2 (grey intervals). Data were
measured in a 10 µm-size CuMnAs/Si bit cell.
16
(B)(cid:1)(cid:1)(cid:1)(cid:1)(cid:1)(A)10μm(C) |
1906.09875 | 1 | 1906 | 2019-06-24T12:13:06 | Characterization of Surface and Structure of in-situ Doped Sol-Gel-Derived Silicon Carbide | [
"physics.app-ph"
] | Silicon carbide (SiC), is an artificial semiconductor used for high-power transistors and blue LEDs, for its extraordinary properties. SiC would be attractive for more applications, but large-scale or large-surface area fabrication, with control over defects and surface is challenging. Sol-gel based techniques are an affordable alternative towards such requirements. This report describes two types of microcrystalline SiC derived after carbothermal reduction from sol-gel-based precursors, one with nitrogen added, the other aluminum. Characterization of their bulk, structure and surface shows that incorporation of dopants affects the formation of polytypes and surface chemistry. Nitrogen leads exclusively to cubic SiC, exhibiting a native oxide surface. Presence of aluminum instead promotes growth of hexagonal polytypes and induces self-passivation of the crystallites surface during growth. This is established by hydrogenation of silicon bonds and formation of a protecting aluminum carbonate species. XPS provides support for the suggested mechanism. This passivation is achieved in only one step, solely by aluminium in the precursor. Hence, it is shown that growth, doping and passivation of SiC can be performed as one-pot synthesis. Material without insulating oxide and a limited number of defects is highly valuable for applications involving surface-sensitive charge-transfer reactions, therefore the potential of this method is significant. | physics.app-ph | physics | Dr. B Friedel, O. Kettner, B. Kunert, Prof. Dr. R. Schennach, Prof. Dr. R. Resel
DOI: 10.1002/adem.201701067
Characterization of Surface and Structure of in-situ Doped Sol-Gel-Derived Silicon
Carbide **
By Olivia Kettner, Sanja Šimić, Birgit Kunert, Robert Schennach, Roland Resel, Thomas
Griesser and Bettina Friedel*
[*] Dr. B. Friedel
illwerke vkw Endowed Professorship for Energy Efficiency, Energy Research Center,
Vorarlberg University of Applied Sciences
6850 Dornbirn, Austria
E-mail: [email protected]
Institute of Solid State Physics, Graz University of Technology
8010 Graz, Austria
S. Šimić
Graz Centre for Electron Microscopy (ZFE)
8010 Graz, Austria
Prof. Dr. T. Griesser
Institute of Chemistry of Polymeric Materials, Montanuniversität Leoben
8700 Leoben, Austria
[**] O.K. and B.F. are grateful to the Austrian Science Fund (FWF) for financial support
under Project No. P 26968.
(Supporting Information is available online from Wiley InterScience or from the author).
Abstract: Silicon carbide (SiC), is an artificial semiconductor used for high-power transistors
and blue LEDs, for its extraordinary properties. SiC would be attractive for more applications,
but large-scale or large-surface area fabrication, with control over defects and surface is
challenging. Sol-gel based techniques are an affordable alternative towards such requirements.
This report describes two types of microcrystalline SiC derived after carbothermal reduction
from sol-gel-based precursors, one with nitrogen added, the other aluminum. Characterization
of their bulk, structure and surface shows that incorporation of dopants affects the formation
of polytypes and surface chemistry. Nitrogen leads exclusively to cubic SiC, exhibiting a native
oxide surface. Presence of aluminum instead promotes growth of hexagonal polytypes and
induces self-passivation of the crystallites' surface during growth. This is established by
1
hydrogenation of silicon bonds and formation of a protecting aluminum carbonate species. XPS
provides support for the suggested mechanism. This passivation is achieved in only one step,
solely by aluminium in the precursor. Hence, it is shown that growth, doping and passivation
of SiC can be performed as "one-pot synthesis". Material without insulating oxide and a limited
number of defects is highly valuable for applications involving surface-sensitive charge-
transfer reactions, therefore the potential of this method is significant.
1. Introduction
Silicon carbide is a high-temperature stable, non-oxidic ceramic and first been synthesized by
A. G. Acheson in 1890,[1] and due to its high Mohs hardness of 9.3, was widely produced and
used as an abrasive.[2] Still today, SiC is an important material, used to improve mechanical and
thermal properties of machine parts,[3] and preventing from wear in abrasive or corrosive
environments.[4-5] But SiC is also a wide band gap semiconductor, highly attractive for high
power transistors, light emitting diodes and solar cells, due to its extraordinary electrical
properties, often referred to as the third generation of semiconductor materials.[6-9] Silicon
carbide appears in more than 200 polytypes, which differ by the stacking sequence of the Si-C
tetrahedrals.[10] However, obtaining semiconductor grade material is comparatively more
challenging: Single crystals from hexagonal polytypes can be grown via sublimation, where
mostly nitrogen-doped or nominally undoped wafers are commercially available.[11] Initially p-
doped material is rare because the deeper valence band of wide bandgap materials makes it hard
to reach sufficient densities,[12] but can be established post-growth via ion-implantation or CVD
epitaxial growth. Cubic polytype 3C-SiC, can merely be grown by heteroepitaxial growth e.g.
on silicon,[13] also nitrogen-doped or nominally undoped. However, lattice mismatch and
different thermal expansion coefficients lead to mechanical tension in the grown material, with
tendency to cracking, therefore 3C-SiC wafers are enormously expensive.[14]
2
SiC preparation via sol-gel processing is a low-cost, solution-processable and environmentally-
friendly addition to those traditional methods,[15] with the potential to expand into further fields
of applications particular by its ability to facilitate the production of large-area and high-surface
area materials. It is suitable for fabrication of nano- to microscale films, of nano- to
microparticles and bulk materials alike.[16] As has been shown, its high degree of freedom
regarding precursor composition even enables in-situ doping, such as highly doped n-type
SiC:N obtained by the introduction of nitric acid or sodium nitride, and p-type SiC:Al with
reasonable charge carrier densities by introduction of aluminum chloride or metallic aluminum
powder during wet-chemical precursor preparation.[17-18] Thereby evidence supporting the
dopants presence and position in the SiC material was found by electron paramagnetic
resonance spectroscopy and associated modelling of the related defect signals for N- and Al-
doping, respectively.[16,19] This motivates a broadening of SiC's application range towards new
fields, such as the electronic acceptor in organic/inorganic hybrid photovoltaics, as robust
photoelectrode in catalysis and for nanostructured blue light emitters, where nanoscale or large
surface area is required.[20-21] In particular from hybrid solar cells, it is a well-known fact that
the surface properties of the inorganic component substantially determine the efficiency of the
device, e.g. via surface trap states fostering trap-mediated recombination or inter-particle charge
transport limitations, caused by their stabilizing ligands.[22-24] In the case of silicon carbide
nanoparticles within a conjugated polymer matrix, Kettner et al. have observed extended
emission lifetimes, assigned to the formation of longer lived polarons upon photoinduced
electron transfer to the inorganic acceptor or surface trap states,[21] which have been suggested
in either case to be beneficial for their function in hybrid photovoltaics.[25] However, also here
the surface is the most crucial point, when incorporating SiC nanoparticles into the organic
semiconductor matrix of a hybrid solar cell.
Investigations on the surface of traditionally derived SiC have been done in the past, e.g. by
Mac Millan et al.[26], Afanesev et al.[27] and Kaplan.[28] Thereby they mainly focused on the
3
formation of native oxide (SiO2) on hexagonal SiC and the interface formation between SiC
and SiO2, also depending on which site was exposed, Si or C. Formation of native oxides on
SiC is a strong disadvantage for most (opto)electronic applications, as the insulating SiO2 layer
hinders charge transfer. Further, it has been reported that SiC can show considerable densities
of surface defects under certain circumstances, for example after etching processes. In one
important example, namely blue emitting porous SiC, those defects can detrimentally diminish
the original emission properties, and subsequent surface passivation is vital.[29]
In this paper, our experimental observations regarding the dopant-related surface termination
of sol-gel derived 3C-SiC crystalline particles are reported. The work aims to give new insights
in the special characteristics of surface formation in doped sol-gel derived cubic SiC and,
therefore, is a further step to exploit the full potential of this material. The work is based on
thorough surface characterization utilizing X-ray photoelectron spectroscopy (XPS) depending
on the introduced dopant. This is accompanied by results on their crystallite size, unit cell and
polytype distribution, as characterized by scanning electron microscopy (SEM) and X-ray
diffraction (XRD) with Rietveld refinement.
2. Materials and Methods
2.1. Preparation of Silicon Carbide Powders
Silicon carbide sub-micrometer powders were fabricated via sol-gel processed precursors and
their carbothermal reduction, which is described in detail elsewhere.[17-18] All chemicals used
for the fabrication of the precursor material were purchased from Sigma-Aldrich and used as
received without any further purification. In brief, sucrose (> 98.5%) as carbon source and
tetraethoxysilane (TEOS) (≥ 99%) as silicon source were dissolved in deionized water and
ethanol (≥ 98%), respectively, to get a carbon-rich silica-sol in ethanolic solution as precursor
material. The silicon-to-carbon ratio was adjusted to 1:4 and hydrochloric acid (37%) was added
as the catalyst. The molar ratio between metal-organic precursor, water and catalyst
4
TEOS:H2O:HCl is at 1:8.2:27. For p- and n-doping, the precursors were modified by adding
nominally 5 at% relative to silicon of aluminum (metallic aluminum powder, 5 µm, 99.5%) or
nitrogen (sodium nitrate, ≥ 99%), respectively. The silica-sol was allowed to gel at 60°C in a
sealed container for about 12 hours, leading to formation of an amber lyogel, and by drying in
an open container at 150°C for 48 hours to get a black xerogel. Next, the precursor was annealed
at 1000°C for 3.5 h under argon gas flow, during which left organic degradation products of the
former sucrose get carbonized and vitrification of the silicate gel is completed. To convert that
gained carbon-rich silicate glass into silicon carbide, the precursor was sintered at 1800°C in
argon atmosphere for 15 minutes in an induction furnace. Some nitrogen-doped samples were
subsequently annealed in oxygen atmosphere at 1000°C for 10 minutes to demonstrate
oxidation effects.
2.2. Characterization
Imaging on the SiC crystals was performed via scanning electron microscopy (SEM) (Zeiss
DSM 982 Gemini). The microscope was equipped with a thermal field emission gun as source
and an Everhart Thornley Detector. Secondary electron images were recorded with an
acceleration voltage of 10 keV. The structure was investigated by X-ray diffraction (XRD)
using a Siemens D 501 diffractometer in Bragg-Brentano geometry operated at 40 kV and 30
mA using Cu Kα radiation and a graphite monochromator at the secondary side. The XRD data
were refined by Rietveld analysis for phase fractions and lattice parameters (Bruker AXS,
Topas Version 3.0). Qualitative phase analysis was performed on basis of PDF-2. XPS was
performed with a monochromatic Thermo Fisher K-Alpha spectrometer equipped with an Al
X-ray source (1486.6 eV) operating at a base pressure in the range of 10-8 to 10-10 mbar. High
resolution scans were acquired at a pass energy of 50 eV and a step size (resolution) of 0.1 eV.
Survey scans were acquired with a pass energy of 200 eV and a step size of 1.0 eV. The
instrument work function was calibrated to give a binding energy (BE) of 83.96 eV for the Au
5
4f7/2 line for metallic gold. All measurements were performed at room temperature. The peaks
were fitted utilizing Gaussian/Lorentzian mixed functions employing a linear background
correction (program XPSPEAK41). Absorption infrared spectroscopy (RAIRS) was used to
obtain electronic and chemical material information. A detailed description of the set-up and
measurement procedure used for performing reflection measurements can be found
elsewhere.[30] In short, the spectra were recorded utilizing a reflection unit with variable angle
and a motorized polarizer (Bruker Optics). The measurements were performed with an
incidence angle of 74° under vacuum with a base pressure of about 4 mbar. A correction for the
substrate absorption (indium foil) was done for the recorded spectra.
3. Results
3.1. Structural Analysis
Appearance and morphology of the nitrogen- (SiC:N) and aluminum-doped SiC (SiC:Al)
powder samples were investigated by SEM. The easiest obtainable form of sol-gel derived SiC
(beside fibers, thin films and porous structures) are microcrystals. Figure 1 shows two
examples of microcrystalline samples of SiC:N (a) and SiC:Al (b). Optical microscopy images
of the samples (not shown) show SiC:N microcrystals in their typical green colour, while
SiC:Al microcrystals appear dark blue. One of the most obvious features of the crystals of
SiC:N are the triangular shaped facets, originating from the truncated tetrahedron, typical and
characteristic for the zinc blende structure of the cubic SiC polytype 3C-SiC. In comparison,
the SiC:Al sample is dominated by hexagonal facets of rather flat crystals, which cannot be
clearly assigned to a certain polytype merely from the image. They might still be truncated
tetrahedra of different tracht and habit, therefore showing full hexagonal instead of triangular
faces and flattened shape. It is another possibility that these are hexagonal flat crystals, which
are typical for the so-called Lely platelets of the hexagonal polytype 6H-SiC.
6
A clear identification and distinction of polytypes in the two doped microcrystalline samples
can be obtained by XRD powder diffraction. The XRD patterns of equally derived SiC:N and
SiC:Al microcrystalline powders are shown in Figure 2 (a) and (b), respectively. In the
displayed angular range both, the SiC:N (a) and SiC:Al (b) patterns show common most
prominent peaks at 35.7°, 41.5°, 60.1°, 71.8° and 75.6°. The according lattice constants were
calculated to d111 = 2.51 Å, d200 = 2.18 Å, d220 = 1.54 Å, d311 = 1.31 Å and d222 = 1.26 Å,
respectively. These are in good agreement with literature values for 3C-SiC and can be assigned
to its (111), (200), (220), (311) and (222) crystal planes.[31] Further, the 3C-SiC (111) diffraction
peak shows two satellites at 34.2° and 38.2° for both samples. With their calculated lattice
constants of d101 = 2.62 Å and d103 = 2.36 Å, they have been identified as the (101) and (103)
crystal planes of the 6H-SiC polytype. Thereby the intensity of the satellites in comparison to
the 3C-SiC (111) is less intense for SiC:N than for SiC:Al, which could indicate a larger
contribution of the 6H-SiC polytype in the SiC:Al sample. The XRD pattern of SiC:N shows
further minor peaks, of which all but one could be assigned to 6H-SiC. This confirms that the
cubic polytype 3C-SiC and a minor portion of the hexagonal 6H-SiC polytype, dominate the
phase composition of the nitrogen-doped SiC microcrystals (Figure 2a). No contributions of
SiO2 or foreign species such as silicon nitride were found. In comparison, the pattern for the
aluminum-doped SiC microcrystals (Figure 2b) is slightly more complex. Here, the 3C-SiC
phase is still the most prominent one, according to their peak intensity, but overlaid with a large
number of further diffraction peaks. Part of them are the aforementioned 6H-SiC satellites at
34.2° and 38.2°, along with the additional minor signals of 6H-SiC, which in general show
higher intensity than for the ones in the SiC:N powder. Two small shoulders to the 6H-SiC
peaks appearing in both sample patterns at 37.7° and 64.7°C are caused by negligible
contributions of the polytype 15R-SiC. The additional diffraction peaks, which are not present
for SiC:N, have been assigned to the hexagonal polytype 4H-SiC. It should be noted that the
peak at 26.6° is a SiO2 contribution (quartz (011)), does not originate from of the actual sample
7
(as confirmed by FTIR and XPS) but is an external contamination with agate. No foreign
impurities such as aluminum oxide have been found. These structural results indicate that the
SiC:Al samples exhibit a different polytype composition than SiC:N despite identical
preparation conditions besides the dopant.
For a distinct quantitative analysis of the phase-composition of the samples, Rietveld
refinement on the respective XRD patterns has been performed to obtain the dopant-dependent
unit cell dimensions and the crystallite sizes of certain polytype phases.[32-33] The obtained
values are summarized in Table 1 and Table 2 for SiC:N and SiC:Al, respectively. The detailed
comparison of the observed pattern with the data gained from Rietveld analysis and their
differential-plots can be found in the supplementary information. According to Rietveld
refinement, the SiC:N powder consists of 93% of the cubic 3C-SiC polytype and 7% of the
hexagonal 6H-SiC. The peaks belonging to 3C-SiC are sharp and can be fitted to crystal sizes
larger than 200 nm. The peaks assigned to the 6H-SiC are broadened, leading to calculated
crystal sizes of around 30 nm. There is no indication of other polytypes in the SiC:N sample
according to Rietveld refinement. In comparison, the evaluation of data collected from the
SiC:Al microcrystalline powder revealed a phase-composition of about 53% 3C-SiC, 37% 6H-
SiC and 10% 4H-SiC polytype. In this case, 3C-SiC and 6H-SiC both, gave particle sizes larger
than 100 nm, while the line broadening of peaks related to 4H-SiC indicate smaller particle
sizes of around 30 nm. This confirms first indications from SEM imaging that SiC:N indeed is
dominated by the cubic 3C-SiC polytype of relatively large sizes, while the smaller particles
can be assigned to the hexagonal 6H-SiC minority. The SiC:Al on the other hand has large
contributions of both, the cubic 3C-SiC and the hexagonal 6H-SiC polytype with intermediate
crystal sizes, while here a minority of small particles of another hexagonal polytype 4H-SiC
have formed. A reason for the difference in polytype formation despite identical synthesis
parameters must be related to the presence of the dopant. Thereby the effect of the dopant on
polytype formation can have a kinetic as well as a chemical origin, as it was shown by Jepps et
8
al. that nitrogen as impurity promotes the growth of 3C-SiC.[10] Aluminum, in contrast, tends
to stabilize the hexagonal modifications.[34]
The unit cell data (axial length a for cubic, basal axial length a and height c for hexagonal unit
cells) of each polytype fraction of the SiC:N and SiC:Al samples were derived by Rietveld
refinement and are also presented in Tables 1 and 2, respectively. The value for the lattice
constant of nominally undoped 3C-SiC at room temperature is specified with a3C-SiC=4.3596
Å,[35] which is in good agreement with the obtained unit cell data for the cubic component of
SiC:N with a3C-SiC:N = (4.359±0.005) Å and SiC:Al with a3C-SiC:Al = (4.36±0.01) Å. Regarding
hexagonal contributions the polytypes 6H-SiC and 4H-SiC are significant for Rietveld
refinement of the samples SiC:N (6H) and SiC:Al (6H, 4H). Thereby literature refers to lattice
parameters of basal axis of a6H-SiC = 3.0806 Å and height of c6H-SiC = 15.1173 Å for nominally
undoped 6H-SiC.[35] The calculated values for SiC:N are a6H-SiC:N = (3.081±0.005) Å and c6H-
SiC:N = (15.118±0.005) Å, and for SiC:Al a6H-SiC:Al = (3.09±0.01) Å and height of c6H-SiC:Al =
(15.13±0.01) Å, respectively. All parameters lie entirely within this range, merely the height of
the 6H-SiC:Al unit cell appears to be slightly elongated, even with the considered deviation of
the refinement. Reported unit cell dimensions for 4H-SiC are a4H-SiC = 3.0730 Å for its axis
and c4H-SiC = 10.0530 Å for its height.[2] Rietveld refinement on the SiC:Al here led to a4H-SiC:Al
= (3.08±0.01) Å and height of c4H-SiC:Al = (10.10±0.01) Å, again in good agreement for the axis
component, but a significant elongation of the height. Comparing these lattice parameters of
sol-gel derived nitrogen- and aluminum-doped SiC with literature, it can be observed that the
hexagonal unit cells for SiC:Al are slightly larger, while they are unchanged for the cubic
component or nitrogen incorporation. The elongation of lattice parameters for hexagonal
SiC:Al, which makes Δc6H-SiC:Al = (0.01±0.01) Å and Δc4H-SiC:Al = (0.05±0.01) Å can be
explained by the larger covalent radius of the Al atom (1.18 Å), which resides on a silicon-
position (Si covalent radius 1.11 Å) in the lattice, which has been proven by Li et al. for 4H-
SiC.[36] Nitrogen, which resides at carbon sites,[37] has a slightly smaller covalent radius of 0.75
9
Å, compared to carbon with 0.77 Å, but does not lead to significantly changed unit cell
dimensions in this case.[38]
3.2. Surface Composition
Surface chemistry of the differently doped sol-gel derived SiC powders was investigated via
XPS to identify any connections between the surface termination and the incorporated dopant,
in particular the effect on native oxide formation. Therefore the silicon Si 2p and carbon C 1s
core level spectra were utilized to distinguish different silicon and carbon species in the samples
and are shown in Figure 3. A comparison of the Si 2p core level spectra of SiC:N (top) and
SiC:Al microcrystalline powder (bottom) is shown in Figure 3a. The Si 2p spectra can be
deconvoluted into peaks arising from the different oxidation states of silicon Si0, Si1+, Si2+, Si3+
and Si4+,[39] which are located at different binding energies, thus allow identification of chemical
bonds and make deductions of related compounds. In the present case, the Si 2p spectrum of
SiC:N shows only two distinct peaks, one at a binding energy of 103.7 eV and another at 100.8
eV, whereas the latter has considerably higher intensity. While the low-energy signal can be
assigned to the Si+ state in SiC, [40] the other at higher energy arises from Si4+ in SiO2, [41-43]
indicating the presence of a significant native surface oxide layer on the SiC:N microcrystals.
In comparison, the Si 2p core level spectrum of the SiC:Al microcystals shows merely one
dominant peak at a binding energy of 99.7 eV with a tail towards the high-energy side. The
strong peak can be assigned to Si0, which originates either from Si-Si bonds in bulk silicon, [44-
45] or from Si-H bonds, as typically observed for hydrogenated SiC at this energy.[46-48] The tail
originates from a quite weak underlying peak centered at 101.3 eV, which can be assigned to a
shifted Si+ signal from SiC or a Si2+ contribution originating from an imperfect sub-oxide layer
(SiOx with 0<x<2) or a combination of both.[39,49] There are no signs of a native oxide layer on
the surface of the SiC:Al crystallites.
10
The C 1s core level spectra of SiC:N (top) and SiC:Al (bottom) microcystalline powders are
shown in Figure 3b. The SiC:N spectrum exhibits an asymmetric line feature built by two
underlying peaks, a strong one located at 282.9 eV and a minor one at 284.8 eV. The low-
energy peak at 282.9 eV can be clearly identified as originating from the C-Si bond of silicon
carbide.[43,50] The less intense peak at 284.8 eV can roughly be assigned to some kind of C-C
bond, but the exact origin is unclear.[50] In literature, C 1s peaks occurring at 285.0 eV ±0.4 eV
have been discussed related to various origins,[51] such as diamond-like sp3 carbon (285.0
eV),[52] graphitic sp2 carbon (284.6 eV), [53] adventitious carbon (285.1 eV) from organic post-
fabrication deposites,[54] or mixed sp2/sp3 phases.[55] Estrade-Szwarckopf investigated
asymmetric C 1s features, where a broad peak occurred, shifted against the graphitic sp2 peak
with its intensity varying with the treatment of the sample and referred to this broad 285 eV
peak as "defect carbon".[56] Also Emtsev et al. and Rani et al. found such a broad C 1s peak
close to 285 eV when investigating graphene grown epitaxially onto SiC or Gold, respectively,
and also ascribed it to sp2 defects at the graphene-substrate interface.[57-58] Finally, Iwanowski
et al. suggested carbonaceous surface exclusions from a carbon-saturated cubic silicon carbide
(Si1-xCx with x>0.54) crystal lattice, being responsible for the observed 285 eV peak. [59] In the
present C 1s spectra for SiC:N, the small signal intensity of that peak makes surface sp2, sp3 or
mixed carbons rather unlikely as origin, and also adventitious carbons can be excluded in
absence of exposure to organics between high-temperature synthesis and characterization.
Therefore, here it is suggested that the observed 284.8 eV feature is related to buried defect C-
C bonds at the interface between the SiC and the native silicon oxide layer.[60]
The C 1s spectrum of the SiC:Al microcrystalline powder indicates at first glance a completely
different surface composition in comparison to SiC:N. The spectrum can be deconvoluted into
two almost equally strong peaks at 281.9 eV and 284.5 eV and a very weak one at 289.1 eV.
The most intense feature, arising at 281.9 eV, was attributed to C-Si bonding in silicon
carbide.[44] Its significant shift in binding energy compared to SiC:N can be ascribed to the
11
notably different sample composition, because binding energies are known to shift with the
polytype.[59] The slightly less intense peak with its maximum at 284.5 eV was clearly identified
as C-C bonding of graphitic surface carbon.[61] The weak signal at 289.1 eV was ascribed to O-
C=O bonding, suggesting the formation of minor carbonates on the surface.[62] These results
indicate that the surface structure of the SiC:Al crystallites is composed of a hydrogenated
silicon interface buried under a graphene surface layer, similar to what has been reported by
Pallechi et al..[63] Thereby the thickness of the graphene layer can be estimated from the
intensity ratio between the carbide and graphitic peak to be nominally between one and two
monolayers.[64] The very small carbonate signal suggests that the Al2(CO3)3 is merely present
as a small density of distributed "defect-like" sites below or above the carbon surface.
3.3. Bulk Properties
The RAIRS spectra of SiC:N and SiC:Al microcrystals are illustrated in Figure 4. Both samples
clearly show a strong peak between 750 and 1000 cm-1, which is referred to as the "reststrahlen
band", a typical feature arising from the phonon-polariton resonance of bulk SiC. This peak
generally dominates spectra recorded from SiC and is caused by vibrations of the Si- and C-
sublattices against each other.[65] The tail of the phonon band towards higher wavenumbers for
SiC:N is an effect that originates from high dopant concentration.[66] The unusual sharpness of
the peak compared to the rather broad reststrahlen band usually seen for SiC single crystal
wafers is a result of the present microcrystalline morphology of the material. Narrowing of the
reststrahlen band has been reported for porous structures and powders and assigned to an
increase in surface area.[26] On the high wavenumber side, where SiC:Al shows no further
features, SiC:N exhibits an additional absorption band at about 1110 cm-1. This peak can be
assigned to Si-O-Si vibrations of SiO2.[67] This peak is superposed with a very broad absorption
in the region between 2350 and 970 cm-1. The latter is a result of plasmon-phonon interactions,
which occur when plasmon and phonon modes appear in a similar frequency range.[65] The
12
plasma frequency of undoped SiC lies typically in the microwave range, but strongly depends
on charge carrier concentration, effective electron mass and the electron mobility. Here, the
shift to higher energies emerges from large charge carrier concentration, which was found to
be ND-NA = 2.1⋅1019cm-3 from Hall-measurements on this synthesized material.[17] The small
absorption peak at about 640 cm-1 for SiC:N can be attributed to the coupling of the longitudinal
optical phonon to the plasmon.[65] The absence of both these features in SiC:Al is caused by its
considerably lower charge carrier concentration from p-doping, which was found to be merely
ND-NA = -6.8⋅1011cm-3.[68] The absence of the 1110 cm-1 oxide peak in the spectrum of the
SiC:Al is in good agreement with the XPS results, which is showing no silicon oxide present
on the surface. The absence of a carbonate contribution in spectrum is not surprising because it
appeared only in trace amounts on the surface, and undetectable for rather bulk sensitive IR
measurements.
4. Discussion
While the incorporation of a nitrogen dopant with this SiC synthesis method results in SiC:N
microcrystals exhibiting typical common surface composition of SiC with a native oxide (SiO2)
layer and some imperfect graphite species, the equivalent incorporation of aluminum in the
process leads to a very different surface composition. The SiC:Al exhibits apparently a quite
stable hydrogen-terminated surface (Si-H), most likely buried under a graphene mono- or
bilayer surface, while oxygen is merely found bound in a very small density of aluminum
carbonate sites near the surface. These seem to cause a self-passivation effect on the SiC:Al
crystallites and hinder formation of a native oxide layer. Hydrogen is known to saturate
dangling bonds in semiconductors, especially in silicon, but also in SiC and therewith
deactivating potential luminescence quenching or charge trapping sites.[29] The presence of a
high density of Si-H bonds and absence of a Si-dangling bond signal (BE < 99eV) for SiC:Al
here, indicates successful deactivation of these defects. Further, the carbonate layer, despite
13
having less than nominally 1nm thickness judged by the intensity of the signal, seems to have
passivated the SiC surface against the typical formation of silicon oxide at exposure to air.
Potential reasons for this different surface composition of the present SiC:Al lay in the fact that
during this "one-pot synthesis" and in particular during the final annealing step, all chemical
components are enclosed in the reaction chamber volume. The carbothermal reduction of
silicon species in the precursor and formation of SiC occurs merely at temperatures > 1700°C,
following the reaction SiO2 (s) + 3C (s) SiC (s) + 2CO (g). This happens in two ways: (1)
directly reacting with solid carbon via SiO2 (s) + C (s) SiO (g) + CO (g) and SiO (g) + 2C
(S) SiC (s) + CO (g), and indirectly by reaction with CO via SiO2 (s) + CO (g) SiO (g) +
CO2 (g) and SiO (g) + 3CO (g) SiC (s) + 2CO2 (g). As also the aluminum compounds go
into the gas phase during this process, where some aluminum is built into the SiC lattice as
dopant, also secondary chemical reactions between Al and CO/CO2 are likely. It is suggested
that here aluminum carbonate is formed at the surface of the SiC crystals during the cooling
below 1700°C, when silicon species are no longer in a stable gas phase nor reacting with C or
CO. Three scenarios are imaginable for their formation: (1) Few separated near-surface
aluminum sites in the SiC lattice react with the CO/CO2 gas; (2) remaining gaseous aluminum
species condensate on the SiC surface and subsequently react with surface carbon or the
CO/CO2 gas; (3) gaseous aluminum species react in the gas phase with CO/CO2 and
subsequently condensate on the SiC surface. However, since aluminum carbonate Al2(CO3)3 is
known not to be environmentally stable, thus would have been decomposed to Al(OH)3 by the
time of measurement. Therefore it is assumed that the carbonate sites are also buried underneath
the graphene surface layer, which stabilizes them. No other aluminum species like Al2O3 or
Al(OH)3 were found. This was also confirmed by FTIR, which also proves the absence of SiO2
in the SiC:Al sample and further demonstrates the typical difference in charge carrier
concentration between the n- and p-doped SiC by the appearance of the plasmon-phonon
interaction. The observed formation of different polytypes, triggered by the incorporated dopant
14
and also the different sizes of the according crystallites, allows certain speculations about the
role of the dopant in the crystal growth process. The dopant might be acting "only" as a catalyst
or by its displacement of carbon or silicon in the lattice, leading to changes of the unit cell size
and thus giving preference to certain polytype formation. The role of the dopant as chemical
pathway with strong impact on the final polytype has also been discussed by Ariyawong et al.
who addressed the link between the crystal chemistry and growth process parameters.[69]
Therein SiC is treated as a solid solution and the polytypes evaluated in terms of their respective
C and Si activities. They found that 3C-SiC is always obtained with high Si activity in SiC, i.e
close to the SiC-Si two-phase boundary. In the present case, the dominant polytype for SiC:N
is 3C-SiC and the crystals are oxidized on the surface. With Al incorporation in SiC:Al, the
hexagonality increases and the SiC surface is not oxidized. Therefore it could be considered
that the reducing conditions on SiC by Al might increase the C activity in the crystals, thus
promoting formation of the hexagonal polytypes.
Most important in terms of usability of this material in electrochemical and (opto)electronic
applications is clearly the role of aluminum for self-passivation of the as-formed crystals,
protecting from oxidation and most likely deactivation of surface defects by saturation with
hydrogen. The deactivation of dangling bonds and prevention of insulating oxide layers is vital
for efficient charge transfer and transport in this material e.g. in LEDs or catalysis. This sort of
surface passivation is usually only achieved by post-treatment such as alumina or titania
deposition on freshly HF-etched SiC surfaces.[29] In contrast, our synthesis route via sol-gel
precursors and carbothermal reduction allows to obtain SiC growth, doping and surface
passivation as a one-step process.
5. Conclusions
In summary, nitrogen- and aluminum-doped SiC microcrystalline powders have been
synthesized from sol-gel based precursors after carbothermal reduction and have been
15
investigated regarding the effect on surface properties and structural changes triggered by the
incorporated dopant. Therefore SEM, XRD, XPS and RAIRS were used to gain knowledge
about the polytype composition, crystallite size, unit cell size, bulk chemistry and surface
chemistry.
It was revealed that SiC:N consists mostly of the cubic polytype 3C-SiC, while SiC:Al is
devided into 50% 3C-SiC and 50% hexagonal polytypes 6H-SiC and 4H-SiC. This might be
caused either by catalytic properties of the aluminum, promoting formation of the more energy
expensive formation of hexagonal phase, or triggered by the change in unit cell size, which
decreases where the smaller nitrogen atom replaces carbon and increases where the larger
aluminum atom replaces silicon. Independent from this bulk effect, also the surface composition
of the two materials is very different. While the nitrogen dopant leads to formation of SiC:N
microcrystals with a typical SiO2 surface layer, the equivalent incorporation of aluminum
generates SiC:Al crystallites with a surface of hydrogenated silicon bonds and a small density
of aluminum carbonate sites buried under a graphene layer. This leads to two valuable effects
for SiC: deactivated defect sites by hydrogen saturation of dangling bonds and passivation of
the crystallites surface to prevent formation of a native oxide.
The SiC:N instead exhibits no signs of hydrogenation and no graphitic carbon, but the
commonly seen native oxide layer, where dangling bonds are saturated by unfavorable oxygen.
This difference between SiC:N and SiC:Al might be caused by the high reactivity of aluminum
and the longer stability of gaseous aluminum species after SiC formation at the end of the high-
temperature process, leading to the natural formation of a passivation layer on the SiC
crystallites directly after growth. The nature of this passivation layer appeared as a carbonate
feature in XPS, which for aluminum is known to be unstable in air. Therefore it is suggested
that the carbonate is formed at the interface between the SiC and the graphene surface, which
stabilizes it. The fact that SiC growth, doping and surface passivation can be achieved in one
step with our „one-pot synthesis" is a vital key for the application of such affordably prepared
16
materials in applications involving surface-sensitive charge-transfer reactions. This is
promising for its potential use in LEDs, photovoltaics and catalysis.
Received: ((will be filled in by the editorial staff))
Revised: ((will be filled in by the editorial staff))
Published online: ((will be filled in by the editorial staff))
_[1] E. G. Acheson, Production of artificial crystalline carbonaceous materials, carborundum.
British Pat. 17911, 1892.
_[2] M. E. Levinshteĭn, S. L. Rumyantsev, M. Shur, Properties of advanced semiconductor
materials: GaN, AlN, InN, BN, SiC, SiGe, Wiley, New York 2001.
_[3] S. Prochazka, Hot pressed silicon carbide. US Pat. US3853566, 1974.
_[4] F. Mubarok, N. Espallargas, J Therm Spray Tech 2015, 24, 817.
_[5] K. Miyoshi, D. H. Buckley, Wear 1982, 75, 253.
_[6] J. Rabkowski, D. Peftitsis, H. Nee, EEE Ind. Electron. Mag. 2012, 6, 17.
_[7] A. Elasser, M. H. Kheraluwala, M. Ghezzo, R. L. Steigerwald, N. A. Evers, J. Kretchmer,
T. P. Chow, IEEE Trans. on Ind. Applicat. 2003, 39, 915.
_[8] R. Dietmueller, H. Nesswetter, S. J. Schoell, I. D. Sharp, M. Stutzmann, ACS applied
materials & interfaces 2011, 3, 4286.
_[9] S. E. Saddow, F. La Via (Eds.), Advanced Silicon Carbide Devices and Processing,
InTech 2015.
_[10] N. W. Jepps, T. F. Page, Progress in Crystal Growth and Characterization 1983, 7, 259.
_[11] S. Müller, R. C. Glass, H. M. Hobgood, V. F. Tsvetkov, M. Brady, D. Henshall, J. Jenny,
D. Malta, C. H. Carter, Journal of Crystal Growth 2000, 211, 325.
_[12] P. Deák, B. Aradi, A. Gali, U. Gerstmann, phys. stat. sol. (b) 2003, 235, 139.
_[13] Y. Hattori, T. Suzuki, T. Murata, T. Yabumi, K. Yasuda, M. Saji, Journal of Crystal
Growth 1991, 115, 607.
_[14] D. L. Smith, Includes index, Thin film deposition and applications, McGraw-Hill, New
York 1995.
_[15] M. Helgesen, R. Søndergaard, F. C. Krebs, J. Mater. Chem. 2010, 20, 36.
_[16] S. Greulich-Weber, B. Friedel, MSF 2009, 615-617, 637.
17
_[17] B. Friedel, S. Greulich-Weber, MRS Proc. 2006, 963.
_[18] B. Friedel, S. Greulich-Weber, MSF 2006, 527-529, 759.
_[19] A. Konopka, B. Aşik, U. Gerstmann, E. Rauls, N. J. Vollmers, M. Rohrmüller, W. G.
Schmidt, B. Friedel, S. Greulich-Weber, IOP Conf. Ser.: Mater. Sci. Eng. 2010, 15, 12013.
_[20] Van de Walle, Chris G, J. Neugebauer, Nature 2003, 423, 626.
_[21] O. Kettner, C. E. Finlayson, B. Friedel, Nanosci Nanotechnol Lett 2015, 7, 56.
_[22] D. V. Talapin, J.-S. Lee, M. V. Kovalenko, E. V. Shevchenko, Chemical reviews 2010,
110, 389.
_[23] F. Gao, Z. Li, J. Wang, A. Rao, I. A. Howard, A. Abrusci, S. Massip, C. R. McNeill, N.
C. Greenham, ACS nano 2014, 8, 3213.
_[24] K. M. Noone, S. Subramaniyan, Q. Zhang, G. Cao, S. A. Jenekhe, D. S. Ginger, J. Phys.
Chem. C 2011, 115, 24403.
_[25] J. Piris, N. Kopidakis, D. C. Olson, S. E. Shaheen, D. S. Ginley, G. Rumbles, Adv. Funct.
Mater. 2007, 17, 3849.
_[26] M. F. MacMillan, R. P. Devaty, W. J. Choyke, D. R. Goldstein, J. E. Spanier, A. D.
Kurtz, Journal of Applied Physics 1996, 80, 2412.
_[27] V. V. Afanasev, M. Bassler, G. Pensl, M. Schulz, phys. stat. sol. (a) 1997, 162, 321.
_[28] R. Kaplan, Surface Science 1989, 215, 111.
_[29] W. Lu, Y. Ou, E. M. Fiordaliso, Y. Iwasa, V. Jokubavicius, M. Syväjärvi, S. Kamiyama,
P. M. Petersen, H. Ou, Scientific reports 2017, 7, 9798.
_[30] M. Djak, E. Gilli, E. Kontturi, R. Schennach, Macromolecules 2011, 44, 1775.
_[31] J. Ruska, L. J. Gauckler, J. Lorenz, H. U. Rexer, J Mater Sci 1979, 14, 2013.
_[32] H. M. Rietveld, J Appl Crystallogr 1969, 2, 65.
_[33] A. L. Ortiz, F. L. Cumbrera, F. Sánchez-Bajo, F. Guiberteau, H. Xu, N. P. Padture,
Journal of the American Ceramic Society 2000, 83, 2282.
_[34] P. A. Kistler-De Coppi, W. Richarz, International Journal of High Technology Ceramics
1986, 2, 99.
_[35] A. Taylor, R. M. Jones, The crystal structure and thermal expansion of cubic and
hexagonal silicon carbide, Proc. Conf. Silicon Carbide 1960 147-154, Pergamon Press, New
York.
_[36] Y. Li, J. Li, C. Chen, Z. Yang, Y. Yang, Z. Lin, Applied Surface Science 2012, 259, 21.
_[37] S. Greulich-Weber, phys. stat. sol. (a) 1997, 162, 95.
18
_[38] B. Zhang, J. Li, J. Sun, S. Zhang, H. Zhai, Z. Du, Journal of the European Ceramic
Society 2002, 22, 93.
_[39] X. Y. Chen, Y. F. Lu, L. J. Tang, Y. H. Wu, B. J. Cho, X. J. Xu, J. R. Dong, W. D. Song,
Journal of Applied Physics 2005, 97, 14913.
_[40] T. Parrill, Y. Chung, Surface Science 1991, 243, 96.
_[41] A. Stoch, J. Stoch, A. Rakowska, Surf. Interface Anal. 1994, 22, 242.
_[42] B. Herreros, T. L. Barr, P. J. Barrie, J. Klinowski, J. Phys. Chem. 1994, 98, 4570.
_[43] A. A. Galuska, J. C. Uht, N. Marquez, Journal of Vacuum Science & Technology A:
Vacuum, Surfaces, and Films 1988, 6, 110.
_[44] I. Kusunoki, Y. Igari, Applied Surface Science 1992, 59, 95.
_[45] S. A. Chambers, V. A. Loebs, Phys. Rev. B 1993, 47, 9513.
_[46] M. M. Kamble, V. S. Waman, A. H. Mayabadi, S. S. Ghosh, B. B. Gabhale, S. R.
Rondiya, A. V. Rokade, S. S. Khadtare, V. G. Sathe, T. Shripathi, H. M. Pathan, S. W. Gosavi,
S. R. Jadkar, Journal of Coatings 2014, 2014, 1.
_[47] T. Seyller, J. Phys.: Condens. Matter 2004, 16, S1755-S1782.
_[48] S. Gallis, M. Huang, H. Efstathiadis, E. Eisenbraun, A. E. Kaloyeros, E. E. Nyein, U.
Hommerich, Appl. Phys. Lett. 2005, 87, 91901.
_[49] W. Zhang, S. Zhang, Y. Liu, T. Chen, Journal of Crystal Growth 2009, 311, 1296.
_[50] S. Contarini, S. P. Howlett, C. Rizzo, B. A. de Angelis, Applied Surface Science 1991,
51, 177.
_[51] S.-H. Nam, M.-H. Kim, J.-S. Hyun, Y. D. Kim, J.-H. Boo, J. Nanosci. Nanotech. 2010,
10, 2741.
_[52] F. Arezzo, N. Zacchetti, W. Zhu, Journal of Applied Physics 1994, 75, 5375.
_[53] M. M. Waite, S. I. Shah, Appl. Phys. Lett. 1992, 60, 2344.
_[54] S. J. Harris, A. M. Weiner, Appl. Phys. Lett. 1989, 55, 2179.
_[55] J. C. Arnault, S. Saada, S. Delclos, L. Intiso, N. Tranchant, R. Polini, P. Bergonzo,
Diamond and Related Materials 2007, 16, 690.
_[56] H. Estrade-Szwarckopf, Carbon 2004, 42, 1713.
_[57] K. V. Emtsev, F. Speck, T. Seyller, L. Ley, J. D. Riley, Phys. Rev. B 2008, 77.
_[58] J. R. Rani, J. Lim, J. Oh, D. Kim, D. Lee, J.-W. Kim, H. S. Shin, J. H. Kim, S. C. Jun,
RSC Adv. 2013, 3, 5926.
19
_[59] R. Iwanowski, K. Fronc, W. Paszkowicz, M. Heinonen, Journal of Alloys and
Compounds 1999, 286, 143.
_[60] C. Önneby, C. G. Pantano, Journal of Vacuum Science & Technology A: Vacuum,
Surfaces, and Films 1997, 15, 1597.
_[61] H. Estrade-Szwarckopf, B. Rousseau, Journal of Physics and Chemistry of Solids 1992,
53, 419.
_[62] W. Si, X. Wu, J. Zhou, F. Guo, S. Zhuo, H. Cui, W. Xing, Nanoscale research letters
2013, 8, 247.
_[63] E. Pallecchi, F. Lafont, V. Cavaliere, F. Schopfer, D. Mailly, W. Poirier, A. Ouerghi,
Scientific reports 2014, 4, 4558.
_[64] M. H. Oliveira, T. Schumann, F. Fromm, R. Koch, M. Ostler, M. Ramsteiner, T.
Seyller, J. M. J. Lopes, H. Riechert, Carbon 2013, 52, 83.
_[65] H. Mutschke, A. C. Andersen, D. Clément, T. Henning, G. Peiter, Astron. Astrophys.
1999, 345, 187.
_[66] R. T. Holm, P. H. Klein, P. E. R. Nordquist, Journal of Applied Physics 1986, 60, 1479.
_[67] Z. An, R. K. Y. Fu, P. Chen, W. Liu, P. K. Chu, C. Lin, J. Vac. Sci. Technol. B 2003,
21, 1375.
_[68] B. Friedel, 3C-Siliziumkarbid auf Sol-Gel-Basis: Entwicklung, Wachstumsmechanismen
und Charakter anwendungsorientierter Morphologien des Wide-Bandgap-Halbleiters,
Doctoral Thesis, University of Paderborn (2007), DOI:10.13140/RG.2.1.2957.1445.
_[69] K. Ariyawong, C. Chatillon, E. Blanquet, J.-M. Dedulle, D. Chaussende,
CrystEngComm 2016, 18, 2119.
20
Fig. 1. SEM images of doped silicon carbide microcrystals as derived from the sol-gel based
precursor after carbothermal reduction for a nitrogen- (a) and aluminum-doped SiC sample
(b).
Fig. 2. X-ray diffraction patterns of pristine nitrogen-doped SiC powder (a) and aluminum-
doped SiC powder (b) in comparison with the reference pattern of the different SiC polytypes
3C (red bars, JCPDS 73-1665), 6H (blue bars, JCPDS 72-0018), 4H (purple bars, JCPDS 22-
1317) and 15R (green bars, JCPDS 39-1196). (Note: The peak at 66.9° is a machine-related
artefact, the peak at 26.6° is a preparation-induced agate contamination).
21
Fig. 3. XPS spectra (solid line) for Si 2p core level (a) and C 1s core level (b) with simulated
peak deconvolution of the components (broken lines) for SiC:N (top) and pristine SiC:Al
(bottom) microcrystalline powders. For better visibility, spectra have been normalized.
Fig. 4. Infrared reflection absorption (RAIRS) spectra of SiC:N (top curve) and SiC:Al (bottom
curve) microcrystalline powders. For better comparability, spectra have been normalized.
22
Table 1. Phase composition, unit cell dimensions and crystallite sizes determined by Rietveld
for a nitrogen-doped SiC powder.
SiC Polytype Composition
a
3C
6H
[%]
93
7
[Å]
4.3585
3.0807
c
[Å]
-
15.1174
Particle size
[nm]
200
30
Table 2. Phase composition, unit cell dimensions and crystallite sizes determined by Rietveld
for an aluminum-doped powder.
SiC Polytype Composition
a
Particle size
[nm]
100
100
30
3C
6H
4H
[%]
53
37
10
[Å]
4.3603
c
[Å]
-
3.0817
15.1275
3.0800
10.0980
23
|
1811.04135 | 1 | 1811 | 2018-11-09T21:09:14 | Metalized polymer tubes for high-frequency electromagnetic waveguiding | [
"physics.app-ph"
] | Low loss electromagnetic energy transport over long distances motivates the development of different types of waveguiding systems. Requirements of high quality optically polished waveguide surfaces needed in high-frequency applications and low-cost manufacturing are practically incompatible in current realizations. Here we demonstrate a new paradigm solution, based on surface functionalization with subsequent electroless plating of conductive micron smooth copper layer on the inner surface of flexible non-conducting poly-carbonate tubes. The structure was shown to support moderately low loss guiding performances (~5-10 dB/m) at Ku-band. The mechanically flexible design of the system allows shaping the waveguide network almost on demand. In particular, an efficient energy guiding over a closed loop with 8 lambda radius was demonstrated. The new platform of high quality metalized flexible waveguiding systems opens new opportunities in designs of cheap and efficient networks, operating over a broad spectral range, approaching tens of GHz and even higher. | physics.app-ph | physics | Metalized polymer tubes for high frequency electromagnetic waveguiding
Dmitry Filonov1,2,‡,=, Hahi Barhom1,=, Andrey Shmidt1, Yelena Sverdlov1,
Yosi Shacham-Diamand1, Amir Boag1, Pavel Ginzburg1,2
1School of Electrical Engineering, Tel Aviv University, Tel Aviv, 69978, Israel
2Light-Matter Interaction Centre, Tel Aviv University, Tel Aviv, 69978, Israel
‡[email protected]
=contributed equally
Abstract: Low loss electromagnetic energy transport over long distances motivates the development of
different types of waveguiding systems. Requirements of high quality optically polished waveguide surfaces
needed in high-frequency applications and low-cost manufacturing are practically incompatible in current
realizations. Here we demonstrate a new paradigm solution, based on surface functionalization with subsequent
electroless plating of conductive micron smooth copper layer on the inner surface of flexible non-conducting
poly-carbonate tubes. The structure was shown to support moderately low loss guiding performances (~5-10
dB/m) at Ku-band. The mechanically flexible design of the system allows shaping the waveguide network
almost on demand. In particular, an efficient energy guiding over a closed loop with 8 lambda radius was
demonstrated. The new platform of high quality metalized flexible waveguiding systems opens new
opportunities in designs of cheap and efficient networks, operating over a broad spectral range, approaching
tens of GHz and even higher.
Introduction
Efficient transport of electromagnetic energy over distances
requires the development of different types of waveguiding systems.
Many architectures have been demonstrated over the years and each
one provides solutions for a specific frequency range and related
applications[1]. Minimization of propagation and bending losses along
with inerrability within larger-scale devices are among the key
parameters, required from efficient interconnecting systems. Losses
become a critical factor, affecting the performance of millimeter wave
devices. Interface roughness on the level of optically polished high-
quality surfaces is required from the implementations. As a result, the
overall prices and bulky realizations of waveguiding components
become a significant factor. Minimization of bending losses can be
achieved by exploiting geometries, where electromagnetic radiation is
enclosed within a confined volume. Typical representative examples
here
circular/elliptical geometries.
Furthermore, the guided mode in those structures is confined in a void
and, as a result, propagation losses are minimized (especially, if the
core is vacuumed). Though some commercial solutions are available,
bending of enclosed waveguides, is quite a complex technological
challenge, as the millimeter-size aperture (Ka-band and higher) should
be maintained along the entire curved trajectory.
rectangular
include
and
is
Metallization of plastic components
fast developing
technological direction, as it can provide significant advantages over
conventional solutions. Several successful demonstrations of this
approach include 3D printed metalized waveguide filters [2],
waveguiding systems and related components [3], [4] (also for high
GHz-THz and without metallization [5],[6]), antenna devices and
components [7],[8],[9], and others. Furthermore,
is worth
mentioning other additive manufacturing techniques, which allow the
production of high-quality RF components, with an emphasis on
antenna devices [10]. Different types of antennas, fabricated with CNC
milling technique [11],[12], Laser Direct Structuring [13],[14],[15],
it
conformal printing of metallic inks [16], conductive inkjet printing
[17], ultrasonic wire mesh embedding [18], and metal deposition
trough a mask on a curve surface [19], [20], were reported.
Furthermore, integration of 3D printed plastic materials within antenna
designs was demonstrated (e.g. [21],[22]) and fabrication of low-
profile devices with several materials has been shown [23].
Fig. 1 Schematics of a metalized bendable dielectric tube
for high frequency waveguiding applications.
In general, three main advantages of 3D printed polymers-based
structures can be identified. The first one is their lightweights in
comparison with solid metal made counterparts. Since only a thin
conductive layer (skin depth of several microns for high GHz
frequencies, if conventional metals are in use) is required for providing
efficient waveguiding properties, lightweight polymers can serve as
bulk rigid materials, supporting the structure. The second advantage of
metalized 3D-printing approach to waveguiding systems is their
potential flexibility to provide quite complex geometries (e.g. complex
3D networks, power divides, and others), which are hard to obtain with
throughput flow of functional solutions through it. The cleaning cycle
is applied at the end of the metallization process. DIW (Di Ionized
Water) wash at room temperature was followed with methanol
absolute dry wash at 50○C for 5 minutes followed by rinsing with DIW.
Then N,N-Dimethylformamide solution at 50○C washed the tube for 5
minutes followed by rinsing with Ethanol then DIW. Etching to reduce
the roughness on the tube walls by Chromo-Phosphoric-Sulfuric acid
at 50○C for 5 minutes was performed and followed by DIW rinsing.
The next step is a sensibilization in a solution, containing SnCl2-70 g/l
and HCl -- 40 ml/l for 30 min at room temperature, followed by DIW
rinsing. Pd-activation for 60 min at room temperature with PdCl2 1g/l
solution, followed by DIW rinsing. Then copper solution 15 g/l, K-Na-
tartrate 30 g/l Na2CO3 10 g/l NaOH 40 g/l Formaldehyde 35%.
Adjusting pH in the range 12.5 -- 12.7 assuring metal layer formation
properly at a proper rate compared to the reduction time of copper.
The surface roughness of copper layer was measured in Olympus
LEXT OLS4100 laser scanning, providing a resolution of 10-20 nm.
The quality of the surface was defined as the standard deviation of
points on the surface from the mean position. The measurement was
performed over mm2-scale area. The surface roughness was estimated
to be around 10µm, while smaller areas of investigation were found to
be as smooth as ± 0.2µm. It indicates that more accurate and repetitive
surface cleaning can allow obtaining those numbers (corresponding to
surface qualities of current commercially available 60 GHz hollow
waveguide systems). The overall thickness of the metal layer,
deposited on the inner side of the tube, is few microns (depending on
copper solutions flow and duration), while the electromagnetic skin
depth in copper at 10 GHz frequency is less than a micron. Fig. 2
(inset) shows a photo of a cut section of the metalized tube. The
topographic image of the surface appears in the main figure (Fig. 2).
The slant of the surface corresponds to the physical profile behavior,
which can be almost perfectly fitted with a quadratic 2D polynomial.
The quasi-periodic strips of the height profile (if projected on the flat
landscape) are observed along the direction of the stream, which was
induced inside the tube during the metallization process.
Straight waveguide sections
A metalized tube section was assessed for the waveguiding at the
next stage and compared with a straight brass duct section. Standard
SMA to waveguide connectors were implemented on short sections of
brass ducts (inner diameter of 13 and 11mm in the case of brass and
metalized sections, respectively), sealed from the outer side (insets to
Fig. 3).
Fig. 2 Surface profile of 2.5 mm × 1 mm section of a metalized
tube. (Inset) Photograph of the cut section of the tube.
conventional milling techniques. It is worth noting that rigid 3D
structures, once being fabricated, cannot be reshaped. Our proposal,
however, is lacking of these limitations. The last and already
mentioned factor is the manufacturing cost, which can be significantly
reduced in the case when additive manufacturing techniques are in use.
Here we demonstrate a new waveguiding system, based on
metallization of flexible polymer tubes (Fig. 1). The distinctive
advantages of this architecture include its extremely low cost and
mechanical flexibility. Inner surfaces of initially bendable polymer
tubes are metalized with an electroless plating technique and, as a
result, high-quality RF conductivity is achieved. Furthermore,
controllable chemical deposition process was demonstrated to provide
high-quality metallization layers with micron-scale roughness along
centimeter-range tube cross-sections. Mechanical flexibility of the
tube waveguides allows bending them almost on demand without
requiring an a priori knowledge of a layout of an end-user.
The Letter is organized as follows: basic technological principles
of the metallization are discussed first and then followed by
experimental demonstration of the new system and its performance,
which is compared with standard existing solutions.
Tubes metallization
Tubes with different polymeric composites can be metalized by
electroless deposition of a copper layer. As a first step, surface of the
tubes was washed from excess or leftover materials that stay after
chemical treatment and pre-production post-processing. The tube was
connected to a homemade peristaltic pump, which maintains a high
Fig. 3 S-parameters (absolute values) of circular waveguide system (S11 -- blue dots, S12 -- red solid lines). (a) Numerical simulation. (b) Straight
section of a brass duct (inner diameter is 13mm). (c) Metalized polymer tube (inner diameter is 11 mm). Sections are 26cm in length.
Fig. 4 S-parameters (absolute values) of the bended waveguide system (S11 -- blue dots, S12 -- red solid lines). (a) Numerical simulation. (b)
Metalized polymer tube (11 mm inner diameter, 1 meter length, 16 cm bending radius).
Transmission and reflection coefficients (S-parameters) of the system
were retrieved with the VNA (N5232B PNA-L Microwave Network
Analyzer) after performing calibration procedures (Fig. 3). Numerical
simulation of perfectly conducting (PEC) tube section was performed
with CST microwave studio (frequency domain solver). Custom-made
connectors were explicitly
the
excitation/collection ports were placed directly on the coaxial cables.
As a result, imperfect mode matching was emulated, replicating the
experimental scenario. The cut-off frequency of the waveguide with
circular cross-section has a well known analytical expression.
Substitution of the tube's parameters leads to the cut-off at ~13.5 GHz,
which agrees well with the numerical results.
into account, while
taken
Fig. 3(b) is the reference measurement, demonstrating the
performances of an etalon structure -- a brass duct. The results agree
well with the numerical data with certain deviations, including the
absolute values of S-parameters, which correspond to the leakage of
the radiation via connection ports. This effect is more pronounced for
higher frequencies, where electromagnetic fields squeeze through gaps
between the connectors, which are imperfectly attached to the duct's
surface. The performance of the metalized tube appears on Fig. 3(c).
The structure has got a slight distortion during the metallization
process. This can be prevented in the future by mechanical
stabilization of the whole section during either pre- or post-processing.
The direct comparison between the brass (etalon) and metalized
sections show remarkable similarities, nevertheless, the overall
amount of loss is 2-6 dB higher. Also, the cut-off frequency is shifted
to ~16GHz, since the inner diameter of the metallized polymer section
is 11mm, which corresponds to this number.
Bended waveguides
The distinctive advantage of the flexible waveguide platform is to
support bending geometries - Fig. 4 (b, inset). Here, the overall length
of the metalized 11mm diameter tube is 1m, while the bending radius
is 16 cm. Fig. 4(a) shows the result of the numerical simulation, carried
out under the same conditions, as for the straight section (the PEC duct
+ connectors). The cut-off frequency is clearly observed at 16 GHz
and it is more pronounced in this case since this bent waveguide
section is 3 times longer than the straight section (Fig. 3). Fig. 4(b)
shows the experimental results, which correspond well to the
numerical estimate. Similarly to the case of the straight section, the
losses of the real structure are higher by ~6 dB, which is consistent
with the qualitative comparison, which was performed on the data
from Fig. 3.
Conclusions
A new platform for cheap and flexible high-frequency guiding was
proposed and experimentally demonstrated. The principle is based on
electroless plating of smooth bendable polymer tubes, whose inner
surfaces undergo metallization. Highly conductive chemically
deposited layers demonstrate properties, comparable with optically
polished surfaces of conventional brass-based waveguide sections. As
a result, efficient waveguiding at Ku-band was demonstrated. It is
worth noting, that the developed methodology can be extended to
higher frequencies, approaching 100 GHz and even higher. Standard
waveguiding solutions demonstrate excessive
those
frequencies. Furthermore, waveguide components become extremely
expensive, preventing high-frequency RF technologies to replace
existing widespread low GHz solutions (e.g. Wi-Fi and many others).
Our solution might suggest a paradigm shift in this area.
losses at
Acknowledgments
The research was supported in part by PAZY Foundation, 3PEMS
Ltd.
References:
[1]
[2]
[3]
[4]
D. M. Pozar, Microwave engineering. Wiley, 2012.
E. Massoni, M. Guareschi, M. Bozzi, L. Perregrini, U. A.
Tamburini, G. Alaimo, S. Marconi, F. Auricchio, and C.
Tomassoni, "3D printing and metalization methodology for
high dielectric resonator waveguide microwave filters," in
2017 IEEE MTT-S International Microwave Workshop
Series on Advanced Materials and Processes for RF and
THz Applications (IMWS-AMP), 2017, pp. 1 -- 3.
M. I. M. Ghazali, K. Y. Park, V. Gjokaj, A. Kaur, and P.
Chahal, "3D Printed Metalized Plastic Waveguides for
Microwave Components," Int. Symp. Microelectron., vol.
2017, no. 1, pp. 000078 -- 000082, Oct. 2017.
E. A. Rojas-Nastrucci, J. T. Nussbaum, N. B. Crane, and T.
M. Weller, "Ka-Band Characterization of Binder Jetting for
3-D Printing of Metallic Rectangular Waveguide Circuits
and Antennas," IEEE Trans. Microw. Theory Tech., 2017.
[5]
M. Weidenbach, D. Jahn, A. Rehn, S. F. Busch, F. Beltrán-
[6]
[7]
[8]
[9]
Mejía, J. C. Balzer, and M. Koch, "3D printed dielectric
rectangular waveguides, splitters and couplers for 120
GHz," Opt. Express, vol. 24, no. 25, p. 28968, Dec. 2016.
D. W. Vogt, J. Anthony, and R. Leonhardt, "Metallic and
3D-printed dielectric helical terahertz waveguides," Opt.
Express, vol. 23, no. 26, p. 33359, Dec. 2015.
B. Zhang, P. Linner, C. Karnfelt, P. L. Tarn, U. Sodervall,
and H. Zirath, "Attempt of the metallic 3D printing
technology for millimeter-wave antenna implementations,"
in 2015 Asia-Pacific Microwave Conference (APMC), 2015,
vol. 2, pp. 1 -- 3.
E. G. Geterud, P. Bergmark, and J. Yang, "Lightweight
Waveguide and Antenna Components Using Plating on
Plastics," 7th Eur. Conf. Antennas Propogation, 2013.
R. Zhu and D. Marks, "Rapid prototyping lightweight
millimeter wave antenna and waveguide with copper
plating," in IRMMW-THz 2015 - 40th International
Conference on Infrared, Millimeter, and Terahertz Waves,
2015.
[10] M. Liang, J. Wu, and X. Yu, "3D printing technology for RF
and THz antennas," Ieee, no. 2, pp. 536 -- 537, Jan. 2016.
[11] M. Ferrando-Rocher, J. I. Herranz, A. Valero-Nogueira, and
B. Bernardo, "Performance Assessment of Gap Waveguide
Array Antennas: CNC Milling vs. 3D Printing," IEEE
Antennas Wirel. Propag. Lett., vol. 1225, no. c, pp. 1 -- 1,
2018.
[18] M. Liang, C. Shemelya, E. MacDonald, R. Wicker, and H.
Xin, "3-D Printed Microwave Patch Antenna via Fused
Deposition Method and Ultrasonic Wire Mesh Embedding
Technique," IEEE Antennas Wirel. Propag. Lett., vol. 14,
pp. 1346 -- 1349, 2015.
[19]
[20]
I. Ehrenberg, S. Sarma, T. Steffeny, and B.-I. Wuy,
"Fabrication of an X-Band conformal antenna array on an
additively manufactured substrate," in 2015 IEEE
International Symposium on Antennas and Propagation &
USNC/URSI National Radio Science Meeting, 2015, pp.
609 -- 610.
B.-I. Wu and I. Ehrenberg, "Ultra conformal patch antenna
array on a doubly curved surface," in 2013 IEEE
International Symposium on Phased Array Systems and
Technology, 2013, pp. 792 -- 798.
[21] M. Mirzaee, S. Noghanian, and I. Chang, "Low-profile
bowtie antenna with 3D printed substrate," Microw. Opt.
Technol. Lett., vol. 59, no. 3, pp. 706 -- 710, Mar. 2017.
[22]
[23]
C. Shemelya, M. Zemba, M. Liang, X. Yu, D. Espalin, R.
Wicker, H. Xin, and E. MacDonald, "Multi-layer
archimedean spiral antenna fabricated using polymer
extrusion 3D printing," Microw. Opt. Technol. Lett., vol. 58,
no. 7, pp. 1662 -- 1666, Jul. 2016.
P. Parsons, M. Mirotznik, P. Pa, and Z. Larimore, "Multi-
material additive manufacturing of embedded low-profile
antennas," Electron. Lett., vol. 51, no. 20, pp. 1561 -- 1562,
Oct. 2015.
[12] M. A. Al-Tarifi and D. S. Filipovic, "On the design and
fabrication of W-band stabilised-pattern dual-polarised horn
antennas with DMLS and CNC," IET Microwaves, Antennas
Propag., vol. 11, no. 14, pp. 1930 -- 1935, Nov. 2017.
[13]
[14]
[15]
[16]
F. Sonnerat, R. Pilard, F. Gianesello, F. Le Pennec, C.
Person, and D. Gloria, "Innovative LDS Antenna for 4G
Applications," IEEE, no. Eucap, pp. 2696 -- 2699, Apr. 2013.
A. Friedrich, M. Fengler, B. Geck, and D. Manteuffel, "60
GHz 3D integrated waveguide fed antennas using laser
direct structuring technology," in 2017 11th European
Conference on Antennas and Propagation, EUCAP 2017,
2017, pp. 2507 -- 2510.
A. Friedrich and B. Geck, "On the Design of a 3D LTE
Antenna for Automotive Applications based on MID
Technology," Eur. Microw. Conf., pp. 640 -- 643, 2013.
J. J. Adams, S. C. Slimmer, T. F. Malkowski, E. B. Duoss, J.
A. Lewis, and J. T. Bernhard, "Comparison of Spherical
Antennas Fabricated via Conformal Printing: Helix,
Meanderline, and Hybrid Designs," IEEE Antennas Wirel.
Propag. Lett., vol. 10, pp. 1425 -- 1428, 2011.
[17] M. Ahmadloo, "Design and fabrication of geometrically
complicated multiband microwave devices using a novel
integrated 3D printing technique," in 2013 IEEE 22nd
Conference on Electrical Performance of Electronic
Packaging and Systems, 2013, pp. 29 -- 32.
|
1803.07137 | 1 | 1803 | 2018-02-21T16:21:56 | 2D Hydrogenated graphene-like borophene as a high capacity anode material for improved Li/Na ion batteries: A first principles study | [
"physics.app-ph",
"physics.comp-ph"
] | Fast-growing electronics industry and future energy storage needs have encouraged the design of rechargeable batteries with higher storage capacities, and longer life times. In this regard, two-dimensional (2D) materials, specifically boron and carbon nanosheets, have garnered enthusiasm due to their fascinating electronic, optical, mechanical and chemical properties. Recently, a hydrogen boride (HB) nanosheet was successfully fabricated showing remarkable stability and superior physical properties. Motivated by this experimental study, we used first principle electronic structure calculations to study the feasibility of this nanosheet to serve as an anode material for Li/Na/Ca/Mg/Al ion batteries. Most active adsorption sites for single adatoms were evaluated and next adatoms were gradually inserted into the anode surface accordingly. The charge transfer, electronic density of sates, storage capacity, structural stability, open-circuit potential and diffusion energy barriers were explored. Our theoretical study predicts that HB shows outstanding electrode properties for Li and Na ion batteries. The intercalation of both Li and Na adatoms into the HB monolayer can lead to a high identical storage capacity of 1133.8 mAh/g which is promising compared to the capacities of the traditional anode materials; such as graphite (372 mAh/g) and TiO2 (200 mAh/g), and other 2D materials; such as germanene (369 mAh/g), stanene (226 mAh/g), and phosphorene (432.8 mAh/g) nanosheets. These results may open a new horizon for the design of rechargeable batteries with higher storage capacitates. | physics.app-ph | physics | 2D Hydrogenated Graphene-like Borophene as a High Capacity Anode
Material for Improved Li/Na Ion Batteries: A First Principles Study
Meysam Makaremi,1 Bohayra Mortazavi,2 and Chandra Veer Singh*,1,3
1Department of Materials Science and Engineering, University of Toronto, 184 College Street,
2Institute of Structural Mechanics, Bauhaus-Universität Weimar, Marienstr. 15,
Suite 140, Toronto, ON M5S 3E4, Canada.
3Department of Mechanical and Industrial Engineering, University of Toronto, 5 King's College
D-99423 Weimar, Germany.
Road, Toronto M5S 3G8, Canada.
Abstract
Fast-growing electronics industry and future energy storage needs have encouraged the design of
rechargeable batteries with higher storage capacities, and longer life times. In this regard, two-dimensional
(2D) materials, specifically boron and carbon nanosheets, have garnered enthusiasm due to their fascinating
electronic, optical, mechanical and chemical properties. Recently, a hydrogen boride (HB) nanosheet was
successfully fabricated showing remarkable stability and superior physical properties. Motivated by this
experimental study, we used first principle electronic structure calculations to study the feasibility of this
nanosheet to serve as an anode material for Li/Na/Ca/Mg/Al ion batteries. Most active adsorption sites for
single adatoms were evaluated and next adatoms were gradually inserted into the anode surface accordingly.
The charge transfer, electronic density of sates, storage capacity, structural stability, open-circuit potential
and diffusion energy barriers were explored. Our theoretical study predicts that HB shows outstanding
electrode properties for Li and Na ion batteries. The intercalation of both Li and Na adatoms into the HB
monolayer can lead to a high identical storage capacity of 1133.8 mAh/g which is promising compared to
the capacities of the traditional anode materials; such as graphite (372 mAh/g) and TiO2 (200 mAh/g), and
other 2D materials; such as germanene (369 mAh/g), stanene (226 mAh/g), and phosphorene (432.8 mAh/g)
nanosheets. These results may open a new horizon for the design of rechargeable batteries with higher
storage capacitates.
1
1. Introduction
Rechargeable metal-ion batteries play critical roles for a variety of advanced technologies such
as mobile-electronics, communication devices and electric vehicles.1–4 In this regard, the use of
advanced materials with higher charge capacities and faster ion diffusion rates are extremely in
demand to replace common anode and cathode materials. Among various candidates, 2D materials
and their hybrid structures are among the most promising solutions owing to their large surface
area and superior electronic, thermal and mechanical properties.5–11 Previous theoretical studies
have confirmed that 2D materials can yield remarkably high charge capacities12–15 and ultralow
diffusion energy barriers16–21 for Li and Na-ion batteries.
Graphene's outstanding physical, chemical and electronic properties22–24 have led to the
introduction of the diverse family of two-dimensional (2D) materials.25 A few of which are single
elemental nanosheets; such as phosphorene, silicene, germanene and stanene; although, they
cannot forms versatile bonding like sp3, sp2 and sp like graphene26, these structures present limited
bond length flexibility, and their equilibrated single-layers exist in out-of-plane buckled
configurations. On the other side, boron, the neighboring element of carbon in the periodic table,
can similarly form diverse bonding and it involves diverse forms, from zero-dimensional to three-
dimensional crystals.26–29
During the last couple of years, 2D boron sheets so called borophene with the buckled30 and
flat31 atomic configurations have been successfully fabricated using the epitaxy growth of boron
atoms on the silver metallic surface. Followed by these experimental reports, numerous theoretical
studies have been achieved to explore the application of 2D boron based nanomembranes for
various applications such as hydrogen storage32,33, rechargeable metal-ion batteries34–40,
superconducting, magnetic, electronic, and chemical devices39,41–45, and mechanically robust
2
components.46–48 Most recently, experimental realization and characterization of 2D hydrogen
boride sheets with an empirical formula of H1B1 was successfully achieved by exfoliation and
complete ion-exchange between protons and magnesium cations in magnesium diboride.49 Worthy
to note that the existence and stability of such a 2D structure has already been theoretically
proven.50
In this regard, pristine and functionalized borophene films have been also theoretically realized
to yield remarkably high charge capacities, good electronic conductivity and low energy diffusion
barriers, which are all highly desirable for the application as anode materials in the rechargeable
battery industry.32–36,51 2D borophene structures were synthesized by growing on a substrate, and
develop of efficient transferring methods for lifting 2D nanosheets from the metallic substrate in
order to reach isolated nanomembranes has been a significant challenge to date.52 Moreover, for
the application in rechargeable batteries, borophene films have to be producible in large content
and low-cost.36 Interestingly, a recently fabricated hydrogenated borophene structure, hydrogen
boride49, was in multi-layer form and more importantly the nanomembranes were not grown on a
substrate, which accordingly can serve as promising signs toward their real application in
commercial ion batteries.
The successful synthesis of hydrogen boride (HB) raises an important question, whether this
new nanomembrane can serve as a promising anode material in rechargeable batteries or not. To
address this important question, we conducted extensive density functional theory (DFT)
simulations to explore the possible application of the HB monolayer for Na, Ca, Al, Mg, and Li-
ion batteries. Our first-principles analysis of charge capacity, adsorption energies, open-circuit
voltage profiles, dynamical stabilities, electronic structures and diffusion energy barriers confirm
the promising candidacy of hydrogen boride for Na and Li-ions storage.
3
2. Computational Details
Implemented in the Vienna Ab-initio Simulation Package (VASP)53 framework, Spin polarized
Perdew–Burke–Ernzerhof (PBE)54 density functional theory (DFT) calculations were used
through generalized gradient approximation (GGA) and projector augmented-wave (PAW)
potentials55. A large kinetic energy cutoff of 500 eV. Since the GGA underestimates the
binding energies56,57, a Grimme dispersion correction approach, DFT-D258 were selected to
accurately calculate energies comparable with experimental cohesive energies as well as
binding energies. For electronic self-consistency and ionic relaxation convergence, criteria of
1x10-6 eV and 1x10-3 eV/Å were considered. And to integrate the Brillouin zone, a Monkhorst-
Pack mesh of 15x15x1 and the tetrahedron scheme with Blöchl corrections were applied.
The binding energy were calculated as,
E"#$% =()*+,-.)*+.$×)-)
,
$
where E1"23, E1", E3, and n are the total interaction energy, the energy of the pristine nanosheet,
accordingly. More negative values of E"#$% indicates the stronger binding between the surface and
the energy of a single adatom in the gas phase, and the number of intercalating foreign atoms;
(1)
adatoms.
The Bader analysis technique59 were employed to probe charge gain and loss. The system charge
difference can be determined by,
Δρ=ρ1"23−ρ1"−ρ3 ,
where, ρ1"23, ρ1", and ρ3 define the electron densities of HB+M, HB, and metallic foreign atom,
(2)
respectively.
4
electrode, and it can be calculated by using the following equation,
Open-circuit voltage (V)60 is a critical parameter indicating the performance of an anode
here, xA, xB, and x ( xA≤x≤xB), respectively, involve the initial, final and average coverage
ratios. e is the electron charge; and E3=<1", E3=;1", and E3+are the interaction energy for the
coverage ratio of xAand xB, and the metallic bulk energy, respectively.
V≈9(:;.:<))-+. ()-=;*+.)-=<*+)>
(:;.:<)?
,
(3)
As illustrated in Figure 1, the HB unitcell consists of 8 (4 hydrogen and 4 boron) atoms and
involves an orthorhombic structure with lattice constants a = 3.02 Å and b = 5.29 Å. More details
about the HB structure are listed in Tables S1 and S2. For all simulations, we used a supercell
composed of 4 × 2 unitcells including 32 H and 32 B atoms, and a vacuum space of 20 Å in the
z direction. For each type of adatom, four possible sites were analyzed to find the most favorable
adsorption point (see Fig. 1b).
To calculate the battery storage capacity, the adatoms were gradually added to the nanosheet at
predefined sites randomly and uniformly until the maximal surface coverage with the highest
possible capacity were achieved. Storage capacity is calculated by using Faraday formula, as
q=1000 F z$KL=3*+ ,
follows:
here F, z, nOP:, and M1" are the Faraday constant, atomic valence number, maximal adatom
(4)
surface coverage, and the molecular mass of the HB single layer.
Thermal stability of relaxed structures was analyzed by performing ab-initio molecular
dynamics (AIMD) simulations in canonical ensemble (NVT) at 300 K with a time step of 1 fs for
10000 steps.
5
3. Results and Discussion
As shown in Fig. 1b, the binding energy of the insertion of five types of adatoms consisting of
Li, Na, Mg, Ca, and Al at four possible sites of the HB monolayer were calculated to evaluate the
ability of HB nanosheet as an anode electrode for alkali and alkali-earth ion batteries. The
calculations predict that all of the foreign atoms prefer to be adsorbed at the hexagonal hollow site
(Site 4 in Fig. 1b) resulting in adsorption energies of -2.32, -1.27, -0.23, -1.46, and -2.17 eV for
Li, Na, Mg, Ca, and Al; respectively. Also, the differential charge density plots of the relaxed
structures of the most stable configurations of the monolayer interacting with two types of
adatoms, Li and Na, are shown in Figure 2. It suggests that the HB surface strongly interacts with
both alkali elements and accept electron charge densities from them. The binding energy
calculations predict pretty intense binding between the HB surface and most of the adatoms (except
Mg); however, since the initial voltage for the intercalation of Mg, Ca, and Al into the surface is
negative (-1.56, -0.60, and -1.57 V for Mg, Ca, and Al; respectively.), the monolayer cannot be
applied for these types of rechargeable batteries. Therefore, in this study we specifically focused
on anode properties of the HB nanosheet for Li/Na ion storage applications. We added different
number of Li/Na adatoms to the HB monolayer (see Figure S1 and S2), and investigated its
response as an anode material.
It is worthwhile nothing that the electronic conductivity of an anode electrode is a critical
parameter influencing the performance and controlling its internal electronic resistance. Since
during charging and discharging, a battery generate the ohmic heat, a proper anode need to be
metallic and keep this property during charge/discharge cycles.35 The density of states (DOS)
6
diagrams for HB monolayers covered with various number of adatoms are presented in Figure 3
suggesting that the pure HB system and all of the systems involving both HB and foreign atoms
contain the zero-energy gap and present the conducting behavior.
Fig. 4a shows the HB binding energy as a function of adatom concentration. Due to the smaller
size and higher activity of the Li element compared to that of the Na type, as presented in Figs. 4b
and 4c the former group of atoms tend to migrate closer to the surface. Therefore, the insertion of
Li adatoms leads to more negative energies compared to the binding of the latter counterparts. It
should be noted that the energy remains almost constant (~ -1.3 eV) during the whole Li adsorption
spectrum, while it increases with respect to the Na intercalation for x > 0.5.
The open-circuit voltage is a key aspect showing the performance and the capacity of an anode
electrode. The negative values of the potential difference suggest that foreign adatoms prefer to
form metallic clusters instead of adsorption to the electrode. It should be noted that the favored
potential range for anode materials involves a spectrum of 0.1-1 V.18 Fig. 5a depicts the potential
as a function of storage capacity, and illustrates that the HB anode contains average voltages of
0.65 vs Li/Li+ and 0.03 V vs Na/Na+ for the Li and Na insertion, respectively, which remain
constant during the whole adsorption process until reaching the optimal capacity point of x = 0.5.
For comparison, the potential range for flat borophene36, borophane35, and TiO261 anodes reported
to be 0.5-1.8, 0.03-0.6, and 1.5-1.8 V; respectively.
It is worthwhile to note that the HB anode presents an outstanding identical capacity of 1133.8
mAh/g for both of the Li and Na ion storage batteries at x = 0.5. Before the optimal capacity (x <=
0.5) the intercalation behavior of Li adatoms is consistent with that of Na ones, leading to the
creation of a single layer of adatoms in each side of the monolayer (see Figs. 4b and 4c). However,
after reaching the optimal point, as can be seen in Fig. 5a the binding energies are not similar for
7
the two types of adatoms resulting in two different adsorption scenarios. After the optimal
concentration (x > 0.5), highly active Li atoms destroy the monolayer structure by attacking the
HB bonds (see Fig. 5b). Whereas, Na atoms prefer to limit the anode capacity by creating the
second layer of adatoms instead of destroying the monolayer structure (see Fig. 5c), and preventing
additional electronic charge transfer (see Figure S3). After reaching the optimal capacity, the
stability of each structure was evaluated by AIMD simulations at 300 K. The variations of
the temperature as a function of the simulation time is illustrated in Figure S4. Our AIMD
simulations confirm that the HB monolayer remains intact upon the adatoms adsorption
and represents thermal stability for the anode material usage.
For comparison it should be noted that the theoretical charge capacities of commercial graphite62
and TiO263 structures for the Li ion storage were reported to be 372 and 200 mAh/g, respectively.
2D materials including germanene64, stanene64, phosphorene65 nanosheets illustrate storage
capacities of 369, 226, and 432.79 mAh/g, accordingly. Also, studied in the last couple of years,
boron 2D structures consisting of buckled borophene66, flat borophene31, and borophane35 were
predicted to include capacities of 1720, 1980, and 504 mAh/g, respectively.
It is worthwhile noting that among all 2D materials considered in the literature for the anodic
application, only a few of them including flat and buckled borophene present higher storage
capacities compared to hydrogen boride. Nevertheless, one should not forget that, these high
capacity materials have been synthesized by growing on a substrate and their isolated layers have
not been fabricated yet. Whereas, the hydrogen boride nanosheet has been successfully fabricated
and isolated by a cation-exchange exfoliation method, which presents stability required for the
metal-ion battery application and further for the mass production.
8
Diffusion of adatoms through the anode material is an essential characteristic indicating the
charge/discharge ability of the material. We used the Nudged-elastic band (NEB) method to find
the minimum energy paths. The result for the diffusion of Li/Na adatoms on the monolayer through
various possible pathways is depicted in Figure 6. For both adatoms, the diffusion via Path 2 (over
the B-B band) leads to the lowest energy barriers involving 0.86 and 0.32 eV for Li and Na;
respectively. The lower barrier for the Na adatom might be described by weaker interaction
between the Na atom with the HB surface (-1.27 eV) compared to that of the small Li element (-
2.32 eV), which causes an easier and faster transfer rate for this adatom. The Li and Na energy
barriers for the HB nanosheet are close to the data for flat borophene including 0.69 and 0.34 eV
reported by an earlier theoretical work36. It should be mentioned that the diffusion barrier for Li
adatom transfer on other 2D monolayers such as graphene67, CEY graphene68, silicene69, MoS270,
and Ti3C2 MXene71 consists of 0.32, 0.58, 0.23, 0.25, and 0.7 eV; respectively. Moreover, the
energy barrier for commercial anode materials based on TiO2 ranges from 0.35 to 0.65 eV72,73, and
for graphite74 it is around 0.4 eV.
In addition, Figs. 6b and 6c show that the movement of single adatoms along other paths is more
energetically demanding. The barriers for the Li and Na movement over the hydrogen atom (along
Path 2) are 1.42 and 0.66 eV, and through the hexagonal hole (along Path 3) consist of 10.41 (not
shown in Fig. 6c) and 2.21 eV; respectively. The result suggests that diffusion of Na along the
latter path might be impossible due to the larger radius of this element compared to that of Li.
9
4. Conclusions
Extensive spin polarized DFT-D2 simulations were carried out to investigate the performance
of a newly fabricated hydrogenated graphene-like borophene, hydrogen boride, nanosheet to serve
as an anode material for Li/Na/Mg/Ca/Al ion storage applications. Different calculations such as
adsorption energy, average potential, diffusion barrier, density of states and Bader charge analyses
were conducted. It is found that while the insertion of Mg, Ca, and Al into the nanosheet may not
be feasible, the binding of Li and Na can result in superior anode electrodes. Our calculations
suggest that the HB monolayer interacting with Li and Na adatoms show high electronic
conductivity, drastic structural stability, low charging voltage, and versatile storage capacity.
Diffusion of Li and Na adatoms upon the monolayer includes barrier energies of 0.86 and 0.32
eV, accordingly. The Li and Na insertions result in open-circuit potentials of 0.65 vs Li/Li+ and
0.03 V vs Na/Na+. The intercalation of both types of adatoms into HB leads to an identical capacity
of 1133.8 mAh/g resulting in one of the highest storage capacities ever calculated for 2D
nanosheets. Only a few 2D nanostructures were theoretically predicted to illustrate similar or
higher capacities; although, since these structures were not physically isolated, yet they cannot be
used as anode materials. The result of this study is promising and we hope it will shed light on the
development of the new generation of Li/Na ion batteries for modern technologies including
portable electronics, communication applications, aerospace devices and electric vehicles.
AUTHOR INFORMATION
Corresponding Author
* [email protected]
10
ACKNOWLEDGMENT
MM and CVS gratefully acknowledge their financial support in parts by Natural Sciences and
Engineering Council of Canada (NSERC), University of Toronto, Connaught Global Challenge
Award, and Hart Professorship. The computations were carried out through Compute Canada
facilities, particularly SciNet and Calcul-Quebec. SciNet is funded by the Canada Foundation for
Innovation, NSERC, the Government of Ontario, Fed Dev Ontario, and the University of Toronto,
and we gratefully acknowledge the continued support of these supercomputing facilities. BM
greatly acknowledges the financial support by European Research Council for COMBAT project
(Grant number 615132).
11
References
(1) Sun, Y.; Liu, N.; Cui, Y. Promises and Challenges of Nanomaterials for Lithium-Based
Rechargeable Batteries. Nature Energy. 2016.
(2) Xu, J.; Dou, Y.; Wei, Z.; Ma, J.; Deng, Y.; Li, Y.; Liu, H.; Dou, S. Recent Progress in
Graphite Intercalation Compounds for Rechargeable Metal (Li, Na, K, Al)-Ion Batteries.
Advanced Science. 2017.
(3) Liu, T.; Ding, J.; Su, Z.; Wei, G. Porous Two-Dimensional Materials for Energy
Applications: Innovations and Challenges. Materials Today Energy. 2017, pp 79–95.
(4) Shi, Z. T.; Kang, W.; Xu, J.; Sun, Y. W.; Jiang, M.; Ng, T. W.; Xue, H. T.; Yu, D. Y. W.;
Zhang, W.; Lee, C. S. Hierarchical Nanotubes Assembled from MoS2-Carbon Monolayer
Sandwiched Superstructure Nanosheets for High-Performance Sodium Ion Batteries.
Nano Energy 2016, 22, 27–37.
(5) Thackeray, M. M.; Wolverton, C.; Isaacs, E. D. Electrical Energy Storage for
Transportation-approaching the Limits Of, and Going Beyond, Lithium-Ion Batteries.
Energy Environ. Sci. 2012, 5 (7), 7854.
(6) Novoselov, K. S.; Mishchenko, A.; Carvalho, A.; Castro Neto, A. H. 2D Materials and
van Der Waals Heterostructures. Science (80-. ). 2016, 353 (6298), aac9439.
(7) Wang, X.; Xia, F. Van Der Waals Heterostructures: Stacked 2D Materials Shed Light.
Nature Materials. 2015, pp 264–265.
(8) Geim, A. K.; Grigorieva, I. V. Van Der Waals Heterostructures. Nature. 2013, pp 419–
425.
(9) Xu, M.; Liang, T.; Shi, M.; Chen, H. Graphene-like Two-Dimensional Materials.
Chemical Reviews. 2013, pp 3766–3798.
(10) Tang, J.; Chen, D.; Yao, Q.; Xie, J.; Yang, J. Recent Advances in Noble Metal-Based
Nanocomposites for Electrochemical Reactions. Materials Today Energy. 2017, pp 115–
127.
(11) Zhu, Y.; Chen, G.; Zhong, Y.; Zhou, W.; Liu, M.; Shao, Z. An Extremely Active and
Durable Mo2C/graphene-like Carbon Based Electrocatalyst for Hydrogen Evolution
Reaction. Mater. Today Energy 2017, 6, 230–237.
(12) Hwang, H.; Kim, H.; Cho, J. MoS2 Nanoplates Consisting of Disordered Graphene-like
Layers for High Rate Lithium Battery Anode Materials. Nano Lett. 2011, 11 (11), 4826–
4830.
(13) Eftekhari, A. Molybdenum Diselenide (MoSe2) for Energy Storage, Catalysis, and
Optoelectronics. Appl. Mater. Today 2017, 8, 1–17.
(14) Hu, J.; Xu, B.; Yang, S. A.; Guan, S.; Ouyang, C.; Yao, Y. 2D Electrides as Promising
Anode Materials for Na-Ion Batteries from First-Principles Study. ACS Appl. Mater.
Interfaces 2015, 7 (43), 24016–24022.
(15) Zhou, Q.; Wu, M.; Zhang, M.; Xu, G.; Yao, B.; Li, C.; Shi, G. Graphene-Based
Electrochemical Capacitors with Integrated High-Performance. Mater. Today Energy
2017, 6, 181–188.
(16) Samad, A.; Noor-A-Alam, M.; Shin, Y.-H. First Principles Study of a SnS2/graphene
Heterostructure: A Promising Anode Material for Rechargeable Na Ion Batteries. J.
Mater. Chem. A 2016, 4 (37), 14316–14323.
(17) Samad, A.; Shafique, A.; Kim, H. J.; Shin, Y.-H. Superionic and Electronic Conductivity
in Monolayer W2C: Ab Initio Predictions. J. Mater. Chem. A 2017, 5 (22), 11094–11099.
12
(18) Cak[i without dot]r, D.; Sevik, C.; Gulseren, O.; Peeters, F. M. Mo2C as a High Capacity
Anode Material: A First-Principles Study. J. Mater. Chem. A 2016, 4 (16), 6029–6035.
(19) Liu, Y.; Peng, X. Recent Advances of Supercapacitors Based on Two-Dimensional
Materials. Appl. Mater. Today 2017, 8, 104–115.
(20) Mortazavi, B.; Shahrokhi, M.; Makaremi, M.; Rabczuk, T. Theoretical Realization of
Mo<inf>2</inf>P; a Novel Stable 2D Material with Superionic Conductivity and
Attractive Optical Properties. Appl. Mater. Today 2017, 9.
(21) Sengupta, A.; Frauenheim, T. Lithium and Sodium Adsorption Properties of Monolayer
Antimonene. Mater. Today Energy 2017, 5, 347–354.
(22) Novoselov, K. S.; Geim, A. K.; Morozov, S. V; Jiang, D.; Zhang, Y.; Dubonos, S. V;
Grigorieva, I. V; Firsov, A. A. Electric Field Effect in Atomically Thin Carbon Films.
Science 2004, 306 (5696), 666–669.
(23) Geim, A. K.; Novoselov, K. S. The Rise of Graphene. Nat. Mater. 2007, 6 (3), 183–191.
(24) Li, W.; Zhang, Z.; Tang, Y.; Bian, H.; Ng, T.-W.; Zhang, W.; Lee, C.-S. Graphene-
Nanowall-Decorated Carbon Felt with Excellent Electrochemical Activity Toward VO 2 +
/VO 2+ Couple for All Vanadium Redox Flow Battery. Adv. Sci. 2015, 3, 1500276 1-7.
(25) Roome, N. J.; Carey, J. D. Beyond Graphene: Stable Elemental Monolayers of Silicene
and Germanene. ACS Appl. Mater. Interfaces 2014, 6 (10), 7743–7750.
(26) Zhang, Z.; Penev, E. S.; Yakobson, B. I. Two-Dimensional Boron: Structures, Properties
and Applications. Chem. Soc. Rev. 2017.
(27) Yang, Y.; Zhang, Z.; Penev, E. S.; Yakobson, B. I. B40 Cluster Stability, Reactivity, and
Its Planar Structural Precursor. Nanoscale 2017, 9 (5), 1805–1810.
(28) Kondo, T. Recent Progress in Boron Nanomaterials. Sci. Technol. Adv. Mater. 2017, 18
(1), 780–804.
(29) Zhang, Y.; Wu, Z. F.; Gao, P. F.; Zhang, S. L.; Wen, Y. H. Could Borophene Be Used as
a Promising Anode Material for High-Performance Lithium Ion Battery? ACS Appl.
Mater. Interfaces 2016, 8 (34), 22175–22181.
(30) Mannix, A. J.; Zhou, X.-F.; Kiraly, B.; Wood, J. D.; Alducin, D.; Myers, B. D.; Liu, X.;
Fisher, B. L.; Santiago, U.; Guest, J. R.; Yacaman, M. J.; Ponce, A.; Oganov, A. R.;
Hersam, M. C.; Guisinger, N. P. Synthesis of Borophenes: Anisotropic, Two-Dimensional
Boron Polymorphs. Science (80-. ). 2015, 350 (6267), 1513–1516.
(31) Feng, B.; Zhang, J.; Zhong, Q.; Li, W.; Li, S.; Li, H.; Cheng, P.; Meng, S.; Chen, L.; Wu,
K. Experimental Realization of Two-Dimensional Boron Sheets. Nat. Chem. 2016, 8,
563–568.
(32) Li, L.; Zhang, H.; Cheng, X. The High Hydrogen Storage Capacities of Li-Decorated
Borophene. Comput. Mater. Sci. 2017, 137 (Supplement C), 119–124.
(33) Rao, D.; Zhang, L.; Meng, Z.; Zhang, X.; Wang, Y.; Qiao, G.; Shen, X.; Xia, H.; Liu, J.;
Lu, R. Ultrahigh Energy Storage and Ultrafast Ion Diffusion in Borophene-Based Anodes
for Rechargeable Metal Ion Batteries. J. Mater. Chem. A 2017, 5 (5), 2328–2338.
(34) Chen, H.; Zhang, W.; Tang, X.-Q.; Ding, Y.-H.; Yin, J.-R.; Jiang, Y.; Zhang, P.; Jin, H.
First Principles Study of P-Doped Borophene as Anode Materials for Lithium Ion
Batteries. Appl. Surf. Sci. 2018, 427 (Part B), 198–205.
(35) Jena, N. K.; Araujo, R. B.; Shukla, V.; Ahuja, R. Borophane as a Benchmate of Graphene:
A Potential 2D Material for Anode of Li and Na-Ion Batteries. ACS Appl. Mater.
Interfaces 2017, 9 (19), 16148–16158.
(36) Mortazavi, B.; Rahaman, O.; Ahzi, S.; Rabczuk, T. Flat Borophene Films as Anode
13
Materials for Mg, Na or Li-Ion Batteries with Ultra High Capacities: A First-Principles
Study. Appl. Mater. Today 2017, 8, 60–67.
(37) Izadi Vishkayi, S.; Bagheri Tagani, M. Current-Voltage Characteristics of Borophene and
Borophane Sheets. Phys. Chem. Chem. Phys. 2017, 19 (32), 21461–21466.
(38) Xiang, P.; Chen, X.; Zhang, W.; Li, J.; Xiao, B.; Li, L.; Deng, K. Metallic Borophene
Polytypes as Lightweight Anode Materials for Non-Lithium-Ion Batteries. Phys. Chem.
Chem. Phys. 2017, 19 (36), 24945–24954.
(39) Zhang, L.; Liang, P.; Shu, H.; Man, X.; Li, F.; Huang, J.; Dong, Q.; Chao, D. Borophene
as Efficient Sulfur Hosts for Lithium–Sulfur Batteries: Suppressing Shuttle Effect and
Improving Conductivity. J. Phys. Chem. C 2017, 121 (29), 15549–15555.
(40) Liu, J.; Wang, S.; Sun, Q. All-Carbon-Based Porous Topological Semimetal for Li-Ion
Battery Anode Material. Proc. Natl. Acad. Sci. 2017, 114 (4), 651–656.
(41) Kulish, V. V. Surface Reactivity and Vacancy Defects in Single-Layer Borophene
Polymorphs. Phys. Chem. Chem. Phys. 2017, 19 (18), 11273–11281.
(42) Khanifaev, J.; Pekoz, R.; Konuk, M.; Durgun, E. The Interaction of Halogen Atoms and
Molecules with Borophene. Phys. Chem. Chem. Phys. 2017.
(43) Penev, E. S.; Kutana, A.; Yakobson, B. I. Can Two-Dimensional Boron Superconduct?
Nano Lett. 2016, 16 (4), 2522–2526.
(44) Sadeghzadeh, S. Borophene Sheets with in-Plane Chain-like Boundaries; a Reactive
Molecular Dynamics Study. Comput. Mater. Sci. 2018, 143, 1–14.
(45) Izadi Vishkayi, S.; Bagheri Tagani, M. Edge-Dependent Electronic and Magnetic
Characteristics of Freestanding β 12-Borophene Nanoribbons. Nano-Micro
Lett. 2017, 10 (1), 14.
(46) Sadeghzadeh, S. The Creation of Racks and Nanopores Creation in Various Allotropes of
Boron due to the Mechanical Loads. Superlattices Microstruct. 2017, 111 (Supplement
C), 1145–1161.
(47) Mortazavi, B.; Rahaman, O.; Dianat, A.; Rabczuk, T. Mechanical Responses of
Borophene Sheets: A First-Principles Study. Phys. Chem. Chem. Phys. 2016, 18 (39),
27405–27413.
(48) Adamska, L.; Sharifzadeh, S. Fine-Tuning the Optoelectronic Properties of Freestanding
Borophene by Strain. ACS Omega 2017, 2 (11), 8290–8299.
(49) Nishino, H.; Fujita, T.; Cuong, N. T.; Tominaka, S.; Miyauchi, M.; Iimura, S.; Hirata, A.;
Umezawa, N.; Okada, S.; Nishibori, E.; Fujino, A.; Fujimori, T.; Ito, S.; Nakamura, J.;
Hosono, H.; Kondo, T. Formation and Characterization of Hydrogen Boride Sheets
Derived from MgB2 by Cation Exchange. J. Am. Chem. Soc. 2017, 139 (39), 13761–
13769.
(50) Jiao, Y.; Ma, F.; Bell, J.; Bilic, A.; Du, A. Two-Dimensional Boron Hydride Sheets: High
Stability, Massless Dirac Fermions, and Excellent Mechanical Properties. Angew. Chemie
2016, 128 (35), 10448–10451.
(51) Zhang, X.; Hu, J.; Cheng, Y.; Yang, H. Y.; Yao, Y.; Yang, S. A. Borophene as an
Extremely High Capacity Electrode Material for Li-Ion and Na-Ion Batteries. Nanoscale
2016, 8 (33), 15340–15347.
(52) Guo, Y.-G.; Hu, J.-S.; Wan, L.-J. Nanostructured Materials for Electrochemical Energy
Conversion and Storage Devices. Adv. Mater. 2008, 20 (15), 2878–2887.
(53) Kresse, G.; Furthmüller, J. Efficient Iterative Schemes for Ab Initio Total-Energy
Calculations Using a Plane-Wave Basis Set. Phys. Rev. B 1996, 54 (16), 11169–11186.
14
(54) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made
Simple. Phys. Rev. Lett. 1996, 77 (18), 3865–3868.
(55) Kresse, G.; Joubert, D. From Ultrasoft Pseudopotentials to the Projector Augmented -
Wave Method. Phys. Rev. B 1999, 59, 1758.
(56) Bučko, T.; Hafner, J.; Lebègue, S.; Ángyán, J. G. Improved Description of the Structure
of Molecular and Layered Crystals: Ab Initio DFT Calculations with van Der Waals
Corrections. J. Phys. Chem. A 2010, 114 (43), 11814–11824.
(57) Wang, Z.; Selbach, S. M.; Grande, T. Van Der Waals Density Functional Study of the
Energetics of Alkali Metal Intercalation in Graphite. RSC Adv. 2014, 4 (8), 4069–4079.
(58) Grimme, S. Semiempirical GGA-Type Density Functional Constructed with a Long-
Range Dispersion Correction. J. Comput. Chem. 2006, 27 (15), 1787–1799.
(59) Tang, W.; Sanville, E.; Henkelman, G. A Grid-Based Bader Analysis Algorithm without
Lattice Bias. J. Phys. Condens. Matter 2009, 21 (8), 84204.
(60) Aydinol, M. K.; Kohan, A. F.; Ceder, G.; Cho, K.; Joannopoulos, J. Ab Initio Study of
Lithium Intercalation in Metal Oxides and Metal Dichalcogenides. Phys. Rev. B 1997, 56
(3), 1354–1365.
(61) Koudriachova, M. V.; Harrison, N. M.; De Leeuw, S. W. Diffusion of Li-Ions in Rutile.
An Ab Initio Study. In Solid State Ionics; 2003; Vol. 157, pp 35–38.
(62) Tarascon, J. M.; Armand, M. Issues and Challenges Facing Rechargeable Lithium
Batteries. Nature 2001, 414 (6861), 359–367.
(63) Yang, Z.; Choi, D.; Kerisit, S.; Rosso, K. M.; Wang, D.; Zhang, J.; Graff, G.; Liu, J.
Nanostructures and Lithium Electrochemical Reactivity of Lithium Titanites and Titanium
Oxides: A Review. Journal of Power Sources. 2009, pp 588–598.
(64) Mortazavi, B.; Dianat, A.; Cuniberti, G.; Rabczuk, T. Application of Silicene, Germanene
and Stanene for Na or Li Ion Storage: A Theoretical Investigation. Electrochim. Acta
2016, 213, 865–870.
(65) Zhao, S.; Kang, W.; Xue, J. The Potential Application of Phosphorene as an Anode
Material in Li-Ion Batteries. J. Mater. Chem. A 2014, 2 (44), 19046–19052.
(66) Mortazavi, B.; Dianat, A.; Rahaman, O.; Cuniberti, G.; Rabczuk, T. Borophene as an
Anode Material for Ca, Mg, Na or Li Ion Storage: A First-Principle Study. J. Power
Sources 2016, 329, 456–461.
(67) Uthaisar, C.; Barone, V. Edge Effects on the Characteristics of Li Diffusion in Graphene.
Nano Lett. 2010, 10 (8), 2838–2842.
(68) Makaremi, M.; Mortazavi, B.; Singh, C. V. Carbon Ene-Yne Graphyne Monolayer as an
Outstanding Anode Material for Li/Na Ion Batteries. Appl. Mater. Today 2018, 10, 115–
121.
(69) Tritsaris, G. A.; Kaxiras, E.; Meng, S.; Wang, E. Adsorption and Diffusion of Lithium on
Layered Silicon for Li-Ion Storage. Nano Lett. 2013, 13 (5), 2258–2263.
(70) Li, Y.; Wu, D.; Zhou, Z.; Cabrera, C. R.; Chen, Z. Enhanced Li Adsorption and Diffusion
on MoS 2 Zigzag Nanoribbons by Edge Effects: A Computational Study. J. Phys. Chem.
Lett. 2012, 3 (16), 2221–2227.
(71) Er, D.; Li, J.; Naguib, M.; Gogotsi, Y.; Shenoy, V. B. Ti₃C₂ MXene as a High Capacity
Electrode Material for Metal (Li, Na, K, Ca) Ion Batteries. ACS Appl. Mater. Interfaces
2014, 6 (14), 11173–11179.
(72) Rong, Z.; Malik, R.; Canepa, P.; Sai Gautam, G.; Liu, M.; Jain, A.; Persson, K.; Ceder, G.
Materials Design Rules for Multivalent Ion Mobility in Intercalation Structures. Chem.
15
(73) Olson, C. L.; Nelson, J.; Islam, M. S. Defect Chemistry, Surface Structures, and Lithium
Insertion in Anatase TiO2. J. Phys. Chem. B 2006, 110 (20), 9995–10001.
Mater. 2015, 27 (17), 6016–6021.
(74) Persson, K.; Hinuma, Y.; Meng, Y. S.; Van der Ven, A.; Ceder, G. Thermodynamic and
Kinetic Properties of the Li-Graphite System from First-Principles Calculations. Phys.
Rev. B 2010, 82 (12), 125416.
16
Figure 1. a) Hydrogen boride (HB) structural configuration including the supercell (black line)
and the unitcell (red line) with lattice constants a = 3.02 Å and b = 5.29 Å. The contours
illustrate electron localization function (ELF), which has a value between 0 and 1,
where 1 corresponds to the perfect localization. b) Different possible adatom adsorption
sites on the HB surface. Green, white, and red balls represent boron, hydrogen, and adatom,
respectively.
17
Figure 2. Binding charge transfer due to the adsorption of a) Li and b) Na to the
hexagonal site of the monolayer. Color coding involves blue for charge sufficient
and red for the charge deficient regions; respectively, (isosurface value is 0.001
e/Å3).
18
Figure 3. Density of states (DOS) for the pristine HB monolayer, and monolayers interacting with
different amount of adatom inserion. The black dashed line represents the Fermi-level.
19
Figure 4. Adsorption of adatoms by the HB surface. a) Binding energy of adatoms with respect
to the coverage ratio of x. The adsorption of b) Li and c) Na adatoms to the anode surface at x
< 0.5. Color coding involves green and yellow for Li and Na, respectively.
20
Figure 5. Storage capacity of the HB nanosheet. a) Average open-circuit voltage the anode with
respect to the capacity. The nanosheet-adatoms configuration after reaching the optimal
capacity (x > 0.5) due to the insertion of b) Li and c) Na adatoms. Color coding is identical to
Figure 4.
21
Figure 6. Intercalation of single adatoms into the HB nanosheet through different pathways
(Path 1, Path 2, and Path 3). a) Li diffusion and b) Na diffusion; Nudged-elastic band (NEB)
energy curvatures for c) Li and d) Na adatoms. Color coding of atoms is similar to Figure 4.
22
Supporting Information
2D Hydrogenated Graphene-like Borophene as a High Capacity Anode
Material for Improved Li/Na Ion Batteries: A First Principles Study
Meysam Makaremi,1 Bohayra Mortazavi,2 and Chandra Veer Singh*,1,3
1Department of Materials Science and Engineering, University of Toronto, 184 College Street,
2Institute of Structural Mechanics, Bauhaus-Universität Weimar, Marienstr. 15,
Suite 140, Toronto, ON M5S 3E4, Canada.
3Department of Mechanical and Industrial Engineering, University of Toronto, 5 King's College
D-99423 Weimar, Germany.
Road, Toronto M5S 3G8, Canada.
1) Hydrogen Boride Structural Details
Table S1: Lattice vectors of the hydrogen boride unitcell.
z [Å]
0.000
0.000
20.000
z [Å]
10.000
10.000
10.000
10.000
9.090
10.900
9.090
10.900
Vector
1
2
3
x [Å]
3.017
0.000
0.000
y [Å]
0.000
5.291
0.000
Table S2: Atomic positions of the hydrogen boride unitcell.
x [Å]
0.000
1.508
1.508
0.000
0.000
1.508
1.508
0.000
Atom
B
B
B
B
H
H
H
H
y [Å]
0.000
0.881
2.645
3.527
4.409
1.763
1.763
4.409
23
2) Li Adsorption on Hydrogen Boride
Figure S1. Adsorption of Li adatoms on HB nanosheet. Color coding is identical to Figure 4.
24
3) Na Adsorption on Hydrogen Boride
Figure S2. Adsorption of Na adatoms on HB nanosheet. Color coding is identical to Figure 4.
25
4) Electronic Charge Density of Before and After Optimal Storage Capacity for Na
Insertion
Figure S3. Differential Charge density of the HB nanosheet interacting with Na
adatoms. The charge difference configuration a) before and b) after reaching the
optimal capacity due to the insertion of Na adatoms. Color coding is identical to
Figures 2 and 4.
26
5) Thermal Stability Analysis by AIMD Simulations
Figure S4. Thermal stability analysis. The variation of the temperature with
respect to the simulation time for the HB monolayer interacting with Li
(green)/Na (yellow) adatoms at the optimal capacity.
27
|
1811.09551 | 1 | 1811 | 2018-10-26T14:23:37 | Dynamic asymptotic homogenization for periodic viscoelastic materials | [
"physics.app-ph"
] | A non-local dynamic homogenization technique for the analysis of a viscoelastic heterogeneous material which displays a periodic microstructure is herein proposed. The asymptotic expansion of the micro-displacement field in the transformed Laplace domain allows obtaining, from the expression of the micro-scale field equations, a set of recursive differential problems defined over the periodic unit cell. Consequently, the cell problems are derived in terms of perturbation functions depending on the geometrical and physical-mechanical properties of the material and its microstructural heterogeneities. A down-scaling relation is formulated in a consistent form, which correlates the microscopic to the macroscopic transformed displacement field and its gradients through the perturbation functions. Average field equations of infinite order are determined by substituting the down-scale relation into the micro-field equation. Based on a variational approach, the macroscopic field equations of a non-local continuum is delivered and the local and non-local overall constitutive and inertial tensors of the homogenized continuum are determined. The problem of wave propagation in case of a bi-phase layered material with orthotropic phases and axis of orthotropy parallel to the direction of layers is investigated as an example. In such a case, the local and non-local overall constitutive and inertial tensors are determined analytically and the dispersion curves obtained from the non-local homogenized model are analysed. | physics.app-ph | physics |
Dynamic asymptotic homogenization for periodic viscoelastic
materials
Rosaria Del Toro1, Andrea Bacigalupo1, Marco Paggi1
1 IMT School for Advanced Studis Lucca, Piazza S. Francesco 19, 55100 Lucca, Italy
Abstract
A non-local dynamic homogenization technique for the analysis of a viscoelastic heterogeneous ma-
terial which displays a periodic microstructure is herein proposed. The asymptotic expansion of the
micro-displacement field in the transformed Laplace domain allows obtaining, from the expression of
the micro-scale field equations, a set of recursive differential problems defined over the periodic unit
cell. Consequently, the cell problems are derived in terms of perturbation functions depending on the
geometrical and physical-mechanical properties of the material and its microstructural heterogeneities.
A down-scaling relation is formulated in a consistent form, which correlates the microscopic to the macro-
scopic transformed displacement field and its gradients through the perturbation functions.
Average field equations of infinite order are determined by substituting the down-scale relation into the
micro-field equation. Based on a variational approach, the macroscopic field equations of a non-local
continuum is delivered and the local and non-local overall constitutive and inertial tensors of the homog-
enized continuum are determined. The problem of wave propagation in case of a bi-phase layered material
with orthotropic phases and axis of orthotropy parallel to the direction of layers is investigated as an
example. In such a case, the local and non-local overall constitutive and inertial tensors are determined
analytically and the dispersion curves obtained from the non-local homogenized model are analysed.
1 Introduction
Over the last few years, the technological progress led to a fast development of composite materials. Such
materials play a crucial role for various applications in civil, naval, aerospace and mechanical engineering,
since they boast remarkable mechanical and physico-chemical properties such as high strength, corrosion
and thermal resistance, enhanced durability, light weight and ease handling, Gibson [2011]. Among them,
the class of polymer matrix composites is very promising since they can achieve performances superior to
metals with a reduced weight, Wang et al. [2011]. The matrix is usually a resin (epoxy or polyester) with
high toughness, reinforced by fibers (glass, aramid, boron, etc.), which have very high strength. The com-
bination of the two materials is very effective: the matrix diffuses the load among the fibers and protects
them from abrasion, fracture, and damage. At the same time, the reinforcing fibres increase the overall
strength and stiffness of the composite. Similarly, in laminates, the polymeric matrix is used to bond other
materials together and increase the toughness of the composite, see e.g. photovoltaic modules, Paggi et al.
[2016]. To reduce the cost of synthetic fiber-reinforced composites and produce environmentally sustainable
materials, bio-fibre-reinforced polymer composites are very promising and are becoming increasing popular
in emergent countries. The variant of hybrid composites, where synthetic and natural fiber reinforcements
are mixed together offer also a possible trade-off solution, Dhakal et al. [2018].
If the heterogeneities of the reinforcement are sufficiently regular and their size is much smaller than the
dimension of the component, then the material is said to be a composite with periodic or quasi-periodic
microstructure. The constitutive response of polymeric composites and their variants is that of viscoelastic
materials, which exhibit creep and stress relaxation phenomena.
An intense knowledge of the behaviour of viscoelastic materials allows manufacturing devices, which can be
applied to a wide range of fields, including biomedical, industrial, defence and construction. Krushynska
et al. [2016] studied the wave dispersion properties and attenuation capability of dissipative solid acoustic
metamaterials with local resonators having subwavelength band gaps. The characterization and simulation
of mechanical problems involving such materials is very expensive, due to the presence of heterogeneities.
1
Therefore, based on the premises above, the theory of homogenization may represent an excellent method-
ology to recognize and model the effects of the microscopic behaviour on the overall properties of materials.
Such a theory allows replacing a heterogeneous material with an equivalent homogenous one, which can be
modelled through either a first order (Cauchy) or a non-local continuum. Generally, three main classes of
homogenization techniques are possible: the asymptotic techniques (Bensoussan et al., 1978; Bakhvalov and
Panasenko, 1984; Gambin and Kroner, 1989; Allaire, 1992; Meguid and Kalamkarov, 1994; Boutin, 1996;
Fish and Chen, 2001; Andrianov et al., 2008; Panasenko, 2009; T.H. Tran and Bonnet, 2012; Bacigalupo,
2014;), the variational-asymptotic techniques (Smyshlyaev and Cherednichenko, 2000; Smyshlyaev, 2009;
Bacigalupo and Gambarotta, 2014b; Bacigalupo et al., 2014) and many identification approaches, involv-
ing the analytical (Bigoni and Drugan, 2007; Bacca et al., 2013a; Bacca et al., 2013b; Bacca et al., 2013c;
Bacigalupo and Gambarotta, 2013; Bacigalupo et al., 2017) and the computational techniques (Forest and
Sab, 1998; Ostoja-Starzewski et al., 1999; Kouznetsova et al., 2002; Forest, 2002; Feyel, 2003; Kouznetsova
et al., 2004; Kaczmarczyk et al., 2008; Yuan et al., 2008; Bacigalupo and Gambarotta, 2010;De Bellis and
Addessi, 2011; Forest and Trinh, 2011; Addessi et al., 2013; Zah and Miehe, 2013; Trovalusci et al., 2015).
Such techniques have been expanded to the multi-field case, such as termomechanics Aboudi et al., 2001;
Kanout´e et al., 2009; Zhang et al., 2007 and thermo-diffusive, Bacigalupo et al., 2016a; Bacigalupo et al.,
2016b and thermo-piezoelectricity phenomena, Fantoni et al., 2017.
In case of viscoelastic materials with periodic microstructure, which are the object of the present article,
there are still a few contributions devoted to homogenization techniques applied to this paricular class of
composites. Specifically, the computational techniques are proposed in the works of Ohno et al., 2000; Haase-
mann and Ulbricht, 2010; Tran et al., 2011 and Q. Chen and Geng, 2017 and the asymptotic techniques are
analysed in Yi et al., 1998 and Hui and Oskay, 2013.
In the context of the computational homogenization techniques, Ohno et al. [2000] dealt with an homog-
enization model for elastic-viscoplastic periodic materials, but without taking into account an asymptotic
expansion of the field variables. Their method allows determining the macroscopic and the microscopic stress
and strain states in nonlinear time-dependent periodic materials and it encompasses any problem where the
history of the macro-strain and the macro-stress depends upon the time. Haasemann and Ulbricht [2010]
considered a microstructure where all the constituents are linear viscoelastic. The constitutive laws at the
microscale were converted into a Laplace-Carson domain, where the constitutive equations have a quite sim-
ilar form to those of a linear elastic material and then a homogenization approach based on the Hill-Mandel
condition was exploited. The Laplace-Carson transform associated with the application of a FE-method
enables the computation of the relaxation tensor in the Laplace domain and the inverse Laplace-Carson
transformation provides the material properties in the time domain. Tran et al. [2011] presented a computa-
tional homogenization method to determine the response of a linear viscoelastic heterogeneous material. The
components of the relaxation tensor, which appear in the constitutive law at the macro-scale, are numerically
determined in the time domain, without involving the Laplace transform.
Although the employment of computational approaches is more and more extensive thanks to up-to-date
computer facilities, they have a high computational cost, they cannot challenge dynamic problems and they
are not able to provide higher-order approximations of the homogenized constitutive tensors.
Concerning with the asymptotic techniques applied to viscoelastic materials, Hui and Oskay [2013] proposed
a non-local homogenization method with multiple length scales for detecting wave propagation in viscoelas-
tic composite materials, by proceeding with an asymptotic expansion of the governing system of equations
defined in the time domain, then recast into the Laplace domain. The higher-order terms, derived from this
approach, identify the micro heterogeneities producing the wave dispersions and predicting the creation of
bandgaps. Neverthless, such a method cannot provide an average field equation of infinite order, containing
local and non-local higher-order tensor components.
Motivated by the state-of-the-art literature on homogenization, the present study proposes a dynamic
variational-asymptotic homogenization technique for the analysis of a viscoelastic material with periodic
microstructure modelled with a non-local continuum, based on the asymptotic and variational methods
(Smyshlyaev and Cherednichenko [2000] and Bacigalupo [2014]) and on the works related to the variational
principles of linear viscoelasticity (Leitman [1966], Fabrizio and Morro [1992] and ).
The field equation at the micro-scale, which describes the heterogeneous viscoelastic domain, is determined in
the time domain and it is converted into the Laplace domain, with the help of the two-sided Laplace trans-
form. The micro-displacement field is expressed as an asymptotic expansion in the transformed Laplace
space and its replacement into the field equation at the micro-scale enables to produce a sequence of re-
2
cursive differential problems defined over the periodic unit cell. Then solvability conditions are imposed to
such nonhomogeneous recursive cell problems to determine the down-scaling relation, linking the microscopic
transformed displacement field to the macroscopic one and its gradients through the perturbation functions.
Such functions rely on the geometrical and physical-mechanical properties of the material and measure the
microstructural heterogeneities. Average field equations of infinite order are determined by substituting the
down-scale relation into the micro-field equations. Its formal solution is provided with the help of an asymp-
totic expansion of the transformed macro-displacement and, by considering only the terms at the zeroth
order, the field equations related to the equivalent viscoelastic Cauchy continuum are retrieved.
Section 2 deals with the description of the field equations in the time domain and in the Laplace domain
at the microscale. Section 3 shows the recursive differential problems and their solutions and Section 4
presents the cell problems and the related perturbation functions. Section 5 defines the down-scaling re-
lation, the up-scaling relation and the average field equations of infinite order. In Section 6, by means of
a variational approach, the overall constitutive tensors and the overall inertial tensor related to the ho-
mogenized continuum are derived in the Laplace domain for the class of periodic viscoelastic materials,
after introducing the energy-like functional in the Laplace domain (Fabrizio and Morro [1992]). Moreover,
the Euler-Lagrangian differential equation at the macro-scale is determined, expressed in terms of the trans-
formed macro-displacement and its gradients up to the fourth order. In Section 7, the variational-asymptotic
homogenization technique is applied to a bi-phase layered material with isotropic phases subject to periodic
body forces. To verify the reliability of the proposed homogenization procedure, the solution of the homog-
enized problem is compared with the one obtained from the heterogeneous problem and a good agreement
between the models is obtained. Finally, the problem of wave propagation and the related dispersion curves
is studied. Concluding remarks complete the article.
2 Problem setting and field equation in the Laplace domain
Let Ω be a three-dimensional viscoelastic heterogeneous material which displays a periodic microstructure.
A generic point of the material is identified by the position vector x = x1e1 + x2e2 + x3e3 related to a system
of coordinates with origin at point O and orthogonal base {e1, e2, e3}. Let A = [0, ε] × [0, δε] × [0, ε] be a
periodic cell with characteristic size ε. A is described by three orthogonal periodicity vectors v1, v2 and v3
defined as v1 = d1e1 = εe1, v2 = d2e2 = δεe2 and v3 = d3e3 = εe3. The material domain is set up by the
repetition of the cell A in accordance with the directions of v1, v2 and v3, see Fig. 1. Since the material
Figure 1: Heterogeneous and homogeneous 3D domain Ω with periodic cell A and the corresponding nondi-
mensional cell Q.
3
(cid:90) t
is A-periodic, the micro relaxation tensor Gm(x, t) = Gm
ijhkei ⊗ ej ⊗ eh ⊗ ek, which depends on time and
accounts for the viscoelastic effects, and the material density ρm(x) comply with the following conditions:
Gm(x + vi, t) = Gm(x, t),
ρm(x + vi, t) = ρm(x, t),
i = 1, 2, 3 ∀x ∈ A,
i = 1, 2, 3 ∀x ∈ A.
(1)
(2)
The micro stress σ(x, t) constitutive relation, which models the viscoelastic elements of the heterogeneous
material, is expressed in terms of the hereditary integral, Christensen [2012]:
Gm(x, t − τ ) ε(x, τ )dτ,
σ(x, t) =
(3)
where the superscript m refers to the microscale and ε(x, t) = εijei⊗ ej is the micro-strain tensor. Moreover
t denotes the time coordinate and the superimposed dot indicates time derivative. The material undergoes
2 (∇u(x, t) + ∇T u(x, t)), where
small displacements and so the micro-strain tensor is defined as ε(x, t) = 1
∇u is the gradient of the micro-displacement u(x, t). In the time domain, the deformation response of the
material under dynamic loading is expressed by the momentum balance equation:
−∞
∇ · σ(x, t) + b(x, t) = ρm(x)u(x, t),
(4)
where u(x, t) is the micro-displacement field and b(x, t) are the body forces. In the derivation of the theory,
the heterogeneous material is supposed to be subjected to a system of L-periodic body forces b(x, t), with
zero mean values over L = [0, L] × [0, δL]. The structural (or macroscopic) length L is assumed to be
much greater than the microstructural length ε, i.e. L>> ε, to allow the scales separation condition and
so L is considered as an actual representative portion of the material. Let Q = [0, 1] × [0, δ] × [0, 1] be the
nondimensional cell reproducing the periodic microstructure. Q is determined by rescaling the size of the
periodic cell A for the characteristic length ε. Accordingly, two variables are introduced to differentiate the
two scales, namely the macroscopic (or slow) one, x ∈ A, which measures the slow fluctuations, and the
ε ∈ Q, which measures the fast propagation of the signal. Thanks to cell
microscopic (or fast) variable, ξ = x
Q, the properties (1) and (2) may be rewritten in terms of the microscopic variable ξ and so Gm and ρm
are assumed to be Q-periodic and defined on Q as
(5)
Bearing in mind the definition of the strain tensor ε, the minor simmetry of the relaxation tensor Gm is
applied to ε in the integral (3) and the substitution of Eq. (3) into Eq. (4) yields to
ρm(x, t) = ρm(x/ε, t).
Gm(x, t) = Gm(ξ = x/ε, t),
∇ ·(cid:104)(cid:90) t
−∞
Gm(cid:16) x
ε
, t − τ
(cid:104)(cid:104)(cid:90) t
(cid:105)
, t − τ
(cid:17)∇ u(x, τ )dτ
Gm(cid:16) x
(cid:16)
(cid:90) +∞
u = u
x
ε
x,
L(cid:16) ∂nf (t)
(cid:17)
∂tn
4
+ b(x, t) = ρm(x)u(x, t).
(6)
Denoting with [[f ]] = f i(Σ)−f j(Σ) the jump of the function values f at the interface Σ between two different
phases i and j in the periodic cell A, the following fully-bonded interface conditions hold
[[u(x)]]x∈Σ = 0,
(7)
where n represents the outward normal to the interface Σ. Since Gm and ρm are Q-periodic and the body
forces are L-periodic, the micro-displacement depends on both the slow variable x and the fast one ξ and
can be expressed as
= 0,
−∞
ε
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)x∈Σ
(cid:17)∇ u(x, τ )dτ · n
(cid:17)
, t
.
The two-sided Laplace transform of an arbitrary, real valued, time varying function, f ∈ R, is defined as,
Paley and Wiener [1934],
(8)
where, the Laplace argument, s, and the Laplace transform, f , are complex valued (i.e. f : C → C ). The
derivative rule for the Laplace transform is provided by
−∞
L(f (t)) = f (s) =
f (t)e−stdt,
s ∈ C,
= sn f (s),
(9)
and the convolution rule of f1 and f2 is given as
L(f1(t) ∗ f2(t)) = L(f1(t))L(f2(t)),
(10)
or, in other terms:
∇ ·(cid:104)L(cid:16)Gm(cid:16) x
∇ ·(cid:104) Gm(cid:16) x
, t
Equation (6) governing the periodic viscoelastic material in the time domain will be recast in the Laplace
domain employing the Laplace transform (8), the convolution rule
(9).
Therefore, in the Laplace domain, it results
(10) and the derivative rule
(cid:16) x
(cid:17)
forces transformed in the Laplace domain. In addition, it is convenient to consider Cm(cid:16) x
(12)
where u and ∇u represent the micro-displacement field and the gradient of the micro-displacement field
converted in the Laplace domain. Moreover, Gm is the micro-relaxation tensor and b(x, s) are the body
.
+ L(b(x, t)) = ρm(cid:16) x
(cid:17)(cid:17)(cid:105)
(cid:17)
+ b(x, s) = ρm(cid:16) x
(cid:17)(cid:105)
(cid:17)(cid:17)L(cid:16)∇ u
(cid:16) x
(cid:16) x
(cid:17)
= s Gm(cid:16) x
(cid:17)L(cid:16)
(cid:16) x
,
(11)
(cid:17)(cid:17)
s∇u
, x, s
,
(cid:17)
(cid:17)
, s
ε
, x, s
, x, t
, x, t
s2 u
u
ε
ε
ε
ε
ε
ε
ε
ε , s
ε , s
The governing equation of the periodic viscoelastic material defined in the Laplace domain is
(13)
Denoting with [[f ]] = f i(Σ)− f j(Σ) the jump of the function values f at the interface Σ between two phases
i and j in the periodic cell A, the following continuity conditions hold for a perfectly bonded interface
∇ · ( Cm∇u) + b = ρms2 u.
[[u(x, s)]]x∈Σ = 0,
(cid:104)(cid:104)(cid:16) Cm(cid:16) x
(cid:16)
(cid:17)∇u
, s
ε
x,
x
ε
, s
(cid:17)(cid:17) · n
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)x∈Σ
= 0,
(14)
where n represents the outward normal to the interface Σ. The solution of Eq. (13) is too expensive from
both a numerical and an analytical point of view, because the coefficients are Q-periodic. In order to cope
with such a drawback, it is convenient to employ a non-local asymptotic homogeneization technique to turn
the heterogeneous material into an equivalent homogeneous one. Such a procedure generates equations,
equivalent to (13), whose coefficients are not affected by oscillations and their solutions are close to those of
the original equation. Moreover, the computational cost to solve (13) significantly reduces.
In the equivalent homogenized material, by considering a reference system {O, e1, e2}, the macro-displacement
transformed in the Laplace domain is denoted as U (x) = Uiei, with respect to a point x, and the transformed
displacement gradiend is defined as ∇ U (x) = ∂ Ui
e1 ⊗ e2.
∂xj
3 Asymptotic expansion of the microscopic displacement
Based on the asymptotic approach developed in Bakhvalov and Panasenko [1984], Smyshlyaev and Chered-
nichenko [2000], Bacigalupo and Gambarotta [2014a], the micro-displacement u is expressed as an asymptotic
expansion in terms of the parameter ε that separates the slow x variable from the fast one ξ = x
ε ,
(cid:16)
(cid:17)
(cid:16)
(cid:17)
(cid:16)
(cid:17)
uh
x,
, t
=
εlu(l)
h = u(0)
h
x,
x
ε
, t
+ εu(1)
h
x,
x
ε
, t
+ ε2u(2)
h
x,
x
ε
, t
+ O(ε3),
(15)
The Laplace transform (8) is applied to Eq. (15) and leads to
uh
x,
x
ε
, t
εl u(l)
h = u(0)
h
x,
x
ε
, s
+ εu(1)
h
x,
x
ε
, s
(cid:16)
(cid:17)
(cid:16)
(cid:17)
(cid:16)
x,
x
ε
(cid:17)
, s
+ ε2 u(2)
h
+ O(ε3),
(16)
(cid:16)
L(cid:16)
x
ε
(cid:16)
(cid:16)
(cid:17)
+∞(cid:88)
l=0
(cid:17)(cid:17)
+∞(cid:88)
l=0
=
(cid:17)
which is equivalent to the asymptotic expansion of the micro-displacement performed in the time domain.
Let us consider the formula
D
Dxk
u
x, ξ =
x
ε
=
+
∂ uh(x, ξ)
∂ξk
∂ξk
∂xk
=
∂xk
uh(x, ξ) +
1
ε
uh,k
,
(17)
(cid:104) ∂ uh(x, ξ)
∂xk
(cid:105)(cid:12)(cid:12)(cid:12)ξ= x
ε
(cid:104) ∂
(cid:105)(cid:12)(cid:12)(cid:12)ξ= x
ε
5
which introduces the macroscopic derivative
Laplace domain, and let us apply it to the asymptotic expansion (16), leading to:
uh and the microscopic derivative uh,k in the transformed
∂xk
∂
(cid:16)
(cid:17)
(cid:104) ∂ u(0)
h
∂xk
D
Dxk
u
x, ξ =
x
ε
=
+ ε
∂ u(1)
h
∂xk
+ ε2 ∂ u(2)
h
∂xk
+ ...
+
1
ε
(cid:105)
(cid:104)
h,k + εu(1)
u0
h,k + ε2 u(2)
h,k + ...
.
(18)
(cid:105)(cid:12)(cid:12)(cid:12)ξ= x
ε
The asymptotic technique searches for the solution of Eq. (13) as a decomposition in increasing powers
of the microscopic lenght ε. To this purpose, the replacement of the asymptotic expansion (16) into the
microscopic field equation (13) in the Laplace domain and the rearrangement of the terms with equal power
ε yield to the asymptotic field equation
ijhk
,j
h,k
ijhk u(0)
+ ε−1(cid:104)(cid:16) C m
ε−2(cid:16) C m
(cid:17)
(cid:17)(cid:17)
(cid:16) ∂ u(1)
+ ε0(cid:104)(cid:16) C m
(cid:104)(cid:16) C m
(cid:16) ∂ u(2)
(cid:17)(cid:17)
+ u(2)
h,k
h
∂xk
ijhk
+ u(3)
h,k
ijhk
+
ε
h
∂xk
,j
,j
h
∂xk
+ u(1)
h,k
(cid:16) ∂ u(0)
(cid:17)(cid:17)
(cid:16) ∂ u(0)
(cid:16) C m
(cid:16) C m
(cid:16) ∂ u(1)
∂
∂xj
h
∂xk
ijhk
+
ijhk
h
∂xk
∂
∂xj
,j
+ u(2)
h,k
+
∂
∂xj
(cid:17)(cid:105)
h,k
ijhk u(0)
(cid:16) C m
(cid:17)(cid:17)
(cid:17)(cid:17) − ρms2 u(1)
h
+ u(1)
h,k
+ bi − ρms2u(0)
h
(cid:105)
+
(cid:105)
(cid:105)(cid:12)(cid:12)(cid:12)ξ= x
ε
+ O(ε2)
= 0.
+
(19)
Interface conditions
(14) are rephrased with respect to the fast variable ξ since the micro-displacement
uh(x, ξ) is supposed to be Q−periodic with respect to ξ and smooth in the slow variable x. Indicating with
Σ1 the interface between two different phases in the unit cell Q and considering the asymptotic expansion
(16) of the micro-displacement, interface conditions read
u(2)
h
u(0)
h
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:104)(cid:104)
(cid:104)(cid:104)(cid:16) C m
(cid:104)(cid:104)(cid:16) C m
+ ε
1
ε
ijhk
+ ε
u(1)
h
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
+ ε2(cid:104)(cid:104)
(cid:104)(cid:104)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
+ ε0(cid:104)(cid:104)(cid:16) C m
(cid:17)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:17)(cid:17)
(cid:16) ∂ u(1)
+ u(2)
h,k
h
∂ xk
nj
nj
ijhk u(0)
h,k
ijhk
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:16) ∂ u(0)
+ +ε2(cid:104)(cid:104)(cid:16) C m
h
∂ xk
+ u(1)
h,k
+ ... = 0
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:17)(cid:17)
(cid:16) ∂ u(2)
ijhk
h
∂ xk
+ u(3)
h,k
Recursive differential problems and their solutions
(20)
+
(cid:17)(cid:17)
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
+ ...+ = 0.
The asymptotic field equation (16) produces a set of recursive differential problems that determine sequen-
In particular, at the order ε−2, the differential problem, which stems from
tially the solutions u0, u1...
problem (19), is
with interface conditions
(cid:104)(cid:104)
u(0)
h
= f (0)
i
(x),
(cid:17)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
ijhk u(0)
h,k
nj
(21)
= 0.
The solvabiliy condition of this differential problem, in the class of Q−periodic solutions u(0)
f (0)
i
(x) = 0 and so the differential problem (21) develops in the form
h , implies that
(cid:16) C m
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0
,j
h,k
ijhk u(0)
(cid:17)
(cid:104)(cid:104)(cid:16) C m
(cid:16) C m
(cid:17)
ijhk u(0)
h,k
= 0.
,j
The solution results to be
u(0)
h (x, ξ, s) = U M
h (x, s),
6
(22)
(23)
where U M
Bearing in mind the solution (23), the differential problem from (19) at the order ε−1 is
h (x, s) is the transformed macroscopic displacement that does not depend on the microstructure.
(cid:16) C m
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:104)(cid:104)
u(1)
h
(cid:17)
(cid:104)(cid:104)(cid:16) C m
,j
(cid:16)
ijhk u(1)
h,k
+ C m
ijhk,j
= f (1)
i
(x),
∂ U M
h
∂xk
(cid:17)(cid:17)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
u(1)
h,k +
∂ U M
h
∂xk
= 0.
Similarly, the solvability condition in the class of Q−periodic functions ensures that
= 0
ijhk
nj
since U M
h,k = 0. Its interface conditions are
(24)
(25)
(26)
(27)
Q(·)dξ and Q = δ. Moreover the Q-periodicity of the components C m
ijhk and the
(cid:82)
where (cid:104)(·)(cid:105) = 1Q
divergence theorem entail f (1)
i
(x) = (cid:104) C m
f (1)
i
ijhk,j(cid:105) ∂ U (M )
h
∂xk
,
(cid:16) C m
(cid:17)
(x) = 0 and the differential problem
ijhk u(1)
h,k
+ C m
ijhk,j
,j
∂ U M
h
∂xk
= 0,
∀ ∂ U M
h
∂xk
has the following solution
u(1)
h (x, ξ, s) = N (1,0)
hpq1
(ξ)
∂ U M
p
∂xq1
,
where N (1,0)
hpq1
are supposed to have zero mean over the unit cell Q and so N (1,0)
is the perturbation function, which depends on the fast variable ξ. The perturbation functions
complies with the normalization condition
hpq1
N (1,0)
hpq1
(ξ)dξ = 0.
(28)
(cid:90)
Q
(cid:104)N (1,0)
hpq1
(cid:105) =
1
Q
ijhk
+ u(1)
h,k
i = f (2)
i
(x)
(29)
Moreover, the perturbation functions exclusively depend on the geometry and on the mechanical properties
of the microstructure. The differential problem at order ε0 is
(cid:16) C m
(cid:16) ∂ u(1)
with interface conditions(cid:104)(cid:104)
h
∂xk
ijhk
(cid:16) C m
(cid:17)
with interface conditions
u(2)
h
+ u(2)
h,k
(cid:17)(cid:17)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:17)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0,
,j
,j
= 0
ijhk
+
∂
∂xj
(cid:16) C m
(cid:104)(cid:104)(cid:16) C m
(cid:16) C m
(cid:16)
(cid:104)(cid:104)(cid:16) C m
+
ijhk
(cid:16)(cid:16) C m
(cid:104)(cid:104)
u(2)
h
(cid:16) ∂ u(0)
(cid:16) ∂ u(1)
h
∂xk
h
∂ xk
(cid:17)(cid:17) − ρms2 u(0)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:17)(cid:17)
nj
= 0.
+ u(2)
h,k
(cid:17)(cid:17) ∂2 U M
p
∂xq1 ∂xj
u(2)
h,k + N (1,0)
hpq1
∂2 U M
p
∂xq1∂xk
(cid:17)(cid:17)
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0.
Considering the solutions (23) and (27) of the differential problems at order ε−2 and ε−1, respectively, the
differential problem (29) is turned into
ijhk u(2)
h,k
+
,j
ijhkN (1,0)
hpq1
+ C m
ijhq1
ijhkN (1,0)
hpq1,k
− ρms2 U M
i = f (2)
i
(x),
(30)
Again, solvability condition of differential problem (30) in the class of Q−periodic functions and the diver-
gence theorem lead to
f (2)
i
(x) = (cid:104) C m
ijhq1
+ C m
ijhkN (1,0)
hpq1,k(cid:105) ∂2 U M
p
∂xq1 ∂xj
− (cid:104)ρm(cid:105)s2 U M
i
and consequentely the solution of the differential problem at the order ε0 is
u(2)
h (x, ξ, s) = N (2,0)
hpq1q2
∂2 U M
p
∂xq1∂xq2
+ N (2,2)
hp
s2U M
p ,
where N (2,2)
hp
is the perturbation function depending on the parameter s.
7
(31)
(32)
4 Cell problems and perturbation functions
In the Section 3, the solutions u(0)
formulate the cell problems, which are classified according to the even power of the parameter s.
h , ... have been established. Such solutions are employed to
h , u(1)
h , u(2)
Cell problems related to s0
The substitution of solution (27) into problem (26) leads to the following cell problem at the order ε−1
ijhkN (1,0)
hpq1,k
(cid:16) C m
(cid:104)(cid:104)(cid:16) C m
(cid:17)
(cid:16)
= 0
(cid:104)(cid:104)
N (1,0)
ipq1
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
+ C m
ijpq1,j = 0
,j
(33)
ijhk
N (1,0)
hpq1,k + δhpδkq1
nj
= 0,
(34)
(cid:17)(cid:17)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
with interface conditions derived in terms of the perturbation function N (1,0)
hpq1,k
where δhp and δkq1 are the Kronecker delta functions. Once the perturbation function N (1,0)
hpq1,k has been
determined, and thanks to equation (30) and its solution (32), the cell problem at the order ε0 is derived
and the symmetrized version with respect to indices q1 and q2 is
ijhkN (2,0)
hpq1q2,k
+ C m
iq2pq1
+
iq2hkN (1,0)
hpq1,k
+
(cid:16) C m
(cid:17)
(cid:104)(cid:16) C m
ikhq2
(cid:17)
,k
N (1,0)
hpq1
(cid:16) C m
+
(cid:17)
,j
,k
+
1
2
+ C m
(cid:17)
(cid:17)
(cid:16) C m
(cid:104)(cid:104)(cid:16) C m
(cid:17)
+
,j
(cid:88)
(cid:104)
1
w + 2
P∗(q)
(cid:17)(cid:105)
(cid:16) C m
(cid:104)(cid:104)
(cid:16) C m
ijhq2
(cid:105)
N (1,0)
hpq2
ikhq1
iq1pq2
iq1hkN (1,0)
hpq2,k
=
=
1
2
(cid:104) C m
iq2hq1
+
with interface conditions
iq2hkN (1,0)
hpq1,k
+ C m
iq1hq2
+
iq1hkN (1,0)
(cid:17)(cid:105),
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
hpq2,k
N (2,0)
ipq1q2
= 0,
ijhkN (2,0)
hpq1q2,k +
1
2
N (1,0)
hpq1
+ C m
ijhq1
N (1,0)
hpq2
(35)
(36)
(cid:17)(cid:17)
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0.
The solution of the cell problem (35) and (36) is the perturbation function N (2,0)
order εw with w ∈ Z and w ≥ 1 is
ipq1q2
. The cell problem at the
ijhkN (w+2,0)
hpq1...qw+2,k
( C m
ijhqw+2
N (w+1,0)
hpq1...qw+1
),j+
(cid:16) C m
(cid:16) C m
+
(cid:16) C m
(cid:88)
P∗(q)
=
1
w + 2
+ C m
iqw+2hjN (w+1,0)
hpq1...qw+1,j + C m
iqw+2hqw+1
N (w,0)
hpq1...qw
=
(cid:104) C m
iqw+2hjN (w+1,0)
hpq1...qw+1,j + C m
iqw+2hqw+1
N (w,0)
hpq1...qw
(cid:105),
and the corresponding interface conditions are
(cid:104)(cid:104)(cid:16) C m
ijhkN (w+2,0)
hpq1...qw+2,k +
1
w + 2
(cid:104)(cid:104)
N (w+2,0)
ipq1...qw+2
(cid:88)
P ∗(q)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0,
C m
ijhqw+2
N (w+1,0)
hpq1...qw+1
(cid:17)
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0,
(37)
(38)
where symbol P∗(q) denotes all the possible permutations of the multi-index q = q1, q2, ..., ql that does not
exhibit fixed indices (see Appendix B). The resolution of cell problem (37) allows to determining the form
of the perturbation function N (w+2,0)
.
ipq1...qw+2
8
Cell problems related to s2
The substitution of solution (32) into Eq. (30) generates the cell problem at the order ε0
with interface conditions:
(cid:104)(cid:104)
(cid:17)
,j
ijhkN (2,2)
hp,k
(cid:16) C m
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
− ρmδip = −δip
(cid:104)(cid:104) C m
N (2,2)
ip
= 0,
ijhkN (2,2)
hp,k
= 0,
(39)
(40)
as well as the cell problem (35) related to the case s0. From the resolution of problem (39) and (40), the
perturbation function N (2,2)
The perturbation function N (3,2)
ipq1
is derived.
ip
is the solution of the cell problem obtained at the order ε1
hp,j − ρmN (1,0)
hp,j − ρmN (1,0)
iq1hjN (2,2)
iq1hjN (2,2)
),j + C m
=
ipq1
ipq1
ijhkN (3,2)
hpq1,k
+
j
( C m
ijhq1
N (2,2)
(cid:69)
(cid:17)
(cid:105)
hp
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:104)(cid:104) C m
N (3,2)
ipq1
= 0,
ijhkN (3,2)
hpq1,k + C m
ijhq1
N (2,2)
hp
)nj
= 0.
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
Meanwhile, at the order ε2, the perturbation function N (4,2)
ipq1q2
is solution of the cell problem
(cid:104)
(cid:104)(cid:104)
(cid:16) C m
(cid:16) C m
with interface conditions
(cid:17)
(cid:104)
ijhkN (4,2)
hpq1q2,k
+
,j
1
2
( C m
ijhq2
N (3,2)
hpq1
),j + C m
iq2hq1
N (2,2)
hp + C m
iq2hkN (3,2)
hpq1,k+
.
ρm(cid:69)
(cid:68)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:68) C m
(41)
(42)
(43)
(44)
(45)
(46)
ijhkN (4,2)
hpq1q2,k +
N (3,2)
hpq1
+ C m
ijhq1
N (3,2)
hpq2
ijhq2
= 0.
Finally at the order εw+2, with w ∈ Z and w ≥ 1, the perturbation function N (w+4,2)
cell problem
ipq1...qw+2
is derived from the
− ρmN (2,0)
hpq1q2
− ρmN (2,0)
hpq2q1
(cid:105)
+ ( C m
ijhq1
N (3,2)
hpq2
),j + C m
iq1hq2
N (2,2)
hp + C m
iq1hkN (3,2)
hpq2,k+
=
(cid:104) C m
iq2hq1
1
2
N (2,2)
hp + C m
iq2hkN (3,2)
hpq1,k − ρmN (2,0)
hpq1q2
+
iq1hq2
+ C m
N (2,2)
iq1hkN (3,2)
whose interface conditions are
hp + C m
(cid:105),
hpq2,k − ρmN (2,0)
(cid:104)(cid:104)
hpq2q1
N (4,2)
ipq1q2
= 0,
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:17)(cid:17)
(cid:17)
nj
(cid:16) C m
1
2
(cid:17)
(cid:88)
(cid:104)(cid:16) C m
1
w + 2
P ∗(q)
ijhkN (w+4,2)
hpq1....qw+2,k
+
,j
ijhqw+2
N (w+3,2)
hpq1...qw+1
+
,j
+ C m
iqw+2hqw+1
N (w+2,2)
hpq1....qw
+ C m
iqw+2hjN (w+3,2)
hpq1...qw+1,j − ρmN (w+2)
ipq1...qw+2
(cid:105)
=
(cid:104)(cid:104)(cid:16) C m
(cid:16) C m
(cid:88)
P ∗(q)
=
1
w + 2
equipped with the interface conditions
(cid:104) C m
iqw+2hqw+1
N (w+2,2)
hpq1....qw
+ C m
iqw+2hjN (w+3,2)
hpq1...qw+1,j − ρmN (w+2,0)
ipq1...qw+2
(cid:105),
(cid:104)(cid:104)(cid:16) C m
ijhqw+2
(cid:104)(cid:104)
N (w+4,2)
ipq1...qw+2
= 0,
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:16) C m
(cid:88)
9
N (w+4,2)
hpq1...qw+2,k +
1
w + 2
P ∗(q)
ijhqw+2
N (w+3,2)
hpq1...qw+1
(cid:17)(cid:17)
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0.
Cell problems related to s2n
In the present subsection, the cell problems related to power s2n are devised and their corresponding per-
turbation functions are established. At the order ε(2n−2), with n ∈ Z and n ≥ 2, the cell problem is
with interface conditions
ijhkN (2n,2n)
hp,k
− ρmN (2n−2,2n−2)
= −(cid:104)ρmN (2n−2,2n−2)
(cid:105),
ip
N (2n,2n)
ip
= 0,
ijhkN (2n,2n)
hp,k
= 0,
(cid:104)(cid:104) C m
ip
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
and its solution is the perturbation function N (2n,2n)
Whereas at the order ε(2n−1), the perturbation function N (2n+1,2n)
ip
.
ipq1
is the solution of the cell problem
ijhkN (2n+1,2n)
hpq1,k
+
,j
N (2n,2n)
hp
ijhq1
+ C m
iq1hkN (2n,2n)
hp,k − ρmN (2n−1,2n−2)
ipq1
=
(cid:105)
,j
(cid:17)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:17)
,j
(cid:69)
,
(cid:16) C m
(cid:104)(cid:104)
(cid:104)(cid:16) C m
(cid:104)(cid:104)
(cid:104)(cid:16) C m
(cid:17)
(cid:17)
(cid:16) C m
(cid:68) C m
=
(cid:16) C m
(cid:68) 1
2
(47)
(48)
(49)
(50)
(51)
(52)
iq1hkN (2n,2n)
hp,k − ρmN (2n−1,2n−2)
ipq1
with interface conditions:
(cid:104)(cid:104)
N (2n+1,2n)
ipq1
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0,
( C m
ijhkN (2n+1,2n)
hpq1,k
+ C m
ijhq1
N (2n,2n)
hp
)nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0.
The cell problem evaluated for ε(2n) is
ijhkN (2n+2,2n)
hpq1q2,k
+
,j
1
2
N (2n+1,2n)
hpq1
ijhq2
N (2n,2n)
hp
+
+ C m
iq2hjN (2n+1,2n)
hpq1,j
− ρmN (2n,2n−2)
ipq1q2
+
N (2n+1,2n)
hpq2
ijhq1
+
,j
+ C m
iq1hq2
N (2n,2n)
hp
+ C m
iq1hjN (2n+1,2n)
hpq2,j
− ρmN (2n,2n−2)
ipq2q1
=
(cid:17)
+ C m
iq2hq1
,j
(cid:16) C m
(cid:17)
(cid:105)
C m
iq2hq1
N (2n,2n)
hp
+ C m
iq2hjN (2n+1,2n)
hpq1,j
− ρmN (2n,2n−2)
ipq1q2
+
+ C m
iq1hq2
N (2n,2n)
hp
+ C m
iq1hjN (2n+1,2n)
hpq2,j
− ρmN (2n,2n−2)
(cid:105),
ipq2q1
with interface conditions
(cid:104)(cid:104)(cid:16) C m
ijhkN (2n+2,2n)
hpq1q2,k +
(cid:16) C m
ijhq2
1
2
(cid:104)(cid:104)
N (2n+2,2n)
ipq1q2
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
N (2n+1,2n)
hpq1
+ C m
ijhq1
N (2n+1,2n)
hpq2
= 0,
(cid:17)(cid:17)
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0,
and its solution is the perturbation function N (2n+2,2n)
Finally, the perturbation function N (w+2n+2,2n)
ipq1q2
.
is the solution of the cell problem for εw+2n
(cid:16) C m
(cid:17)
(cid:88)
ipq1...qw+2
(cid:104)(cid:16) C m
ijhqw+2
1
w + 2
P ∗(q)
ijhkN (w+2n+2,2n)
hpq1...qw+2,k
+
,j
(cid:17)
+
,j
N (w+2n+1,2n)
hpq1...qw+1
10
N (w+2n,2n)
hpq1...qw
+ C m
iqw+2hjN (w+2n+1,2n)
hpq1...qw+1,j − ρmN (w+2n,2n−2)
ipq1...qw+2
(cid:105)
=
+ C m
iqw+2hqw+1
(cid:88)
P (q)
=
1
w + 2
(cid:104) C m
iqw+2hqw+1
N (w+2n,2n)
hpq1...qw
+ C m
iqw+2hjN (w+2n+1,2n)
hpq1...qw+1,j − ρmN (w+2n,2n−2)
ipq1...qw+2
(cid:105),
whose interface conditions are
(cid:104)(cid:104)(cid:16) C m
ijhkN (w+2n+2,2n)
hpq1...qw+2,k +
N (w+2n+2,2n)
ipq1...qw+2
= 0,
(cid:104)(cid:104)
(cid:88)
1
w + 2
P ∗(q)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:16) C m
ijhqw+2
(53)
(cid:17)(cid:17)
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0.
N (w+2n+1,2n)
hpq1...qw+1
In Bakhvalov and Panasenko [1984], it is emphasized that the uniqueness of the perturbation functions
N (i,2r)
, derived from the cell problems (33)-(53), is guaranteed by imposing the normalization condition
hpq1...qi−2r
(cid:104)N (i,2r)
(cid:105) = 0.
hpq1...qi−2r
5 Down-scaling relation, average field equation of infinite order
and macroscopic problems
The down-scaling relation referred to the transformed micro-displacement is expressed as an asymptotic
expansion of powers of the microscopic length ε relying on the transformed macro-displacement U M
h (x, s),
its gradients and the Q-periodic perturbation functions. Such functions are delivered by solving the cell
problems that are listed in the Section (4). Therefore, the replacement of the solutions of the recursive
differential problems (23), (27), (32), (133) and (137) into the asymptotic expansion (16) enables establishing
the transformed micro-displacement uh(x, ξ, s) as
=
(54)
(cid:17)
+
+ +N (2,2)
hp
(ξ)s2 U M
p
(cid:17)
, s
=
(cid:16) +∞(cid:88)
l,j=0
εj+l(cid:88)
q=l
h (x, s) + εN (1,0)
hpq1
(ξ)
∂ U M
p
∂xq1
N (2j+l,2j)
hpq
(ξ)
∂l U M
p
∂xq
+ ε2(cid:16)
N (2,0)
hpq1q2
(ξ)
N (3,0)
hpq1q2q3
(ξ)
∂3 U M
p
∂xq1∂xq2∂xq3
+ N (3,2)
hpq1
(ξ)s2 ∂ U M
p
∂xq1
s2j(cid:17)(cid:12)(cid:12)(cid:12)ξ= x
ε
∂2 U M
p
∂xq1∂xq2
(cid:17)
+
x
ε
=
x,
uh
(cid:16)
(cid:16) U M
+ ε3(cid:16)
ε4(cid:16)
(cid:17)
+ O(ε5)
(cid:17)(cid:12)(cid:12)(cid:12)ξ= x
ε
.
N (4,0)
hpq1q2q3q4
(ξ)
∂4 U M
p
∂xq1∂xq2∂xq3∂xq4
+ N (4,2)
hpq1q2
(ξ)s2 ∂2 U M
∂xq1∂xq2
p
+ N (4,4)
hp
(ξ)s4 U M
p
= ∂l(·)
∂xq1 ...xql
. Moreover, the perturbation function N (0,0)
In Eq. (54), q describes the lenght of the multi-index and the derivative with respect to q is written as
∂l(·)
stands for the Kronecker delta δhp. There is
∂xq
point in observing that the Q-periodic perturbation functions N (2j+l,2j)
are affected by the microstructural
inhomogeneities of the material and this is emphasized by their dependency on the fast variable ξ = x
ε .
h (x, s) is L-periodic and relies on the slow
On the other hand, the transformed macro-displacement U M
variable x and the time. The transformed macro-displacement field is supposed to be the mean value of the
transformed micro-displacement field over the unit cell Q
hpq
hp
(cid:68)
(cid:16)
(cid:17)(cid:69)
U M
h (x, s) =
uh
x,
x
ε
+ ζ, s
.
(55)
11
Eq. (55) is said to be the up-scaling relation and it links the transformed macro-displacement field with
the transformed micro-displacement field. In Eq. (55) the variable ζ ∈ Q identifies a family of translations
of the heterogeneous domain respect to the L−periodic body forces b(x, t), see Smyshlyaev and Chered-
nichenko [2000], Bacigalupo [2014]. Therefore, the transformed body forces in the Laplace space b(x, s) are
L−periodic.
Replacing the down-scaling relation (54) into the micro-field Eq. (13) and assembling the terms with equal
powers of ε, the average field equations of infinite order read
n(n+3,0)
ipq
∂n+3 U M
p
∂xq
+
n(n+3,2)
ipq
s2 ∂n+1 U M
p
∂xq
+
+∞(cid:88)
εn+1 (cid:88)
n=0
q=n+1
∂2 U M
p
∂xq1∂xq2
− n(2,2)
ip
s2 U M
p +
ε2n+2n(2n+4,2n+4)
ip
s2n+4 U M
n(2,0)
ipq1q2
− +∞(cid:88)
n=0
εn+1 (cid:88)
+∞(cid:88)
p − +∞(cid:88)
n=0
q=n+3
ε2n+n+3 (cid:88)
n,n=0
q=n+1
n(2n+n+5,2n+4)
ipq
s2n+4 ∂n+1 U M
p
∂xq
+ bi(x, s) = 0,
(56)
where the coefficients of the gradients of the transformed macro-displacement are the known terms of the
corresponding cell problems. Therefore it results
+ C m
iq2hkN (1,0)
hpq1,k + C m
iq1hq2
+ C m
iq1hkN (1,0)
hpq2,k(cid:105),
(57)
(58)
(59)
(60)
(61)
(62)
(63)
(64)
n(2,0)
ipq1q2
=
(cid:104) C m
iq2hq1
1
2
ip = δip(cid:104)ρm(cid:105),
n(2,2)
(cid:88)
w + 2
P ∗(q)
n(w+2,0)
ipq1...qw+2
=
1
(cid:104) C m
iqw+2hjN (w+1,0)
hpq1...qw+1,j + C m
iqw+2hqw+1
N (w,0)
hpq1...qw
(cid:105),
n(3,2)
ipq1
= (cid:104)ρmN (1,0)
ipq1
− C m
hp,j (cid:105),
iq1hjN (2,2)
n(2n+4,2n+4)
ip
= (cid:104)ρmN (2n+2,2n+2)
ip
(cid:105),
n(2n+5,2n+4)
ipq1
= (cid:104)ρmN (2n+3,2n+2)
ipq1
− C m
iq1hkN (2n+4,2n+4)
hp,k
(cid:105),
n(2n+6,2n+4)
ipq1q2
=
1
2
(cid:104)ρmN (2n+4,2n+4)
ipq1q2
− C m
iq2hq1
N (2n+4,2n+4)
hp
− C m
iq2hjN (2n+5,2n+4)
hpq1,j
+
+ ρmN (2n+4,2n+2)
ipq2q1
− C m
iq1hq2
N (2n+4,2n+2)
hp
− C m
iq1hjN (2n+5,2n+4)
hpq2,j
(cid:105),
n(2n+w+6,2n+w+4)
ipq1...qw+2
=
(cid:104)ρmN (2n+w+4,2n+2)
ipq1...qw+2
− C m
iqw+2hqw+1
N (2n+w+4,2n+4)
hpq1...qw
+
(cid:88)
1
w + 2
P ∗(q)
− C m
iqw+2hjN (2n+w+5,2n+4)
hpq1...qw+1,j
(cid:105),
with w ∈ Z, w ≥ 1, n ∈ Z and n ≥ 0. The average field equations of infinite order
solved by performing an asymptotic expansion of the transformed macro-displacement U M
ε, namely
(56) are formally
p (x) in power of
U M
p (x) =
εjU j
p (x).
(65)
+∞(cid:88)
j=0
12
p + ε U (1)
p + ...)+
(cid:16) ∂4 U (0)
p
∂4 U (1)
p
+ ε
∂xq1...∂xq4
∂xq1...∂xq4
(66)
(cid:17)
+ ...
+
(cid:17)
(cid:17)
(cid:17)
The substitution of Eq.
(65) into Eq.
(cid:16)
n(2,0)
ipq1q2
+ ε
p
∂2 U (1)
∂xq1∂xq2
+ εn(3,0)
ipq1...q3
∂xq1 ...∂xq3
∂xq1...∂xq3
+ ε
p
p
p
∂xq1
ε0 ∂2 U (0)
∂xq1∂xq2
(cid:16) ∂3 U (0)
s2(cid:16) ∂ U (0)
s4(cid:16) U (0)
s4(cid:16) ∂ U (0)
s6(cid:16) ∂8 U (0)
p
∂xq1
+ ε
+ ε
ip
p
∂xq1
∂3 U (1)
p
ip
+ ...
+ ...
ipq1...q4
s2( U (0)
+ ε2n(4,0)
(56) leads to
(cid:17) − n(2,2)
(cid:17)
s2(cid:16) ∂2 U (0)
(cid:17) − ε2n(4,2)
s6(cid:16) U (0)
(cid:17) − ε4n(6,6)
(cid:17) − ε4n(6,4)
s4(cid:16) ∂2 U (0)
s6(cid:16) ∂2 U (0)
(cid:17) − ε6n(8,6)
∂xq1∂xq2
ipq1q2
ipq1q2
ip
p
p
p
ipq1q2
∂xq1∂xq2
... − εn(3,2)
ipq1
+ ε
+ ...
∂ U (1)
p
∂xq1
+ ε
p
∂2 U (1)
∂xq1∂xq2
+ ...
+
∂xq1∂xq2
... − ε2n(4,4)
p + ε U (1) + ...
p + ε U (1) + ...
+ ...+
(cid:17)
ε3n(5,4)
ipq1
ε5n(7,6)
ipq1
∂ U (1)
p
∂xq1
+ ...
∂ U (1)
p
∂xq1
+ ...
+ ε
p
∂2 U (1)
∂xq1∂xq2
+ ...
+ ...+
+ ε
p
∂2 U (1)
∂xq1 ∂xq2
+ ...
+ ... + bi(x, s) = 0,
which provides the following macroscopic recursive problems for the different orders of ε. Namely, at the
order ε0 it results
n(2,0)
ipq1q2
p
∂2 U (0)
∂xq1∂xq2
− n(2,2)
ip
s2 U (0)
p + bi(x, s) = 0,
at the order ε the problem is
n(2,0)
ipq1q2
p
∂2 U (0)
∂xq1∂xq2
− n(2,2)
ip
s2 U (0)
p + n(3,0)
ipq1...q3
instead at the order ε2 it reads
n(2,0)
ipq1q2
p
∂2 U (0)
∂xq1∂xq2
− n(2,2)
ip
s2 U (2)
p + n(3,0)
ipq1...q3
∂3 U (0)
p
∂xq1 ...∂xq3
− n(3,2)
ipq1
s2 ∂ U (0)
p
∂xq1
= 0,
∂3 U (1)
p
∂xq1 ...∂xq3
+ n(4,0)
ipq1...q4
∂4 U (0)
p
∂xq1...∂xq4
+
− n(3,2)
ipq1
s2 ∂ U (0)
p
∂xq1
− n(4,2)
ipq1q2
s2 ∂2 U (0)
∂xq1∂xq2
p
− n(4,4)
ip
s4 U 0
p = 0.
(67)
(68)
(69)
A generic recursive problem, at odd order ε2 w−1, with w ∈ Z and w ≥ 2 is
∂r U (2 w+1−r)
− n(2 w−1,2 w−1)
ip
s2 U (2 w−1)
p
+
n(2 w−1,0)
ipq1q2
∂2 U (2 w−1)
∂xq1 ∂xq2
p
n(r,0)
ipq
p
∂xq
+
(70)
− s2
2 w+1(cid:88)
2 w−1−2n(cid:88)
r=3
(cid:88)
(cid:88)
q=r−2
+
r=3
q=r−2
n(r,2)
ipq
∂r−2 U (2 w+1−r)
p
∂xq
− (1 − δ2 w)
s2n+4(cid:16)
n(2n+4,2n+4)
ip
U (2 w−3−2n)
p
(cid:17)
+
n(r+2+2n,2n+4)
ipq
∂r−2 U (2 w−1−r−2n)
p
∂xq
− n(2 w,2 w)
ip
13
s2 w U (1)
p − n(2 w,2 w)
ipq1
s2 w
p
U (0)
∂xq1
= 0,
(cid:88)
2 w+1(cid:88)
w−3+δ2 w(cid:88)
q=r
r=3
n=0
whereas a generic recursive problem at even order ε2 w is
n(2 w,0)
ipq1q2
p
∂2 U (2 w)
∂xq1∂xq2
− n(2 w,2 w)
ip
s2 U (2 w)
p
+
∂r U (2 w+1−r)
p
∂xq
+
(71)
n(r,0)
ipg
(cid:88)
2 w+2(cid:88)
s2n+4(cid:16)
− w−2(cid:88)
q=r
r=3
n=0
− s2
2 w+2(cid:88)
2 w−2n(cid:88)
r=3
(cid:88)
(cid:88)
q=r−2
r=3
q=r−2
+
n(r,2)
ipq
∂r−2 U (2 w+2−r)
p
∂xq
n(2n+4,2n+4)
ip
U (2 w−2−2n)
p
+
n(r+2+2n,2n+4)
ipq
∂r−2 U (2 w−r−2n)
p
∂xq
− n(2 w+2,2 w+2)
ip
s2 w+2 U (0)
p = 0,
(cid:17)
(cid:90)
(cid:16)
(cid:17)
Λ =
λm
L
x,
x
ε
dx =
(cid:90)
(cid:16) 1
2
L
where n ∈ Z and n ≥ 2. There is no point in managing the averaged equation of infinite order (56). In
addition, the ellipticity of the differential problem could be not guaranteed if Eq. (56) is truncated at a
certain order. To overcome such a disadvantage, an asymptotic-variational approach is pursued.
6 Asymptotic expansion of the energy and second order homoge-
nization
In this Section a finite order governing equation is provided by exploiting a variational-asymptotic procedure,
see Smyshlyaev and Cherednichenko [2000], Bacigalupo and Gambarotta [2014a]. Let Λ be the energy-like
functional written in terms of the energy-like density λm at the microscale and referred to the periodic
domain L, Fabrizio and Morro [1992],
ρm u ∗ u +
1
2
∇u ∗ (Gm ∗ ∇ u) − u ∗ b
dx.
(72)
Let L(Λ) be the energy-like functional in the Laplace domain, which is expressed in terms of the energy-like
density λm in the Laplace domain
Λ = L(Λ) =
∇u : Gm∇u − u · b
ρms2 u · u +
(cid:16) 1
dx =
(cid:17)
(cid:16)
(cid:17)
(cid:90)
(cid:90)
(73)
dx,
x,
λm
L
x
ε
2
L
1
2
be written as Gm,ζ(cid:16)
where the symbol : denotes the second order inner product. Specifically, the tranformed energy-like functional
Λ and its corresponding energy-like density λm are influenced by the translation variable ζ ∈ Q. Such a
variable is introduced because the actual "phase" of the microstructure is undetectable and a family of
translated microstructures is taken into account.
Therefore, the transformed micro-relaxation tensor Gm depends on the translation variable ζ and it may
, ...,, which are
solutions of the cell problems determined in Section 4, reckon on variable ζ. In addition, the energy-like
density λm in the Laplace domain complies with the property λζ
and so the Laplace
m
transform of the energy-like functional Λ, depending on the parameter ζ, is
and the perturbation functions N (1,0)
hpq1
= Gm(cid:16)
, N (2,0)
hpq1q2
= λm
ε + ζ
ε + ζ
x, x
ε
x, x
ε
(cid:17)
(cid:17)
(cid:16)
(cid:17)
(cid:16)
(cid:17)
x, x
x, x
(cid:90)
(cid:16)
(cid:17)
(cid:16)
Λζ = Λ(ζ) =
λζ
m
L
x,
x
ε
dx =
λm
L
x,
x
ε
+ ζ
(cid:90)
(cid:90)
1
Q
Q
dx.
(cid:17)
(cid:68)λm
(cid:16)
(cid:90)
L
Let Λm be the average transformed energy-like functional at the microscale
Λm =(cid:104)Λζ(cid:105) =
Λ(ζ)dζ =
Λζdζ =
+ ζ
dx,
x,
(cid:17)(cid:69)
x
ε
(cid:90)
1
Q
Q
(74)
(75)
where the Fubini theorem is applied. The average transformed energy-like functional (cid:104)Λζ(cid:105) at the microscale
does not rely on the translation variable ζ because the energy-like functional Λζ is averaged with regard to
14
the translated realizations of the microstructure and so the transformed energy-like density at the micoscale
satisfies
(cid:16)
(cid:68)λm
(cid:17)(cid:69)
(cid:90)
(cid:16)
(cid:17)
x,
x
ε
+ ζ
=
1
Q
λm
Q
x,
x
ε
+ ζ
dζ =
1
Q
Q
(cid:90)
(cid:16)
(cid:17)
λm
x, ξ
dξ = (cid:104)λm(x, ξ)(cid:105).
(76)
Two methods are herein proposed to determine the governing field equation at the macroscale and the overall
constitutive and inertial tensors.
Approximation of the energy-like functional through truncation of its asymptotic
expansion
Let us consider the down-scaling relation related to the transformed micro-displacement u(x, ξ, s), i.e.
uh(x, ξ, s) = Uh(x, s) + εN (1,0)
hpq1
(ξ)
∂ U M
p
∂xq1
N (2,0)
hpq1q2
(ξ)
∂2 U M
p
∂xq1∂xq2
+ N (2,2)
hp
(ξ)s2 U M
p
+
(77)
(cid:17)
+ ε3(cid:16)
N (3,0)
hpq1q2q3
(ξ)
∂3 U M
p
∂xq1∂xq2∂xq3
+ N (3,2)
hpq1
(ξ)s2 ∂ U M
p
∂xq1
+ O(ε4).
+ ε2(cid:16)
(cid:17)
Let us replace the down-scaling relation (77) into the tranformed energy-like functional (74) and let us
suppose that Λm is truncated at the second order. After applying the divergence theorem, the transformed
energy-like functional at the second order is
ΛII
m =
(cid:104)λII
m (x, ξ)(cid:105)dx =
s2(cid:104)ρm(cid:105)
U M
h
h dx + εs2(cid:104)ρmN (1,0)
U M
rpq1
(cid:105)
∂ U M
p
∂xq1
U M
r dx+
(78)
L
(cid:90)
L
+ εs3(cid:104) Gm
hkijB(1,0)
hkpq1
ijr (cid:105)
B(2,2)
∂ U M
p
∂xq1
U M
r dx+
+ εs(cid:104) Gm
hkijB(1,0)
hkpq1
B(2,0)
ijrw1w2
(cid:105)
∂ U M
p
∂xq1
∂2 U M
r
∂xw1∂xw2
L
ρmN (1,0)
hpq1
N (1,0)
hrq2
− ρmN (2,0)
rpq1q2
+ ε2s3(cid:104) Gm
hkijB(1,0)
hkpq1
B(3,2)
ijrq2
− Gm
hkijB(2,0)
hkpq1q2
ijr (cid:105)
B(2,2)
∂ U M
p
∂xq1
∂ U M
r
∂xq2
L
dx + ε2s4(cid:104)ρmN (2,2)
(cid:105)
rp
+ ε2s5 1
2
(cid:104) Gm
hkijB(2,2)
ijr (cid:105)
hkp B(2,2)
U M
p
U M
r dx +
1
2
s(cid:104) Gm
L
hkijB(1,0)
hkpq1
B(1,0)
ijrw1
∂ U M
p
∂xq1
∂ U M
r
∂xw1
dx+
L
dx + ε2s2(cid:68) 1
(cid:90)
2
(cid:69)(cid:90)
(cid:69)(cid:90)
L
∂ U M
p
∂xq1
∂ U M
r
∂xq2
dx+
U M
p
U M
r dx+
(cid:90)
L
(cid:90)
L
U M
h
bhdx.
+ ε2s
Gm
hkijB(2,0)
hkpq1q2
B(2,0)
ijrw1w2
− Gm
hkijB(1,0)
ijrw1
B(3,0)
hkpq1q2w2
∂2 U M
p
∂xq1∂xq2
∂2 U M
r
∂xw1∂xw2
dx −
L
It is important to note that second-order gradients of the transformed macro-displacement are takent into
account as well as the first-order gradient of the displacement (i.e., strain), by generalizing the standard
continuum mechanics.
The localization tensors, appearing in Eq. (78), assume the form
B(1,0)
hkpq1
= δhpδkq1 + N (1,0)
hpq1,k,
(cid:16)
B(2,0)
hkpq1q2
=
1
2
δkq2N (1,0)
hpq1
+ δkq1N (1,0)
hpq2
(cid:17)
+ N (2,0)
hpq1q2,k,
15
(79)
(80)
1
2
(cid:90)
L
(cid:90)
(cid:90)
(cid:90)
L
(cid:68) 1
2
(cid:90)
(cid:90)
(cid:105)
(cid:90)
L
(cid:90)
L
∂ U M
p
∂xq1
L
(cid:90)
δ U M
r dx+
∂3 U M
p
(cid:90)
(cid:90)
(cid:90)
∂2 U M
p
∂xq1 ∂xq2
L
hkp = N (2,2)
B(2,2)
hp,k ,
B(3,0)
hkpq1q2q3
=
1
3
(cid:16)
δkq3N (2,0)
hpq1q2
+ δkq1N (2,0)
hpq2q3
+ δkq2 N (2,0)
hpq3q1
(cid:17)
+ N (3,0)
hpq1q2q3,k,
B(3,2)
hkpq1
= δkq1N (2,2)
hp + N (3,2)
hpq1,k.
(81)
(82)
(83)
The localization tensors are periodic functions with regard to the fast coordinate ξ since the perturbation
functions and their gradients are Q- periodic functions. Both B(2,0)
are symmetrized
with respect to the indices q1, q2 and q1, q2, q3, respectively (see Appendix B). The governing equation of
a non-local homogeneous continuum is delivered by determining the stability condition of the transformed
energy-like functional ΛII
m , which is found to be the first variation of the average transformed energy-like
functional δ ΛII
m ,
and B(3,0)
hkpq1q2q3
hkpq1q2
δ ΛII
m ( U M
t
, δ U M
t ) = s2(cid:104)ρm(cid:105)
U M
t δ U M
t dx + εs2(cid:104)ρm(N (1,0)
rpq1
− N (1,0)
prq1
)(cid:105)
∂ U M
p
∂xq1
δ U M
r dx+
(84)
L
(cid:90)
+ εs3(cid:104) Gm
hkijB(1,0)
hkpq1
ijr − Gm
B(2,2)
hkijB(1,0)
hkrq1
ijp (cid:105)
B(2,2)
+ εs(cid:104) Gm
hkijB(1,0)
hkpq1
B(2,0)
ijrw1w2
− Gm
hkijB(1,0)
hkrw1
B(2,0)
ijpq1w2
(cid:105)
∂xw1∂xw2 ∂xq1
L
δ U M
r dx+
− ε2s2(cid:104)ρmN (1,0)
hpq1
N (1,0)
hrq2
− ρm(N (2,0)
rpq1q2
+ N (2,0)
prq2q1
)(cid:105)
∂2 U M
p
∂xq1∂xq2
δ U M
r dx − ε2s3(cid:104) Gm
hkijB(1,0)
hkpq1
L
B(3,2)
ijrq2
− Gm
hkijB(2,0)
hkpq1q2
B(2,2)
ijp +
+ Gm
hkijB(1,0)
hkrq2
B(3,2)
ijpq1
− Gm
hkijB(2,0)
hkrq2q1
ijp (cid:105)
B(2,2)
δ U M
r dx + ε2s4(cid:104)ρm(N (2,2)
rp + N (2,2)
pr
)(cid:105)
U M
p δ U M
r dx+
(cid:90)
L
(cid:90)
∂2 U M
p
∂xw1∂xq1
L
δ U M
r dx+
(cid:90)
+ ε2s5(cid:104) Gm
hkijB(2,2)
ijr (cid:105)
hkp B(2,2)
U M
p δ U M
r dx − s(cid:104) Gm
hkijB(1,0)
hkpq1
B(1,0)
ijrw1
(cid:105)
+ ε2s(cid:104) Gm
hkijB(2,0)
hkpq1q2
B(2,0)
ijrw1w2
− Gm
hkijB(1,0)
hkrq1
B(3,0)
ijpq1q2w2
− Gm
hkijB(1,0)
hkpq1
B(3,0)
ijrw1w2q2
(cid:105)
(cid:90)
L
−
δ U M
t
btdx,
∂4 U M
p
∂xw2 ∂xw1∂xq1∂xq2
L
δ U M
r dx+
where the Q−periodicity of the functions Gm
B(3,2)
hkpq1
Lagrangian differential equation associated with the variational problem (84) in the Laplace domain is
hkij, N (1,0)
ipq1
is taken into account. The first variation δ ΛII
and the localization tensors B(1,0)
hkpq1
and so the Euler-
m must vanish for all admissible δ U M
, N (2,0)
ipq1q2
, B(2,0)
hkpq1q2
t
,...,
s2ρ U M
t + s2ρ( Itpq1 − Itq1p)
∂ U M
p
∂xq1
− s2ρ Itq2pq1
∂2 U M
p
∂xq1 ∂xq2
+ s4ρ I (cid:93)
tp
U M
p =
− s3( Jtpq1 − Jtq1p)
∂ U M
p
∂xq1
− s3 J 1
tq2pq1
∂2 U M
p
∂xq1∂xq2
+
16
− s5 J (cid:93)
tp
p + s Gtr1pq1
U M
∂2 U M
p
∂xq1 ∂xr1
+ s( Ytr1pq1r2 − Ytr1r2pq1 )
∂3 U M
p
∂xr1 ∂xr2 ∂xq1
+
− s S1
tr1r2pq1q2
∂4 U M
p
∂xq1∂xq2 ∂xr1 ∂xr2
+ bt,
(85)
which is formulated in terms of the transformed macro-displacement and its gradients up to the fourth order.
The components of the constitutive tensors in the Laplace domain related to the homogenized continuum
are defined as
Gtr1pq1 = (cid:104) Gm
(86)
(cid:105),
hkijB(1,0)
hkpq1
B(1,0)
ijtr1
Ytr1pq1r2 = ε(cid:104) Gm
hkijB(1,0)
hktr1
B(2,0)
ijpq1r1
(cid:105),
Ytr1r2pq1 = Ypq1tr1r2 = ε(cid:104) Gm
hkijB(1,0)
hkpq1
B(2,0)
ijtr1r2
(cid:105),
S1
tr1r2pq1q2
= ε2(cid:104) Gm
hkijB(2,0)
hkpq1q2
B(2,0)
hktr1r2
− Gm
hkijB(1,0)
ijtr1
B(3,0)
hkpq1q2r2
− Gm
hkijB(1,0)
ijpq1
B(3,0)
hktr1r2q2
(cid:105),
Jtpq1 = ε(cid:104) Gm
hkijB(1,0)
hkpq1
B(2,2)
ijt
(cid:105),
Jtq1p = Jptq1 = ε(cid:104) Gm
hkijB(1,0)
hktq1
ijp (cid:105),
B(2,2)
tp = Jtp + Jpt =
J (cid:93)
ε2
2
(cid:104) Gm
hkijB(2,2)
hkp B(2,2)
ijt
(cid:105) +
ε2
2
(cid:104) Gm
hkijB(2,2)
hkt B(2,2)
ijp (cid:105) = ε2(cid:104) Gm
hkijB(2,2)
hkp B(2,2)
ijt
(cid:105),
J 1
tq2pq1
= J 1
pq1tq2
= −ε2(cid:104)− Gm
hkijB(1,0)
hkpq1
B(3,2)
ijtq2
+ Gm
hkijB(2,0)
hkpq1q2
ijp − Gm
B(2,2)
hkijB(1,0)
hktq2
B(3,2)
ijpq1
+ Gm
hkijB(2,0)
hktq2q1
(87)
(88)
(89)
(90)
(91)
(92)
(cid:105),
B(2,2)
ijt
(93)
tr1pq1
, Ytr1pq1r2 and S1
where the components Gm
tr1r2pq1q2
computed due to the micro-fluctuation functions N (1,0)
with the ones determined in Bacigalupo [2014].
The transformed inertial tensor components are given as
ρ = (cid:104)ρm(cid:105),
ikl
of the constitutive tensors in the Laplace domain are
, N (2,0)
iklpq. Such tensors are in accordance
iklp and N (3,0)
Itpq1 = ε(cid:104)ρmN (1,0)
tpq1 (cid:105) 1
ρ
,
Itq1p = Iptq1 = ε(cid:104)ρmN (1,0)
ptq1 (cid:105) 1
ρ
,
pt = Itp + Ipt = ε2(cid:104)ρm(N (2,2)
I (cid:93)
tp + N (2,2)
pt
)(cid:105) 1
ρ
,
Itp =
Ipt =
ε2
ρ
ε2
ρ
(cid:104)ρmN (2,2)
tp
(cid:105),
(cid:104)ρmN (2,2)
pt
(cid:105),
Itq2pq1 = ε2(cid:104)ρmN (1,0)
hpq1
N (1,0)
htq2
− ρm(N (2,0)
tpq1q2 + N (2,0)
tpq2q1)(cid:105) 1
ρ
.
17
(94)
(95)
(96)
(97)
(98)
(99)
(100)
(cid:16)
(cid:90)
L
1
2
(cid:90)
L
(cid:90)
L
(cid:90)
ijr (cid:105)
B(2,2)
(cid:90)
L
U M
p
(cid:69)(cid:90)
(cid:90)
L
(cid:68) 1
2
(cid:90)
(cid:90)
L
(cid:90)
(cid:90)
Approximation of the energy-like functional through truncation of the down-
scaling relation
An alternative approach is here presented to evaluate the overall constitutive and inertial tensors. To this
purpose, the gradient referred to the down-scaling relation (77) is approximated at the first order as
=
∂ Uh
∂xk
+ N (1,0)
hpq1,k
∂ U M
p
∂xq1
+ ε
N (1,0)
hpq1
∂2 U M
p
∂xq1∂xq2
+ +N (2,0)
hpq1q2,k(ξ)
∂2 U M
p
∂xq1∂xq2
+ N (2,2)
hp,k (ξ)s2 U M
p
(cid:16) Duh
(cid:17)I
Dxk
(cid:17)
.
(101)
+ ε2(cid:16)
(cid:17)
In addition, the transformed micro-displacement at the second order is formulated as
h (x, ξ, s) = Uh(x, s) + εN (1,0)
uII
hpq1
(ξ)
∂ U M
p
∂xq1
N (2,0)
hpq1q2
(ξ)
∂2 U M
p
∂xq1 ∂xq2
+ N (2,2)
hp
(ξ)s2 U M
p
,
(102)
where the perturbation functions are helpful to determine a consistent approximation of the gradient at the
first order. Then the gradient approximation in Eq. (101) and the displacement approximation in Eq. (102)
are replaced into the transformed energy-like functional (75), which is approximated at the second order as
ΛII
m =
(cid:104)λII
m (x, ξ)(cid:105)dx =
s2(cid:104)ρm(cid:105)
U M
h
h dx + εs2(cid:104)ρmN (1,0)
U M
rpq1
(cid:105)
∂ U M
p
∂xq1
U M
r dx+
(103)
L
+ εs3(cid:104) Gm
hkijB(1,0)
hkpq1
ijr (cid:105)
B(2,2)
∂ U M
p
∂xq1
U M
r dx+
+ εs(cid:104) Gm
hkijB(1,0)
hkpq1
B(2,0)
ijrw1w2
(cid:105)
∂ U M
p
∂xq1
∂2 U M
r
∂xw1∂xw2
dx + ε2s2(cid:68) 1
2
ρmN (1,0)
hpq1
N (1,0)
hrq2
− ρmN (2,0)
rpq1q2
(cid:69)(cid:90)
L
∂ U M
p
∂xq1
∂ U M
r
∂xq2
dx+
+ ε2s3(cid:104)− Gm
hkijB(2,0)
hkpq1q2
∂ U M
p
∂xq1
∂ U M
r
∂xq2
dx + ε2s4(cid:104)ρmN (2,2)
rp
(cid:105)
L
U M
p
U M
r dx+
+ ε2s5 1
2
(cid:104) Gm
hkijB(2,2)
ijr (cid:105)
hkp B(2,2)
U M
r dx +
s(cid:104) Gm
hkijB(1,0)
hkpq1
1
2
B(1,0)
ijrw1
(cid:105)
∂ U M
p
∂xq1
∂ U M
r
∂xw1
dx+
L
+ ε2s
Gm
hkijB(2,0)
hkpq1q2
B(2,0)
ijrw1w2
∂2 U M
p
∂xq1 ∂xq2
∂2 U M
r
∂xw1 ∂xw2
dx −
L
L
U M
h
bhdx.
In accordance with the procedure proposed earlier, the Euler-Lagrangian equation deriving from first varia-
tion of the transformed energy-like functional (103) in the Laplace domain is
s2ρ U M
t + s2ρ( Itpq1 − Itq1p)
∂ U M
p
∂xq1
− s2ρ Itq2pq1
∂2 U M
p
∂xq1 ∂xq2
+ s4ρ I (cid:93)
tp
U M
p =
= −s3( Jtpq1 − Jtq1p)
∂ U M
p
∂xq1
− s3 J 2
tq2pq1
∂2 U M
p
∂xq1∂xq2
+
− s5 J (cid:93)
tp
U M
p + s Gtr1pq1
∂2 U M
p
∂xq1 ∂xr1
+ s( Ytr1pq1r2 − Ytr1r2pq1 )
∂3 U M
p
∂xr1 ∂xr2 ∂xq1
+
18
− s S2
tr1r2pq1q2
∂4 U M
p
∂xq1∂xq2 ∂xr1 ∂xr2
+ bt,
(104)
where the overall inertial tensors are (94),(95), (97) and (100), whereas the overall constitutive tensors are
(86), (87), (90), (92), respectively, and
(cid:105),
= ε2(cid:104) Gm
B(2,0)
(105)
S2
tr1r2pq1q2
hkijB(2,0)
hkpq1q2
hktr1r2
= −ε2(cid:104) Gm
pq1tq2
hkpq1q2
= J 2
B(2,2)
hkijB(2,0)
ijp + Gm
J 2
(106)
tq2pq1
By applying the inverse Laplace transform L−1 to Eq. (85) and Eq. (104), the field equation at the macro-
scale corresponding to Eq. (4) is recast in the time domain as
tp ∗ ....
U
hkijB(2,0)
t + ρ(Itpq1 − Itq1p) ∗ ∂ U M
p
∂xq1
− ρItq2pq1 ∗ ∂2 U M
∂xq1∂xq2
B(2,2)
+ ρI (cid:93)
M
p =
ρ U M
hktq2q1
ijt
p
(cid:105).
= −( Jtpq1 − Jtq1p) ∗ ∂ U M
p
∂xq1
− J i
tq2pq1
∗ ∂2 U M
∂xq1∂xq2
p
+
− J (cid:93)
tp ∗ ....
U
M
p + Gtr1pq1 ∗ ∂2 U M
∂xq1 ∂xr1
p
+ (Ytr1pq1r2 − Ytr1r2pq1 ) ∗
∂3 U M
p
∂xr1 ∂xr2 ∂xq1
+
∂4 U M
p
∗
tr1r2pq1q2
∂xq1∂xq2 ∂xr1∂xr2
− Si
(107)
where the superscript i = 1, 2 and the symbol ∗ stands for the convolution and the time derivative can be
moved from the constitutive tensor to the variable.
The constitutive tensor components in the time domain are
+ bt,
Gm
tr1pq1
= L−1( Gm
tr1pq1
),
(Ytr1pq1r2 − Ytr1r2pq1 ) = L−1( Ytr1pq1r2 − Ytr1r2pq1 ),
Si
tr1r2pq1q2
= L−1( Si
tr1r2pq1q2
),
Jtpq1 − Jtq1p = L−1(s( Jtpq1 − Jtq1p)),
(108)
(109)
(110)
(111)
= L−1(s J i
),
tq2pq1
tp),
(112)
J i
tq2pq1
tp = L−1(s J (cid:93)
J (cid:93)
(113)
with i = {1, 2}, whereas in the time domain the inertial tensor components and the acceleration result to be
(Itpq1 − Itq1p) = L−1( Itpq1 − Itq1p),
Itq2pq1 = L−1( Itq2pq1),
tp = L−1( I (cid:93)
I (cid:93)
t = L−1(s2 U M
U M
t ).
In case of a locally homogeneous material, i.e. if the miscrostructure disappears, the perturbation functions
N (1,0)
iklpq are zero and the components of the localization tensors defined in (80)-(83) vanish
except for B(1,0)
and so the equation of motion of a classical
hkpq1
homogeneous continuum is retrieved.
, which becomes B(1,0)
hkpq1
iklp ,...., N (3,0)
δhpδkq1 + δhq1δkp
, N (2,0)
(117)
(114)
(116)
(115)
= 1
2
(cid:17)
(cid:16)
tp),
ikl
19
6.1 Dispersive wave propagation
In this Subsection, the Laplace and the Fourier transforms are applied to Eq. (107) with respect to time t
and to the slow variable x to obtain the field equation at the macroscale within the frequency and the wave
vector domain. In particular the two-sided Fourier transform of an arbitrary function f is defined as, Paley
and Wiener [1934]:
F(f (x)) = f (k) =
f (x)eιk·xdx =
f (x)eιksxsdx, k ∈ R2,
(118)
(cid:90) +∞
−∞
(cid:90) +∞
−∞
where k is a bidimensional vector and so the field equation at the macroscale in the transformed space is
rephrased as
s2ρ
U M
t + s2ρι( Itpq1 + Itq1p)
U M
p kq1 + s2ρ Itq2pq1
U M
p kq1 kq2 + s4ρ I (cid:93)
tp
U M
p =
− s3ι( Jtpq1 − Jtq1p)
p kq1 − s3 Jtq2pq1
U M
U M
p kq1kq2 +
− s5 J (cid:93)
tp
p − s Gtr1pq1
U M
p kq1 kr2 − sι( Ytr1pq1r2 − Ytr1r2pq1)
U M
U M
p kr1kr2 kq1+
− s Si
The vector k is written respect to n as k = kn (k = k and n = 1), therefore the Eq. (119) becomes
U M
p kq1kq2 kr1 kr2 +
i = {1, 2}.
tr1r2pq1q2
bt,
(119)
s2ρ
U M
t + s2ρι( Itpq1 + Itq1p)
U M
p knq1 + s2ρ Itq2pq1
U M
p k2nq1nq2 + s4ρ I (cid:93)
tp
U M
p =
= −s3ι( Jtpq1 − Jtq1p)
p knq1 − s3 J i
U M
tq2pq1
p k2nq1nq2 − s5 J (cid:93)
U M
tp
U M
p +
− s Gtr1pq1
p k2nq1nr2 − sι( Ytr1pq1r2 − Ytr1r2pq1)
U M
U M
p k3nr1nr2 nq1+
− s Si
tr1r2pq1q2
U M
p k4nq1nq2nr1 nr2 +
bt,
i = {1, 2}.
(120)
In case of an orthotropic material the motion equation in the Laplace domain (104) along the direction eβ
(β = 1, 2) is rephased as
s2ρ U M
α − s2ρ Iαβαβ
∂2 Uα
∂x2
β
+ s4ρ Iαα U M
α =
(121)
= s3 J i
αβαβ
∂2 U M
α
∂x2
β
− s5 Jαα U M
α + s Gαβαβ
∂2 U M
α
∂x2
β
− s Si
αββαββ
∂4 U M
α
∂x4
β
,
with i = {1, 2} and α = 1, 2. The Fourier transform (118) is applied to Eq. (121) with respect to the slow
variable x to retrieve the Christoffel equation depending on the complex angular frequency s and the wave
vector kβ,
(s2ρ + s2ρ Iαβαβk2
β + s4ρ Iαα + s3 J i
αβαβk2
β + s5 Jαα + s Gαβαβk2
β + s Si
αββαββk4
β)
U M
α (kβ, s) = 0,
i = {1, 2},
(122)
U (kβ, s) stands for the Fourier transform of the transformed macro-displacement U (x, s). The Christof-
where
fel equation (122) defines the wave propagation in the viscoelastic medium that is embedded in the Laplace-
Fourier space. The dispersion function stemmed from Eq. (122) is
β + s Si
β + s5 Jαα + s Gαβαβk2
β + s4ρ Iαα + s3 J i
s2ρ + s2ρ Iαβαβk2
i = {1, 2}.
αββαββk4
αβαβk2
β = 0,
(123)
The dispersion function describes the longitudinal and the transverse oscillatory motion of the viscoelastic
homogeneous continuum.
20
7 Homogenization of a bi-phase layered material
The model proposed in Section 6 is herein applied to a domain made of two layered materials, which have
thickness s1 and s2, and subject to L-periodic body forces b(x). The domain displays orthotropic phases and
the orthotropic axis is supposed to be parallel to the direction e1, see Fig. 2. In case of isotropic phases, for
the plane-stress state we have E = E and ν = ν, whereas for the plane-strain state E = E
1−ν ,
where E is the Young's modulus and ν is the Poisson's ratio.
For sake of simplicity but without loss of generality, the components of the viscoelastic tensor are
1−ν2 and ν = ν
Gi
1111 = Gi
2222 = Gi,∞
1111(e
− t
τr + 1), Gi
1122 = Gi,∞
1122(e
− t
τr + 1),
− t
1212(e
τr + 1),
1111 = E
i ∈ {1, 2},
1−ν2 , Gi,∞
1212 = Gi,∞
Gi
where Gi,∞
1111 =
Gi,∞
2222. It can be noticed that the viscoelastic tensor can be deemed as a term of the Prony series [Ferry,
1980] and so the proposed homogenization technique can be applied to any kind of sufficiently regular kernel,
since the overall constitutive and inertial tensor components have a general structure.
In addition, τr stands for the relaxation time and the superscript i represents either phase 1 or the phase 2.
The Laplace transform (8) applied to the components of the viscoelastic tensor (124) leads to
2(1+ν) are the equilibrium elastic modulus with Gi,∞
1−ν2 and Gi,∞
1122 = E ν
1212 =
(124)
E
Gi
1111 = Gi
2222 =
E
1 − ν2
2τrs + 1
s(sτr + 1)
,
Gi
1122 =
E ν
1 − ν2
2τrs + 1
s(sτr + 1)
,
E
Gi
1212 =
2τrs + 1
s(sτr + 1)
2(1 + ν)
Finally the relation Cm(cid:16) x
,
(cid:17)
i ∈ {1, 2}.
= s Gm(cid:16) x
ε , s
(cid:17)
ε , s
provides the viscoelastic tensor components as
(125)
C i
1111 = Gi
2222 =
E
1 − ν2
2τrs + 1
(sτr + 1)
,
C i
1122 =
E ν
1 − ν2
2τrs + 1
(sτr + 1)
,
Figure 2: Heterogeneous bidimensional domain with its layered periodic cell.
21
C i
1212 =
E
2(1 + ν)
2τrs + 1
(sτr + 1)
,
i ∈ {1, 2}.
(126)
The transformed components of the viscoelastic tensor (126) are employed to determine the perturbation
functions of first, second and third order, obtained by solving the cell problems (33), (36), (42), (see Appendix
C for the structure of the perturbation functions).
The following dimensionless quantities are introduced
(cid:115)
rρ =
ρ1
ρ2
,
rE =
E1
E2
,
τς =
τr
s2
E2
ρ2
,
η =
s1
s2
,
22
22
22
ρ2
) E2
) E2
E2
22
E2
22
E2
22
) E2
and its imaginary part Im(N (2,2)
are taken into account since s is a complex number.
with respect to the imaginary part of s, Im(s), and ξ2.
along the periodic cell vs. the real part of s, Re(s), and the vertical
is shown in Fig. 3-(c) vs. Re(s) and in Fig. 3-(d) vs. Im(s), by varying the
where rρ stands for the ratio between the densities, rE is the ratio between the Young's moduli related to
phase 1 and phase 2, τς is the relaxation time and η is the ratio between the thicknesses of the layers s1 and
s2.
The perturbation function N (2,2)
is analytically computed by solving the cell problem (39) supplied with
the interface conditions (40) in Sec. 4, with respect to the phase 1 and the phase 2, and the structure of
N (2,2)
is reported in terms of the geometric and mechanical properties of the periodic domain in Eq. (165).
Such function depends on the fast variable ξ2, which is perpendicular to the transversal direction e1, as well
as on the complex parameter s. The perturbation function is non-dimensionalized as N (2,2)
and its real
ρ2
part Re(N (2,2)
22
ρ2
Fig. 3-(a) shows the real part of N (2,2)
ρ2
coordinate ξ2.
Fig. 3-(b) depicts the dependence of Re(N (2,2)
22
ρ2
The imaginary part of N (2,2)
ρ2
vertical coordinate ξ2.
The cell problem (41) equipped with the interface conditions (42) provides the perturbation function N (3,2)
222
and its formulation is explicitely expressed in Eq. (193), where there is point in noticing they take into ac-
count of the effect of the microstructural heterogeneities of the domain. The real part and the imaginary part
of the dimensionless perturbation function N (3,2)
222 ) E2
222
ρ2
ρ2
Fig. 4-(a) and Fig. 4-(b) show how the real part of N (3,2)
222
ρ2
by varying the vertical coordinate ξ2.
Fig. 4-(c) and Fig. 4-(d) show that the imaginary part of N (3,2)
222
ρ2
vertical coordinate ξ2.
In particular, the poles of dimensionless perturbation functions are emphasized in Fig. 3-a, Fig. 3-c, Fig. 4-a
and Fig. 4-c. In all the previous figures, it is straightforward that the dimensionless perturbation functions
are Q-periodic, they have vanishing mean values
Re(N (2,2)
over the unit cell Q and they are smooth along the boundaries of Q, as expected.
The Poisson ratios are assumed to be equal for both phases ν1 = ν2 = 0.2 and the dimensionless relaxation
time is τς = 10. In addition, the dimensionless Young's modulus is rE = 10, the density is rρ = 10 and the
ratio between the thicknesses is η = 1.
1111 and I11, referred to the wave propagation along the
The transformed inertial tensor components J11, J 2
transversal direction e1, are taken into account. They are affected by the complex angular frequency s and
are considered, which are Re(N (3,2)
222 ) E2
ρ2
depends on the real and imaginary parts of s
depends on Re(s) and Im(s) and the
and Im(N (3,2)
, Re(N (3,2)
222 ) E2
ρ2
and Im(N (3,2)
222 ) E2
ρ2
) E2
, Im(N (2,2)
22
ρ2
) E2
22
ρ2
E2
E2
E2
.
their dimensionless components are written as
be negative and therefore the poles of the dimensionless components are visible in Figs. 5, 6 and 7.
. The Re(s)-axis is assumed to
,
J 2
1111
ε2s2ρ2
and I11 E2
ε2ρ2
The second order transformed inertial tensor component
is computed by means of Eq. (92) and it
(cid:114) E2
J11
ρ2
ε2s2ρ2
(cid:114) E2
(cid:114) E2
ρ2
J11
ρ2
ε2s2ρ2
is decomposed into its real part
complex frequency s.
Fig. 5-(a) shows the behaviour of
Re( J11)
ε2s2ρ2
(cid:114) E2
(cid:114) E2
ρ2
ρ2
ε2s2ρ2
Re( J11)
(cid:114) E2
ρ2
Im( J11)
ε2s2ρ2
and its imaginary part
, due to the presence of the
by varying the real part and the imaginary part of s.
In
22
Figure 3: (a) Dimensionless real part Re( N (2,2)
(b) Dimensionless real part Re( N (2,2)
imaginary part Im( N (2,2)
imaginary part Im( N (2,2)
τς = 10, ν1 = ν2 = 0.2 and η = 1.
) E2/ρ2 vs. the real part of s, Re(s), and the coordinate ξ2.
) E2/ρ2 vs. the imaginary part of s, Im(s), and ξ2. (c) Dimensionless
) E2/ρ2 vs. the real part of s, Re(s), and the coordinate ξ2. (d) Dimensionless
) E2/ρ2 vs. the imaginary part of s, Im(s), and ξ2 obtained for rρ = 10, rE = 10,
22
22
22
22
23
Figure 4: (a) Dimensionless real part of the perturbation function N (3,2)
of s, Re(s), and the coordinate ξ2. (b) Dimensionless real part Re( N (3,2)
s, Im(s), and ξ2. (c) Dimensionless imaginary part Re( N (3,2)
imaginary part Re( N (3,2)
and η = 1.
222 ) E2/ρ2, vs. the real part
222 ) E2/ρ2 vs. the imaginary part of
222 ) E2/ρ2 vs. Re(s), and ξ2.(d) Dimensionless
222 ) E2/ρ2 vs. Re(s), and ξ2, obtained for rρ = 10, rE = 10, τς = 10, ν1 = ν2 = 0.2
222 , Re( N (3,2)
24
Figure 5: (a) Dimensionless real part component Re( J11)
imaginary part of s, Im(s). (b) Dimensionless imaginary part component Im( J 2
/ε2s2ρ2 vs. the real
part of s, Re(s), and the imaginary part of s, Im(s), with rρ = 10, rE = 10, τς = 10, ν1 = ν2 = 0.2 and
η = 1.
/ε2s2ρ2 vs. the real part of s, Re(s), and the
11)
ρ2
ρ2
(cid:113) E2
(cid:113) E2
(cid:114) E2
J11
ρ2
ε2s2ρ2
(cid:114) E2
ρ2
Im( J11)
particular it can be observed that the real part of
is symmetric with respect to the Re(s)-axis.
ε2s2ρ2
is plotted with respect to the real part of s, Re(s), and the
In Fig. 5-(b) the imaginary part
imaginary part of s, Im(s). The function assumes either positive or negative values and it is symmetric
respect to zero.
Eq. (98) allows computing the transformed second order inertial tensor component I11 E2
ε2ρ2
posed into its real part Re( I11) E2
ε2ρ2
and Im( I11) E2
The dependency of Re( I11) E2
6-(a) and 6-(b). The symmetry of Re( I11) E2
are shown.
Finally, in Figs. 7-(a) and 7-(b), the real part of the dimensionless fourth order inertial tensor component
on the real part and the imaginary part of s is shown in Figs.
with respect to Re(s)-axis and of Im( I11) E2
and its imaginary part Im( I11) E2
with respect to zero
, which is decom-
ε2ρ2
ε2ρ2
ε2ρ2
ε2ρ2
ε2ρ2
.
(cid:114) E2
ρ2
Re( J 2
1111)
ε2s2ρ2
(cid:114) E2
ρ2
Im( J 2
1111)
ε2s2ρ2
and its imaginary part
complex frequency s. It is worth to noticing that the inertial tensor component J 1
approach 1 is vanishing.
are plotted vs. the real and the imaginary parts of the
1111 stemming from the
Fig. 8-(a) shows that the magnitudes of the constitutive tensor components S1
111111 decreases more steadily than S2
111111, deter-
mined for the compressional waves along the direction e1, are compared in terms of the magnitude of s and
the non-dimensional Young's modulus rE. S1
10 < rE < 1,
and both are equal to zero when the material is homogeneous, i.e. when rE = 1. With increasing the di-
111111. Finally
mensionless Young's modulus rE, a significant rise in S2
111111 decreases more slowly than S2
as s goes up, S1
In Fig. 8-(b) the magnitudes of S1
211211 are plotted with respect to the magnitude of the complex
angular frequency s and the non-dimensional Young's modulus rE. They are computed for the shear waves
travelling along the layering direction e1. It can be observed that if 1
and S2
S1
211211 and S2
S2
211211 to decrease.
211211
211211 decrease down to zero when the material is homogeneous. A rise in rE (with rE > 1) makes
211211 and
211211 increasing rapidly. Moreover, a growth in the magnitude of s leads S1
111111 is observed as compared to S1
111111.
111111 and S2
111111, as 1
10 < rE < 1, the magnitudes of S1
211211 and S2
25
Figure 6: (a) Dimensionless real part component Re( I11) E2/ε2s2 vs. the real part of s, Re(s), and the
imaginary part of s, Im(s). (b) Dimensionless component Im( I11) E2/ε2s2 vs. the real part of s, Re(s), and
the imaginary part of s, Im(s), obtained for rρ = 10, rE = 10, τς = 10, ν1 = ν2 = 0.2 and η = 1.
Fig. 8-(c) depicts the behaviour of the magnitudes of the transformed viscoelastic component G1212 related
to the shear wave travelling along e2. It is observed that the trend of G1212 steadely increases by a rise in
the Young's modulus and by varying the magnitude of the complex frequency s.
Fig. 8-(d) shows the magnitude of the transformend viscoelastic component G1111 concerning with the com-
pressional wave along e1. G1111 significantly grows up in the interval 1
10 < rE < 4 by varying s and then
its trend becomes a plateau.
Figure 7: (a) Dimensionless real part component Re( J 2
1111)
(cid:113) E2
ρ2
the imaginary part of s, Im(s). (a) Dimensionless imaginary part component Im( J 2
/ε2s2ρ2 vs. the
real part of s, Re(s), and the imaginary part of s, Im(s). Both are retrieved for rρ = 10, rE = 10, τς = 10,
ν1 = ν2 = 0.2 and η = 1.
1111)
ρ2
/ε2s2ρ2 vs. the real part of s, Re(s), and
(cid:113) E2
26
111111 (violet) by
Figure 8: (a) The magnitude of the constitutive tensor S1
211211
varying the dimensionless Young's modulus rE and the magnitude of s. (b) S1
(green) are compared with respect to rE and s.(c) G1212 is depicted by varying the Young's modulus rE
and the magnitude of s. (d) G1111 is represented by changing rE and the magnitude of s. The graphs are
obtained for rρ = 10, τς = 10, ν1 = ν2 = 0.2, η = 1 and the argument of the s is θ = 0.
111111 (red) is compared with S2
211211 (gold) and S2
27
7.1 Benchmark test: homogenized problem vs. heterogeneous material
A horizontal (or vertical) sample of length L = L1(= L2) is taken into account since the body forces and
the heterogeneous domain are periodic. The solution of the heterogeneous model is provided by a numerical
procedure, accounting for the actual heterogeneous composition of the layered composite, and it is compared
with the solution deriving from the homogenized model. The field equation (121) is written as
∂2 U M
α
∂x2
β
− s Si
∂4 U M
α
∂x4
β
= −bα,
αββαββ
s Gαβαβ
where i = {1, 2} and the indices α and β are not summed. The shear problem takes place for α (cid:54)= β whereas
the compressional problem for α = β. The transformed macro-displacement is determined from Eq. (127)
(127)
(cid:16) Lβ
(cid:17)2
U M
α (xβ) =
2π
s Gαβαβ
1 +
(cid:104)
(cid:17)2 Si
bα
(cid:16) 2πk
Lβ
αββαββ
Gαβαβ
(cid:105) ,
(128)
(129)
with i = 1, 2 and k = 1, 2, .. The dimensionless transformed macro-displacement is written as
U M
α =
U M
α
E1
(cid:16) Lβ
(cid:17)2 ,
Υα
2π
where E1 is the Young's modulus related to the phase 1. A FE method provides the macro-displacements
U M
α , computed for each individual unit cell composing the sample of length L/ε = 11, with amplitute Υα = 1
N/mm3, E1 = 10000 MPa, E2 = 1000 MPa, ν1 = ν2 = 0.2, τς = 5, η = 1 and rρ = 10.
The homogenized and the heterogeneous problems are supposed to be in a plane stress state. In Fig. 9-(a)
the magnitude of the dimensionless transformed macro-displacement (129) with α = 1, obtained by solving
analytically the field equation (127) along the direction x1, is compared with the solution provided by a finite
element analysis of the heterogeneous domain equipped with proper periodic boundary conditions on the
displacement. The continuous curve stands for the analytical dimensionless transformed macro-displacement
U M
1 (x1), which is derived from the method 1 providing the overall constitutive tensor S1
111111, Eq. (234).
The dotted curve represents the analytical dimensionless transformed macro- displacement
U M
1 (x1) and
derives from the alternative method 2, which provides the overall constitutive tensor S2
111111, Eq. (240).
Finally, the diamonds stand for the numerical results related to the heterogeneous model and obtained from
the corresponding microscopic solution through the up-scaling relation (55) and considering the imaginary
part of the body force b1 = Υ1 sin 2πk
L1
x1, with k = 1.
ρ2
All the methods are compared in Fig. 9-(a) and three values for the dimensionless parameter sς = ss2/
are considered. The red curves and diamonds are obtained for sς = −2, the blue curves and diamonds are
given for sς = −0.5 and finally the green ones stand for sς = −0.3. By increasing the values of sς a gradual
In Fig. 9-(b) the magnitude of the analytical
reduction in the behaviour of the solution is emphasized.
U M
2 (x1), which is obtained from the method 1 (continuous
dimensionless transformed macro-displacement
curve), and the transformed macro-displacement stemming from the method 2 (dotted curve) are put in
relation with the heterogeneous solution (diamonds) for three increasing values of sς , which are sς = −2 (red
one), sς = −0.5 (blue one) and sς = −0.3 (green one) respectively. The dotted and the continuous curves
are almost coincident.
Both for the compressional and shear problems 9-(a) and 9-(b), there is a very agreement deal between the
solution of both homogenized models and the numerical solution of the heterogeneous approach in the static
case. Along the orthotropic direction x2, the
122122 and Si
overall constitutive tensors Si
222222, with i = 1, 2, are equal to zero and so the transformed
, with α = 1, 2. Therefore, the
macro-displacement (128) assumes the form U M
(cid:17)2 Υα sin 2π
(cid:16) L2
x2
α (x2) =
2π
L2
s Gα2α2
U M
1 (x2) (Fig. 9-(c)) and the magnitude of the
magnitude of the dimensionless macro-displacement field
dimensionless macro- displacement U M
2 (x2) (Fig. 9-(d)), represented by continuous curves, are compared
with the numerical macro-displacements associated with the heterogeneous model (diamonds), by varying
three values of the dimensionless parameter sς , which are sς = −2 (red one), sς = −0.5 (blue one) and
sς = −0.3 (green one) respectively.
(cid:113) E2
28
U M
Figure 9: (a) Magnitude of the dimensionless macro-displacement component
1 (x1) induced by the harmonic
2 (x1) caused by b2(x1) along direction x1.(c) Magnitude of the component
body force b1(x1) and (b) U M
U M
1 (x2)
2 (x2) produced by b2(x2) along x2. The solution given by the homogenized model
induced by b1(x2) and (d) U M
1 (continuous curves) and the homogenized model 2 (dotted curves) are compared with the solution given by the
heterogeneous one (diamonds). The red curves are given for the dimensionless parameter sς = −2, the blue curves
for sς = −0.5 and the green ones for sς = −0.3.
29
7.2 Dispersion curves for layered materials
The dispersion relation (123) is an implicit function depending on the real part of s, Re(s), its imaginary
part, Im(s), and the wave number kβ, β = 1, 2, which have been non-dimensionalized as Re(s)s2(cid:114) E2
ρ2
, Im(s)s2(cid:114) E2
ρ2
and kβε. Eq. (123) provides the dispersion curves related to the dynamic homogenization model and its
simplified version presented in Sec. (6).
The dispersion curves obtained with two approaches are compared in Fig. 10 and Fig. 11. The curves
given by the dynamic homogenization model are plotted with a thicker line than the ones obtained with the
simplified version and it is observed that the curves produced from both approaches are pretty coincident
in the range of the considered wave-number. The dual nature of a viscoelastic material is reflected by the
presence of the complex parameter s. Its real part is related to the viscosity of the material whereas its
imaginary part deals with its elastic behaviour. Therefore, the dispersion curves lay both on the real axis of
s, which describes the wave propagation, and its imaginary axis, which characterizes the damping.
Fig. 10 and Fig. 11 illustrate the variation of the dimensionless real part of s in terms of the dimensionless
imaginary part of s and the dimensionless wave number. In case of dispersion curves that travel along the
axis eβ, the wave number kβε belongs to the
interval [0, π], with β = 1, 2.
Fig. 10-(a) shows the curves obtained for three values of the Young's modulus. The orange one corresponds
to rE = 20, the blue one to rE = 15 and the green one to rE = 10. The values for τς = 2, η = 1 and rρ = 10
are supposed to be fixed. The curves are monotonically increasing and as rE goes down they gradually
decline.
Fig. 10-(b) shows curves obtained for three values of the relaxation time τς . For low values of τς , namely
τς = 1, which is represented by the orange curve, and τς = 2, identified by the blue curve, the viscoelasticity
strongly affects the dispersion curves, which are enterely embedded in the real and the imaginary plane. By
increasing the relaxation time τς = ∞, the elastic behaviour is retrieved (green curve), infact the curve is
squeezed in the imaginary axes of the complex frequency.
Fig. 10-(c) highlights the trends of three curves developped by setting three different values of the dimen-
sionless ratio between the thicknesses of the material. The blue curve is obtained for η = 1, the orange one
for η = 2 and the green one for η = 6, with fixed values of the dimensionless parameters rE = 10, τς = 2
and rρ = 10. A rise in the ratio implies that the corresponding curves become more flatten.
Finally, Fig. 10-(d) represents how for three increasing values of the dimensionless density the related curves
are influenced. The green curve is given with rρ = 5, the blue one with rρ = 10 and the orange one corre-
sponds to rρ = 50, by fixing rE = 10, τς = 2 and η = 1.
In Fig. 11 dispersion curves that travel along the axis e2 are taken into account. In Fig. 11 the curves
achieved from approach 1 and approach 2 are compared.
In Fig. 11-(a), the values of the dimensionless parameters are set as τς = 2, η = 1 and rρ = 10. The
orange curve obtained for rE = 100 and the blue one for rE = 3 have a similar trend. By considering a
lower dimensionless Young's modulus rE = 0.1, the green curve is recovered, which is strongly overwhelmed
with respect to the previous ones. Fig. 11-(b) shows that by increasing the relaxation time τς up to high
values (τς = ∞) the green curve is achieved and so the viscoelastic effect is negligible. As the relaxation
time assumes low values (τς = 1 and τς = 7) the corresponding curves (the orange and the blue one) are
strongly influenced by the viscoelastic response. The dimensionless parameters are set as rE = 10, τς = 2
and rρ = 10.
In Fig. 11-(c) by setting three values for the dimensionless ratio between the thicknesses of the material,
the following curves are accomplished: the blue curve is stemmed for η = 1, the orange one for η = 2 and
the green one for η = 6, with fixed values of the non-dimensional parameters rE = 10, τς = 2 and rρ = 10.
As the ratio increases the curves become more and more flatten. Fig. 11-(d) shows the curves obtained for
three values of the dimensionless density. The green one relates to rρ = 5, the blue curve corresponds with
the value rρ = 10 and the orange one rρ = 50, with the constant parameters rE = 10, τς = 2 and η = 1.
The diagram emphasises that the higher the density is, the more sqeezed the curve is.
30
Figure 10: dispersion curves along the direction e1 derived with two approaches (a) by varying the dimen-
sionless Young's modulus: rE = 20 (orange), rE = 15 (blue) and rE = 10 (green), (b) by modifying the
dimensionless relaxation time: τς = 1 (orange), τς = 2 (blue) and τς = ∞ (green), (c) with different values of
the dimensionless thickness ratio: η = 2 (orange), η = 1 (blue) and η = 6 (green), (d) with different values
of the dimensionless density: rρ = 50 (orange), rρ = 10 (blue) and rρ = 5 (green).
31
Figure 11: dispersion curves along e2 derived with two approaches, (a) by varying the Young's modulus:
rE = 100 (orange), rE = 3 (blue) and rE = 0.1 (green), (b) by modifying the nondimensional relaxation
time: τς = 1 (orange), τς = 7 (blue) and τς = ∞ (green), (c) with three values of the thickness ratio: η = 2
(orange), η = 1 (blue) and η = 6 (green), (d) with different values of the dimensionless density: rρ = 50
(orange), rρ = 10 (blue) and rρ = 5 (green).
32
8 Conclusions
The paper proposed a variational-asymptotic homogenization model for viscoelastic materials having a peri-
odic microstructure. Specifically, the field equations at the micro-scale have been derived and transposed into
the Laplace domain to treat viscoelasticity, relying on the complex frequency and the micro-relaxation tensor.
Next, the down-scaling relation and the up-scaling relation have been detailed. In particular, the down-scaling
relation relates the transformed micro-displacement field to the transformed macro-displacement field and
its gradients by means of the perturbation functions, which are the solutions of the cell problems defined
over the unit cell Q. Perturbation functions are Q-periodic and have vanishing mean values over the unit
cell. On the other hand, the up-scaling relation defines the transformed macro-displacement field as the
mean value of the transformed micro-displacement field over the unit cell Q.
By introducing the down-scaling relation into the field equations at the micro-scale of the viscoelastic ma-
terial, the average field equations of infinite order have been determined. Their truncation at an arbitrary
order could not ensure the ellipticity of the differential problem.
To avoid such an issue, two methods, based on a variational-asymptotic approach, have been invoked. Ac-
cording to the first method, the down-scaling relation is replaced into the transformed energy-like functional,
which is truncated at the second order. By imposing the first variation of the truncated energy-like functional
equal to zero, the field equation at the macro-scale and the overall inertial and constitutive tensors have been
determined. In the second method, the gradient related to the down-scaling relation has been approximated
at the first order and the transformed micro-displacement has been truncated at second order. Both are
introduced into the transformed energy-like functional and, from its first variation, the field equation at the
macro-scale and the corresponding overall inertial and constitutive tensors have been derived. In both meth-
ods, the overall constitutive tensors depend on the localization functions, whereas the overall inertial tensors
are expressed through the perturbation functions. In the limit case of an homogeneous material without
heterogeneities, the perturbation functions consistently become identically equal to zero, the components of
the localization tensors vanish and the equation of motion of the classical continuum has been retrieved.
The proposed model has been applied to a bi-phase layered material with isotropic phases, subject to L-
periodic body forces. In such a case, some of the perturbation functions and some of the components of the
overall inertial tensors have been computed and displayed to remark their dependence on the real part and
on the imaginary part of the complex frequency.
Moreover, the analytical solutions for the transformed macro-displacement field derived from the two homog-
enized models and subjected to harmonic volume forces have been compared with the reference numerical
solution obtained from a finite element analysis of the heterogeneous model, for three different values of the
complex frequency. A good agreement between the three models has been achieved, proving the validity of
the proposed homogenized methods.
Finally, a detailed parametric study of the dispersion function, derived from the generalized Christoffel
equation, allowed determining the behaviour of the dispersion curves through the viscoelastic medium along
two orthogonal axis, for both homogenized methods. An optimal matching between the dispersion curves
derived from the two methods has been pointed out. The variational-asymptotic homogenization method
herein described is helpful to detect the effective viscoelastic properties of many composite materials and it
may be adopted for the manufacture and the design of more efficient and sophisticated devices, for a large
spectra of applications.
References
J. Aboudi, M.-J. Pindera, and S. Arnold. Linear thermoelastic higher-order theory for periodic multiphase materials.
Journal of Applied Mechanics, 68(5):697 -- 707, 2001.
D. Addessi, M. L. De Bellis, and E. Sacco. Micromechanical analysis of heterogeneous materials subjected to overall
cosserat strains. Mechanics Research Communications, 54:27 -- 34, 2013.
G. Allaire. Homogenization and two-scale convergence. SIAM Journal of Mathematical Analisys, 23:1482 -- 1518, 1992.
I. V. Andrianov, V. I. Bolshakov, V. V. Danishevs' kyy, and D. Weichert. Higher order asymptotic homogenization
and wave propagation in periodic composite materials. Proceedings of the Royal Society of London A: Mathematical,
Physical and Engineering Sciences, 464(2093):1181 -- 1201, 2008.
33
M. Bacca, F. Dal Corso, D. Veber, and D. Bigoni. Mindlin second-gradient elastic properties from dilute two-
phase cauchy-elastic composites part i: closed form expression for the effective higher-order constitutive tensor.
International Journal of Solids and Structures, 50(24):4010 -- 4019, 2013a.
M. Bacca, F. Dal Corso, D. Veber, and D. Bigoni. Anisotropic effective higher-order response of heterogeneous cauchy
elastic materials. Mechanics Research Communications, 54:63 -- 71, 2013b.
M. Bacca, F. Dal Corso, D. Veber, and D. Bigoni. Mindlin second-gradient elastic properties from dilute two-phase
cauchy-elastic composites part ii: Higher-order constitutive properties and application cases. International Journal
of Solids and Structures, 50(24):4020 -- 4029, 2013c.
A. Bacigalupo. Second-order homogenization of periodic materials based on asymptotic approximation of the strain
energy: formulation and validity limits. Meccanica, 49(6):1407 -- 1425, 2014.
A. Bacigalupo and L. Gambarotta. Second-order computational homogenization of heterogeneous materials with
periodic microstructure. ZAMM -- Journal of Applied Mathematics and Mechanics/Zeitschrift fr Angewandte Math-
ematik und Mechanik, 90:796 -- 811, 2010.
A. Bacigalupo and L. Gambarotta. A multi-scale strain-localization analysis of a layered strip with debonding
interfaces. International Journal of Solids and Structures, 50(13):2061 -- 2077, 2013.
A. Bacigalupo and L. Gambarotta. Second-gradient homogenized model for wave propagation in heterogeneous
periodic media. International Journal of Solids and Structures, 51(5):1052 -- 1065, 2014a.
A. Bacigalupo and L. Gambarotta. Computational dynamic homogenization for the analysis of dispersive waves in
layered rock masses with periodic fractures. Computers and Geotechnics, 56:61 -- 68, 2014b.
A. Bacigalupo, L. Morini, and A. Piccolroaz. Effective elastic properties of planar sofcs: A non-local dynamic
homogenization approach. International Journal of Hydrogen Energy, 39(27):15017 -- 15030, 2014.
A. Bacigalupo, L. Morini, and A. Piccolroaz. Overall thermomechanical properties of layered materials for energy
devices applications. Composite Structures, 157:366 -- 385, 2016a.
A. Bacigalupo, L. Morini, and A. Piccolroaz. Multiscale asymptotic homogenization analysis of thermo-diffusive
composite materials. International Journal of Solids and Structures, 85:15 -- 33, 2016b.
A. Bacigalupo, M. Paggi, F. Dal Corso, and D. Bigoni. Identification of higher-order continua equivalent to a cauchy
elastic composite. Mechanics Research Communications, 2017. doi: https://doi.org/10.1016/j.mechrescom.2017.
07.002.
N. Bakhvalov and G. Panasenko. Homogenization: Averaging Processes in Periodic Media. Kluwer Academic Pub-
lishers, Dordrecht-Boston-London, 1984.
A. Bensoussan, J.-L. Lions, and G. Papanicolaou. Asymptotic analysis for periodic structures. North-Holland,
Amsterdam, 1978.
D. Bigoni and W. Drugan. Analytical derivation of cosserat moduli via homogenization of heterogeneous elastic
materials. Journal of Applied Mechanics, 74(4):741 -- 753, 2007.
C. Boutin. Microstructural effects in elastic composites. International Journal of Solids and Structures, 33:1023 -- 1051,
1996.
R. Christensen. Theory of viscoelasticity: an introduction. Elsevier, 2012.
M. L. De Bellis and D. Addessi. A cosserat based multi-scale model for masonry structures. International Journal
for Multiscale Computational Engineering, 9(5):543, 2011.
H. N. Dhakal, S. O. Ismail, S. O. Ojo, M. Paggi, and J. R. Smith. Abrasive water jet drilling of advanced sustainable
bio-fibre-reinforced polymer/hybrid composites: a comprehensive analysis of machining-induced damage responses.
The International Journal of Advanced Manufacturing Technology, pages 1 -- 15, 2018.
M. Fabrizio and A. Morro. Mathematical problems in linear viscoelasticity, volume 12. Siam, 1992.
F. Fantoni, A. Bacigalupo, and M. Paggi. Multi-field asymptotic homogenization of thermo-piezoelectric materials
with periodic microstructure. International Journal of Solids and Structures, 120:31 -- 56, 2017.
34
J. D. Ferry. Viscoelastic properties of polymers. John Wiley & Sons, 1980.
F. Feyel. A multilevel finite element method (fe2) to describe the response of highly non-linear structures using
generalized continua. Computer Methods in applied Mechanics and engineering, 192(28-30):3233 -- 3244, 2003.
J. Fish and W. Chen. Higher-order homogenization of initial/boundary-value problem. Journal of engineering
mechanics, 127(12):1223 -- 1230, 2001.
S. Forest. Homogenization methods and mechanics of generalized continua-part 2. Theoretical and applied mechanics,
(28-29):113 -- 144, 2002.
S. Forest and K. Sab. Cosserat overall modeling of heterogeneous materials. Mechanics Research Communications,
25(4):449 -- 454, 1998.
S. Forest and D. K. Trinh. Generalized continua and non-homogeneous boundary conditions in homogenisation
methods. ZAMM-Journal of Applied Mathematics and Mechanics/Zeitschrift fur Angewandte Mathematik und
Mechanik, 91(2):90 -- 109, 2011.
B. Gambin and E. Kroner. Higher order terms in the homogenized stressstrain relation of periodic elastic media.
physica status solidi (b). International Journal of Engineering Science, 151(2):513 -- 519, 1989.
R. F. Gibson. Principles of composite material mechanics. CRC press, 2011.
G. Haasemann and V. Ulbricht. Numerical evaluation of the viscoelastic and viscoplastic behavior of composites.
Technische mechanik, 30(1-4):122 -- 135, 2010.
T. Hui and C. Oskay. A nonlocal homogenization model for wave dispersion in dissipative composite materials.
International Journal of Solids and Structures, 50(1):38 -- 48, 2013.
(cid:32)L. Kaczmarczyk, C. J. Pearce, and N. Bi´cani´c. Scale transition and enforcement of rve boundary conditions in
second-order computational homogenization. International Journal for Numerical Methods in Engineering, 74(3):
506 -- 522, 2008.
P. Kanout´e, D. Boso, J. Chaboche, and B. Schrefler. Multiscale methods for composites: a review. Archives of
Computational Methods in Engineering, 16(1):31 -- 75, 2009.
V. Kouznetsova, M. Geers, and W. Brekelmans. Advanced constitutive modeling of heterogeneous materials with a
gradient-enhanced computational homogenization scheme. International Journal for Numerical Methods in Engi-
neering, 54:1235 -- 1260, 2002.
V. Kouznetsova, M. Geers, and W. Brekelmans. Multi-scale second-order computational homogenization of multi-
phase materials: a nested finite element solution strategy. Computer Methods in Applied Mechanics and Engineer-
ing, 193(48):5525 -- 5550, 2004.
A. Krushynska, V. Kouznetsova, and M. Geers. Visco-elastic effects on wave dispersion in three-phase acoustic
metamaterials. Journal of the Mechanics and Physics of Solids, 96:29 -- 47, 2016.
M. J. Leitman. Variational principles in the linear dynamic theory of viscoelasticity. Quarterly of Applied Mathematics,
24(1):37 -- 46, 1966.
S. Meguid and A. Kalamkarov. Asymptotic homogenization of elastic composite materials with a regular structure.
International Journal of Solids and Structures, 31(3):303 -- 316, 1994.
N. Ohno, X. Wu, and T. Matsuda. Homogenized properties of elastic viscoplastic composites with periodic internal
structures. International Journal of Mechanical Sciences, 42:1519 -- 1536, 2000.
M. Ostoja-Starzewski, S. D. Boccara, and I. Jasiuk. Couple-stress moduli and characteristic length of a two-phase
composite. Mechanics Research Communications, 26(4):387 -- 396, 1999.
M. Paggi, M. Corrado, and I. Berardone. A global/local approach for the prediction of the electric response of cracked
solar cells in photovoltaic modules under the action of mechanical loads. Engineering Fracture Mechanics, 168:
40 -- 57, 2016.
R. E. A. C. Paley and N. Wiener. Fourier transforms in the complex domain, volume 19. American Mathematical
Soc., 1934.
35
G. Panasenko. Boundary conditions for the high order homogenized equation: laminated rods, plates and composites.
Comptes Rendus MEcanique, 337(1):8 -- 14, 2009.
X. C. Q. Chen, G. Wang and J. Geng. Finite-volume homogenization of elastic/viscoelastic periodic materials.
Composite Structures, 182:457 -- 470, 2017.
V. P. Smyshlyaev. Propagation and localization of elastic waves in highly anisotropic periodic composites via two-scale
homogenization. Mechanics of Materials, R59:434 -- 447, 2009.
V. P. Smyshlyaev and K. Cherednichenko. On rigorous derivation of strain gradient effects in the overall behaviour
of periodic heterogeneous media. Journal of the Mechanics and Physics of Solids, 48(6):1325 -- 1357, 2000.
V. M. T.H. Tran and G. Bonnet. A micromechanics-based approach for the derivation of constitutive elastic coefficients
of strain-gradient media. International Journal of Solids and Structures, 49(5):783 -- 792, 2012.
A. Tran, J. Yvonnet, Q.-C. He, C. Toulemonde, and J. Sanahuja. A simple computational homogenization method
for structures made of linear heterogeneous viscoelastic materials. Computer Methods in Applied Mechanics and
Engineering, 200(45-46):2956 -- 2970, 2011.
P. Trovalusci, M. Ostoja-Starzewski, M. L. De Bellis, and A. Murrali. Scale-dependent homogenization of random
composites as micropolar continua. European Journal of Mechanics-A/Solids, 49:396 -- 407, 2015.
R.-M. Wang, S.-R. Zheng, and Y. G. Zheng. Polymer matrix composites and technology. Elsevier, 2011.
Y.-M. Yi, S.-H. Park, and S.-K. Youn. Asymptotic homogenization of viscoelastic composites with periodic mi-
crostructures. International Journal of Solids and Structures, 35(17):2039 -- 2055, 1998.
X. Yuan, Y. Tomita, and T. Andou. A micromechanical approach of nonlocal modeling for media with periodic
microstructures. Mechanics Research Communications, 35(1-2):126 -- 133, 2008.
D. Zah and C. Miehe. Computational homogenization in dissipative electro-mechanics of functional materials. Com-
puter Methods in Applied Mechanics and Engineering, 267:487 -- 510, 2013.
H. Zhang, S. Zhang, J. Y. Bi, and B. Schrefler. Thermo-mechanical analysis of periodic multiphase materials by a
multiscale asymptotic homogenization approach. International Journal for Numerical Methods in Engineering, 69
(1):87 -- 113, 2007.
36
Appendix A. Higher order recursive differential problems
The recursive differential problems are established at order ε, ε2, ε2 w−1 and ε2 w, with w ∈ Z and w ≥ 2 and
they are helpful to formulate the cell problems in section 4.
(27) and (32) related to the differential problems at order ε−1 and ε0,
Taking account of the solutions
respectively, the differential problem at order ε, stemmed from equation (19), is
(cid:16) C m
ijhk u(3)
h,k
(cid:17)
,j
+
(cid:17)
(cid:16)(cid:16) C m
N (2,2)(cid:17)
(cid:104)(cid:16) C m
(cid:16)
(cid:104)(cid:104)(cid:16) C m
ijhq1
ijhk
+
with interface conditions
(cid:104)(cid:104)
u(3)
h
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0,
ikhjN (2,0)
hpq1q2
+ C m
ikhjN (2,0)
hpq1q2,j +
,k
N (1,0)
hpq1
∂3 U M
p
∂xq1∂xq2∂xk
+
(130)
+ C m
iq1hjN (2,2)
hp,j − ρmN (1,0)
ipq1
,j
= f (3)
i
(x),
u(3)
h,k + N (2,0)
hpq1q2
∂3 U M
p
∂xq1∂xq2∂xk
+ N (2,2)s2 ∂ U M
p
∂xq1
(cid:17)(cid:17)
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0.
(131)
(cid:17)(cid:17)
(cid:16) C m
(cid:105)
s2 ∂ U M
p
∂xq1
ijhq2
The solvability condition of differential problem (130) in the class of Q−periodic functions and the
divergence theorem yield to
(cid:68) C m
f (3)
i
(x) =
ikhjN (2,0)
hpq1q2,j +
(cid:16) C m
ikhq2
N (1,0)
hpq1
(cid:17)(cid:69)
∂3 U M
p
∂xq1 ∂xq2∂xk
+
(cid:68) C m
iq1hjN (2,2)
hp,j − ρmN (1,0)
ipq1
(cid:69)
s2 ∂ U M
p
∂xq1
,
(132)
and consequentely the solution of the differential problem at order ε is
h (x, ξ, s) = N (3,0)
u(3)
hpq1q2q3
∂3 U M
p
∂xq1 ∂xq2∂xq3
+ N (3,2)
hpq1
s2 ∂ U M
p
∂xq1
.
(133)
Again, considering the solutions (32) and (133) related to the differential problems at order ε0 and ε,
respectively, the differential problem at order ε2, derived from equation (19), is
ijhk u(4)
h,k
+
,j
ikhjN (3,0)
hpq1q2q3
+ C m
iq3hkN (2,0)
hpq1q2
+
,k
ikhjN (3,0)
∂4 U M
p
∂xq1 ∂xq2∂xq3 ∂xk
+ (134)
(cid:17)
N (3,2)
hpq1
ijhq2
,j
+ C m
iq2hq1
N (2,2)
hp + C m
iq2hkN (3,2)
hpq1,k − ρmN (2,0)
ipq1q2
− ρmN (2,2)
ip
s4 U M
p = f (4)
i
(cid:16) C m
(cid:17)(cid:17)
hpq1q2q3,j
(cid:105)
s2 ∂ U M
p
∂xq1
(cid:104)(cid:104)
u(4)
h
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0,
(x),
(135)
(cid:17)(cid:17)
nj
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0.
(cid:17)
(cid:16) C m
(cid:104)(cid:16) C m
+
(cid:16)(cid:16) C m
(cid:17)
with interface conditions
(cid:104)(cid:104)(cid:16) C m
(cid:16)
.
divergence theorem lead to
(cid:68)(cid:16) C m
(cid:68) C m
+
ijhk
u(4)
h,k + N (3,0)
hpq1q2q3
∂4 U M
p
∂xq1∂xq2∂xq3∂xk
+ N (2,2)s2 ∂ U M
p
∂xq1
+ N (3,2)
hpq1
s2 ∂2 U M
∂xq1∂xk
p
The solvability condition of differential problem (134) in the class of Q−periodic functions and the
(cid:17)(cid:69)
(cid:69)
s2 ∂ U M
p
∂xq1
f (4)
i
(x) =
iq3hkN (2,0)
hpq1q2
+ C m
ikhjN (3,0)
hpq1q2q3,j
∂4 U M
p
∂xq1∂xq2 ∂xq3∂xk
+
N (2,2)
hp + C m
iq2hkN (3,2)
hpq1,k − ρmN (2,0)
ipq1q2
iq2hq1
− (cid:104)ρmN (2,2)
ip
(cid:105)s4 U M
p ,
(136)
and consequentely the solution of the differential problem at order ε2 is
u(4)
h (x, ξ, s) = N (4,0)
hpq1q2q3q4
∂4 U M
p
∂xq1 ∂xq2∂xq3 ∂xq4
+ N (4,2)
hpq1q2
s2 ∂2 U M
∂xq1∂xq2
p
+ N (4,4)
hp
s4 U M
p .
(137)
37
The generic recursive differential problem of odd order ε2 w−1, with w ∈ Z and w ≥ 2, is
( C m
ijhku(2 w+1)
h,k
)j +
1
2 w + 1
ijhq2 w+1
N (2 w,0)
hpq1...q2 w
+ C m
iq2 w+1hjN (2 w,0)
hpq1...q2 w,j+
(cid:104)(cid:16) C m
(cid:88)
(cid:105) ∂2 w+1 U M
P ∗(q)
p
(cid:17)
(cid:88)
,j
P ∗(q)
+ C m
iq2 w+1hq2 w
N (2 w−1,0)
hpq1...q2 w−1
∂xq1 ...∂xq2 w+1
+
1
2 w − 1
(cid:104)(cid:16) C m
ijhq2 w−1
N (2 w,2)
hpq1...q2 w−2
(cid:17)
(cid:105) ∂2 w−1 U M
,j
p
+
ipq1...q2 w−1
∂xq1...∂xq2 w−1
+ C m
iq2 w+1−2nhq2 w−2n
(138)
s2+
N (2 w−1,2n)
hpq1...q2 w−1−2n
+
+ C m
iq2 w−1hq2 w−2
N (2 w−1,2)
hpq1...q2 w−3
+ C m
iq2 w−1hjN (2 w,2)
n= w−1(cid:88)
(1 − δ1n)
+
n=1
1
2 w − 2n + 1
(cid:88)
P ∗(q)
(cid:104)(cid:16) C m
+ C m
iq2 w+1−2nhjN (2 w,2n)
hpq1...q2 w−2n,j − ρmN (2 w−1,2n−2)
(cid:17)
ipq1...q2 w+1−2n
N (2 w,2n)
hp
+ C m
iq1hkN (2 w,2 w)
hp,k
,j
(cid:104)(cid:16) C m
ijhq1
ijhq2 w+1−2n
hpq1...q2 w−2n
N (2 w,2n)
hpq1...q2 w−2,j − ρmN (2 w−1,0)
(cid:17)
(cid:105) ∂2 w+1−2n U M
(cid:105) ∂ U m
∂xq1 ...∂xq2 w+1−2n
,j
p
− ρmN (2 w−1,2 w−2)
ipq1
s2 w =
p
∂xq1
s2n+
1
2 w + 1
(cid:104) C m
iq2 w+1hjN (2 w)
hpq1...q2 w,j + C m
iq2 w+1hq2 w
N (2 w−1)
hpq1...q2 w−1
(cid:105) ∂2 w+1 U M
∂xq1...∂xq2 w+1
p
+
(cid:88)
(cid:88)
P ∗(q)
P ∗(q)
+
=
+
1
2 w − 2
n= w−1(cid:88)
n=1
+
(cid:104) C m
iq2 w−1hq2 w−2
N (2 w−1,2)
hpq1...q2 w−3
+ C m
iq2 w−1hjN (2 w,2)
hpq1...q2 w−2,j − ρmN (2 w−1)
ipq1...q2 w−1
(cid:105) ∂2 w−1 U M
p
∂xq1...∂xq2 w−1
(cid:88)
P ∗(q)
1
2 w − 2n + 1
(cid:104) C m
iq2 w+1−2nhq2 w−2n
N (2 w−1,2n)
hpq1...q2 w−1−2n
+
(1 − δ1n)
+ C m
iq2 w+1−2nhjN (2 w,2n)
hpq1...q2 w−2n,j − ρmN (2 w−1,2n−2)
ipq1...q2 w+1−2n
(cid:105) ∂2 w+1−2n U M
p
∂xq1 ...∂xq2 w+1−2n
+ (cid:104) C m
iq1hkN (2 w,2 w)
hp,k
− ρmN (2 w−1,2 w−2)
ipq1
and their interface conditions are
s2 w,
(cid:105) ∂ U m
p
(cid:104)(cid:104)
∂xq1
u(2 w+1)
h
and(cid:104)(cid:104)(cid:16) C m
ijhk u(2 w+1)
h,k
(cid:88)
1
2 w + 1
P ∗(q)
+
(cid:16) C m
ijhq2 w+1
N (2 w,0)
hpq1...q2 w
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:17) ∂2 w+1 U M
= 0
p
∂xq1...∂xq2 w+1
38
s2n+
+
s2+
(139)
(cid:17) ∂2 w+1−2n U M
p
∂xq1...∂xq2 w+1−2n
s2n+
N (2 w,2n)
hpq1...q2 w−2n
(cid:16) C m
(cid:88)
P ∗(q)
ijhq2 w−1
(1 − δ1n)
2 w − 2n + 1
+
1
2 w − 1
n= w−1(cid:88)
n=1
+
+ C m
ijhq1
N (2 w,2 w)
hp
∂ U M
p
∂xq1
∂xq1...∂xq2 w−1
s2+
p
1
P ∗(q)
hpq1...q2 w−2
ijhq2 w+1−2n
N (2 w,2)
(cid:17) ∂2 w−1 U M
(cid:16) C m
(cid:88)
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
s2 w(cid:17)
(cid:104)(cid:16) C m
(cid:88)
(cid:105) ∂2 w+2 U M
N (2 w+1)
ijhq2 w+2
P ∗(q)
= 0.
nj
p
(cid:88)
∂xq1...∂xq2 w+2
+
1
2 w
P ∗(q)
(cid:17)
(cid:104)(cid:16) C m
,j
The generic recursive differential problem of even order ε2 w is
( C m
ijhku(2 w+2)
h,k
)j +
1
2 w + 2
hpq1...q2 w+1
+ C m
iq2 w+2hjN (2 w+1)
hpq1...q2 w+1,j+
(cid:17)
+
,j
(cid:105)
(cid:17)
(cid:105) ∂2 w+2−2n U M
,j
p
+ C m
iq2 w+2hq2 w+1
N (2 w)
hpq1...q2 w
ijhq2 w
N (2 w+1,2)
hpq1...q2 w−1
+ C m
iq2 whq2 w−1
N (2 w,2)
hpq1...q2 w−2
n= w−1(cid:88)
(1 − δ1n)
+
n=1
1
22 w−2n+2
(cid:88)
P ∗(q)
(cid:104)(cid:16) C m
+ C m
iq2 whjN (2 w+1,2)
hpq1...q2 w−1,j − ρmN (2 w)
ipq1...q2 w
∂2 w U M
p
∂xq1...∂xq2 w
s2+
ijhq2 w−2n+2
N (2 w+1,2n)
hpq1...q2 w−2n+1
+ C m
iq2 w−2n+2hq2 w+1−2n
+ C m
iq2 w+2−2nhjN (2 w+1,2n)
hpq1...q2 w+1−2n,j − ρmN (2 w,2n−2)
ipq1...q2 w+2−2n
∂xq1 ...∂xq2 w+2−2n
s2n+
N (2 w+1,2 w)
hpq1
ijhq2
+ C m
iq2hq1
N (2 w,2 w)
hp
,j
+ C m
iq2hjN (2 w+1,2 w)
hpq1,j
− ρmN (2 w,2 w−2)
ipq1q2
+
+ C m
ijhq1
N (2 w+1,2 w)
hpq2
+ C m
iq1hq2
N (2 w,2 w)
hp
,j
+ C m
iq1hjN (2 w+1,2 w)
hpq2,j
− ρmN (2 w,2 w−2)
ipq2q1
(cid:17)
(cid:17)
(cid:105) ∂2 U m
p
∂xq1∂xq2
(140)
N (2 w,2n)
hpq1...q2 w−2n
+
s2 w − ρmN (2 w,2 w)
ip
U M
p s2 w =
(cid:104)(cid:16) C m
+
1
2
=
1
2 w + 2
1
2 w
(cid:88)
n= w−1(cid:88)
P (q)
n=1
+
+
(cid:88)
P ∗(q)
(cid:104) C m
iq2 m+2hjN (2 w+1)
hpq1...q2 w+1,j + C m
iq2 w+2hq2 w+1
N (2 w)
hpq1...q2 w
(cid:105) ∂2 w+2 U M
∂xq1 ...∂xq2 w+2
p
+
(cid:104) C m
iq2 whq2 w−1
N (2 w,2)
hpq1...q2 w−2
+ C m
iq2 whjN (2 w+1,2)
hpq1...q2 w−1,j − ρmN (2 w)
ipq1...q2 w
(cid:105)
∂2 w U M
p
∂xq1 ...∂xq2 w
s2+
(cid:88)
P ∗(q)
1
2 w − 2n + 2
(cid:104) C m
iq2 w−2n+2hq2 w+1−2n
N (2 w,2n)
hpq1...q2 w−2n
+
(1 − δ1n)
+ C m
iq2 w+2−2nhjN (2 w+1,2n)
hpq1...q2 w+1−2n,j − ρmN (2 w,2n−2)
ipq1...q2 w+2−2n
(cid:105) ∂2 w+2−2n U M
p
∂xq1...∂xq2 w+2−2n
s2n+
39
− ρmN (2 w,2 w−2)
ipq2q1
s2 w − (cid:104)ρmN (2 w,2 w)
ip
(cid:105) U M
p s2 w,
(141)
(142)
+
(cid:104)(cid:16) C m
iq2hq1
+
1
2
N (2 w,2 w)
hp
+ C m
iq2hjN (2 w+1,2 w)
hpq1,j
− ρmN (2 w,2 w−2)
ipq1q2
+
+ C m
iq1hq2
N (2 w,2 w)
hp
+ C m
iq1hjN (2 w+1,2 w)
hpq2,j
and their interface conditions are
and(cid:104)(cid:104)(cid:16) C m
ijhk u(2 w+2)
h,k
(cid:88)
1
2 w + 2
P ∗(q)
+
(cid:104)(cid:104)
u(2 w+2)
h
p
∂xq1∂xq2
(cid:105) ∂2 U m
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
(cid:17) ∂2 w+2 U M
= 0
p
N (2 w+1,0)
hpq1...q2 w+1
∂xq1 ...∂xq2 w+2
ijhq2 w
N (2 w+1,2)
hpq1...q2 w−1
∂xq1 ...∂xq2 w
s2+
ijhq2 w+2
(cid:16) C m
(cid:17) ∂2 w U M
(cid:16) C m
(cid:88)
p
P ∗(q)
(cid:16) C m
P ∗(q)
1
2 w
(cid:88)
n= w−1(cid:88)
(cid:104)(cid:16) C m
n=1
1
2
+
+
+
(1 − δ1n)
1
2 w − 2n + 2
ijhq2 w−2n+2
N (2 w+1,2n)
hpq1...q2 w−2n+1
s2n+
∂xq1...∂xq2 w+2−2n
p
(cid:17) ∂2 w+2−2n U M
(cid:105)(cid:105)(cid:12)(cid:12)(cid:12)ξ∈Σ1
= 0.
(cid:17) ∂2 U M
p
∂xq1 ∂xq2
s2 w(cid:105)(cid:17)
nj
N (2 w+1,2 m)
hpq1
+ C m
ijhq1
N (2 w+1,2 w)
hpq2
ijhq2
Appendix B. Symmetrization of the localization tensors
In order to perform the symmetrization of a tensor Zhkpq1...qn with respect to the indices q1...qn, the set
P ∗(q), which consists of all permutations with no fixed indices, is considered. For instance, if q = n, it
results that P ∗(q) =
and
the tensor Zhkpq1...qn is symmetrized with respect to q1, ..., qn as
q1 → q1, q2 → q2, ..., qn → qn
q1 → qn, q2 → q1, ..., qn → q2
, ..., fn =
f1 =
(cid:17)
(cid:16)
(cid:110)
(cid:16)
(cid:17)
1
n
Zhkpq1...qn =
1
n
Zhkpq1...qn + ... + Zhkpqnq1..q2
.
In particular, if q = 2 then the permutations set P ∗(q) with no fixed points is P ∗(q) =
q1, q2 → q2
tensor B(2,0)
q1 →
and the symmetrization with respect to q1 and q2 of the localization
q1 → q2, q2 → q1
δkq2N (1,0)
hpq1
+ N (2,0)
results
, f2 =
f1 =
hpq1q2,k
(cid:17)
hkpq1q2
=
(cid:17)(cid:111)
(cid:16)
(cid:16)
δkq2N (1,0)
hpq1
+ δkq1N (1,0)
hpq2
+ N (2,0)
hpq1q2,k + N (2,0)
hpq2q1,k
=
(cid:17)
(cid:17)(cid:111)
(cid:16)
(cid:110)
(cid:16)
(cid:88)
P ∗(q)
(cid:17)
δkq2N (1,0)
hpq1
+ δkq1N (1,0)
hpq2
+ N (2,0)
hpq1q2,k.
In case of q = 3, the permutations set P ∗(q) having no fixed points is P ∗(q) =
q2, q3 → q3
q1 → q3, q2 → q1, q3 → q2
tensor B(3,0)
q1 → q2, q2 → q3, q3 → q1
δkq3 N (2,0)
hpq1q2
+ N (3,0)
hpq1q2q3,k
, f3 =
, f2 =
hkpq1q2q3
=
is symmetrized with respect to q1, q2 and q3 as
f1 =
q1 → q1, q2 →
and the localization
(cid:16)
(cid:17)
(cid:17)
(143)
(cid:16)
(cid:110)
(cid:17)(cid:111)
(cid:16)
(cid:16)
(cid:16)
B(3,0)
hkpq1q2q3
=
1
3
δkq3N (2,0)
hpq1q2
+ δkq1N (2,0)
hpq2q3
+ δkq2 N (2,0)
hpq3q1
+
40
(cid:17)
(cid:16)
(cid:17)
B(2,0)
hkpq1q2
=
1
2
(cid:16)
=
1
2
(cid:17)
=
+ N (3,0)
hpq1q2q3,k + N (3,0)
hpq2q3q1,k + N (3,0)
hpq3q1q2,k
(cid:16)
=
1
3
δkq3N (2,0)
hpq1q2
+ δkq1N (2,0)
hpq2q3
+ δkq2 N (2,0)
hpq3q1
(cid:17)
+ N (3,0)
hpq1q2q3,k,
(144)
as it appears in Eq. (79).
Appendix C. Perturbation functions of first, second and third order
Appendix C.1. Perturbation function of first order N (1,0)
hpq
Let Ω be a layered domain obtained as a d2− periodic arrangement of two different layers whose thick-
ness is s1 and s2 (here d2 = s1 + s2 and η = s1/s2 are defined). The phases are supposed to be homogeneous
and orthotropic, with an orthotropic axis coincident with the layering direction e1. The micro-fluctuation
are analytically obtained by solving the first cell problems (33). The superscript i = {1, 2}
functions N (1,0)i
stands for the phase 1 and the phase 2 and they are formulated as
hpq
(cid:16) C 1
(cid:16) C 1
211 = −
N (1,0)1
222 = −
N (1,0)1
(cid:17)
(cid:17)
1122 − C 2
C 2
2222 η + C 1
1122
2222
ξ2
, N (1,0)2
211 =
2222 − C 2
C 2
2222 η + C 1
2222
2222
ξ2
, N (1,0)2
222 =
(cid:17)
(cid:16) C 1
1212 − C 2
C 2
1212 η + C 1
1212
1212
ξ2
(cid:16) C 1
(cid:16) C 1
η
η
1122 − C 2
2222 η + C 1
C 2
1122
2222
2222 − C 2
C 2
2222 η + C 1
2222
2222
(cid:17)
(cid:17)
ξ2
ξ2
,
,
η
N (1,0)1
112 = N (1,0)1
121 = −
, N (1,0)2
112 = N (1,0)2
121 =
(145)
(146)
(147)
(cid:17)
ξ2
.
(cid:16) C 1
1212 − C 2
1212 η + C 1
C 2
1212
1212
Such functions depend on the fast variable ξ, since the microstructure enjoys the simmetry property. In the
following it is assumed that the coordinate ξ2 is centered in both layers.
Appendix C.2. Perturbation functions of second order N (2,0)
hpqr
The perturbation functions N (2,0)i
hpqr , i = {1, 2}, deriving from the cell problem (35) are:
N (2,0)1
1111 = A2
1111ξ2
2 + A0
1111, N (2,0)2
1111 = B2
1111ξ2
2 + B0
1111,
N (2,0)1
2211 = A2
2211ξ2
2 + A0
2211, N (2,0)2
2211 = B2
2211ξ2
2 + B0
2211,
N (2,0)1
2222 = A2
2222ξ2
2 + A0
2222, N (2,0)2
2222 = B2
2222ξ2
2 + B0
2222,
N (2,0)1
1122 = A2
1122ξ2
2 + A0
1122, N (2,0)2
1122 = B2
1122ξ2
2 + B0
1122,
(148)
(149)
(150)
(151)
where the constants A2
1111, A0
1111, B2
1111, B0
1111, A2
2211, A0
2211, B2
2211, B0
2211, A2
2222, A0
2222, B2
2222, B0
2222, A2
1122,
A0
1122, B2
1122 and B0
1122 are determined as follows
1111 = − 1
A2
2
(η + 1)
A2,0
(cid:16) C 2
1111 + ηA2,1
2222 η + C 1
1111
2222
(cid:17) C 1
1212
,
(152)
41
A0,3
1111 η3 + A0,2
1111 η2 + A0,1
1111 η + A0,0
1111
(cid:17)
,
C 1
1212
1212
A0
1111 =
1
24
B2
1111 =
1
2
B0
1111 =
1
24
A2
2211 =
1
2
,
η
1111
2222
1212
ηB2,1
(η + 1)
2222
(cid:17) C 2
(cid:17) C 2
2222 η + C 1
1111 + η2B2,2
2222 η + C 1
(cid:16)
(η + 1)4(cid:16) C 2
(cid:16) C 2
(cid:16)−2A0,3
(η + 1)4(cid:16) C 2
(cid:16) C 1
(cid:17) C 1
(cid:16) C 2
1212 − C 2
(cid:17)
(cid:16)
(cid:16) C 1
(η + 1)3(cid:16) C 2
1111 η3 + B0,2
1212 − C 2
2222 η + C 1
1212 η + C 1
(cid:17)
C 1
1122
1212
1212
2222
1212
2222
η
η
,
(cid:17) C 1
(cid:17)
,
1111 η2 + B0,1
1111 η − A0,0
1111
2
C 2
1212
1212
1122 η2 C 2
C 1
2222 + 3 C 1
1122 η C 2
1212 η + C 1
1212
2222
(cid:17) C 2
1122
2222 + 2 C 2
C 1
2222
2211 = − 1
A0
24
2211 = − ηC 2
B2
C 2
(cid:16)
B0
2211 =
1
24
C 1
1122
2222C 1
1122
2222
A2
2211,
A2
2222 =
2222 − C 2
C 1
2 C 2
2222 η + 2 C 1
2222
2222
, A0
2222 = − 1
24
2222 = −ηA2
B2
2222, B0
2222 = − 2η + 1
η + 2
A0
2222,
A2
1122 =
1212 − C 2
C 1
2 C 2
1212 η + 2 C 1
1212
1212
, A0
1122 = − 1
24
1122 = −ηA2
B2
1122, B0
1122 = − 2η + 1
η + 2
A0
1122.
(cid:17)
,
C 1
2222
(cid:17)
η
,
2 C 1
1122 η2 C 2
2222 + 3 η C 2
(η + 1)3(cid:16) C 2
1122
1212 η + C 1
1212
C 1
(cid:17) C 2
2222
C 1
2222 + C 2
1122
2222
1212 − C 2
1212
C 1
(cid:17)(cid:16) C 1
(cid:17)
(cid:17)
2222
(cid:16) C 1
(cid:16) C 2
(cid:16) C 1
(cid:16) C 2
2222 − C 2
2222 η + C 1
2222
2222
η (η + 2)
(η + 1)2
(cid:17)
(cid:17)
1212 − C 2
1212 η + C 1
1212
1212
η (η + 2)
(η + 1)2
,
,
Appendix C.3. Perturbation functions N (2,2)
hp
The perturbation functions N (2,2)i
hp
, i = {1, 2}, obtained by performing the cell problem (39) are:
N (2,2)1
11
= A2
11ξ2
2 + A0
11, N (2,2)2
11
= B2
11ξ2
2 + B0
11,
42
(153)
(154)
(155)
(156)
(157)
(158)
(159)
(160)
(161)
(162)
(163)
(164)
N (2,2)1
22
= A2
22ξ2
2 + A0
22, N (2,2)2
22
= B2
22ξ2
2 + B0
22,
where the constants A2
11, A0
11, B2
11, B0
22, B2
22 and B0
22 are
A2
11 =
(ρ1 − ρ2)
(η + 1) C 1
1212
1
2
, A0
11 = − 1
24
1212 η2 + 3 C 2
1212 (η + 1)4 C 1
C 2
1212
1212 η + 2 C 1
1212
11, A2
22, A0
(ρ1 − ρ2) η
11 = − 1
B2
2
(ρ1 − ρ2) η
(η + 1) C 2
1212
, B0
11 =
1
24
A2
22 =
(ρ1 − ρ2)
(η + 1) C 1
2222
1
2
, A0
22 = − 1
24
22 = − 1
B2
2
(ρ1 − ρ2) η
(η + 1) C 2
2222
, B0
22 =
1
24
(ρ1 − ρ2) η
(ρ1 − ρ2) η
(ρ1 − ρ2) η
1212 η2 + 3 C 1
2 C 2
1212 (η + 1)4 C 1
C 2
1212
1212 η + C 1
1212
2222 η2 + 3 C 2
2222 (η + 1)4 C 1
C 2
2222
2222 η + 2 C 1
2222
2 C 2
2222 η2 + 3 C 1
2222 (η + 1)4 C 1
C 2
2222
2222 η + C 1
2222
(cid:16) C 2
(cid:16)
(cid:16) C 2
(cid:16)
(cid:17)
(cid:17)
(cid:17)
(cid:17)
,
(165)
,
(166)
(167)
,
(168)
.
(169)
Appendix C.4. Perturbation functions of third order N (3,0)
hpqrs
The non-vanishing micro-fluctuation functions N (3,0)i
lem (37), with w = 1, are:
hpqrs, i = {1, 2}, obtained by performing the cell prob-
N (3,0)1
21111 = A3
21111ξ3
2 + A1
21111ξ2, N (3,0)2
21111 = B3
21111ξ3
2 + B1
21111ξ2,
N (3,0)1
11222 = A3
11222ξ3
2 + A1
11222ξ2, N (3,0)2
11222 = B3
11222ξ3
2 + B1
11222ξ2,
N (3,0)1
12111 = A3
12111ξ3
2 + A1
12111ξ2, N (3,0)2
12111 = B3
12111ξ3
2 + B1
12111ξ2,
N (3,0)1
22222 = A3
22222ξ3
2 + A1
22222ξ2, N (3,0)2
22222 = B3
22222ξ3
2 + B1
22222ξ2,
(170)
(171)
(172)
(173)
where the constants A3
21111, A1
21111, B3
21111, B1
21111, A3
11222, A1
11222, B3
11222, B1
11222, A3
12111, A1
12111, B3
12111,
A1
21111 =
η(A1,4
21111η4 + A1,3
72
2222 η + C 1
2222
21111η3 + A1,2
21111η2 + A1,1
(η + 1)4 C 1
21111η + A1,0
C 1
C 2
1212
2222
1212
21111)
B3
21111 =
18
21111η + B3,1
21111)
η(B3,2
2222 η + C 1
2222
(η + 1) C 2
1212
C 2
2222
B1
21111 =
η(B1,4
21111η4 + B1,3
72
2222 η + C 1
2222
21111η3 + B1,2
21111η2 + B1,1
(η + 1)4 C 2
21111η + B1,0
C 2
C 1
1212
2222
1212
21111)
,
,
(174)
(175)
(176)
(177)
B1
12111, A3
22222, A1
22222, B3
22222 and B1
22222 are
A3
21111 =
18
21111η + A3,0
21111
A3,1
2222 η + C 1
2222
(η + 1) C 1
1212
C 1
2222
(cid:16) C 2
(cid:16) C 2
(cid:16) C 2
(cid:16) C 2
(cid:17)
(cid:17)
(cid:17)2
(cid:17)2
,
,
43
1212 η3 +
1212 + 2 C 2
1212
η2 +
2 C 1
1212 + C 2
1212
η + C 1
1212
1212 − C 2
1212
(cid:16)
(cid:17)
(cid:17)
(cid:16) C 2
(cid:16) C 2
2
1212 η + C 1
1212
1212 η + C 1
1212
(cid:17)
(cid:17)2
(cid:16)− 3
(cid:17)2
2
(η + 1)2
C 1
1212 + C 2
1212
(η + 1)2
(cid:17)
η
(cid:17)(cid:16) C 1
(cid:17)(cid:16) C 1
(cid:17)
(cid:17)
,
,
C 2
1212 η3 +
1212 − 3
C 1
4
C 2
1212
η2 +
1212 − C 2
1212
(cid:17)
72 (η + 1)2(cid:16) C 2
1212 − 6 C 2
1212
4 C 1
(cid:16)−6 C 1
(cid:17)2
1212 η + C 1
1212
η2 +
1212 + 3 C 2
1212
(cid:17)
η − C 1
1212
(cid:17)
,
11222 = − 1
A3
18
11222 = − 1
A1
18
(cid:16) C 2
(cid:16)− 1
(cid:16) C 1
(cid:16) C 1
4
B3
11222 =
η
18 C 2
1212 − C 2
1212 η + 18 C 1
1212
η
B1
11222 =
1212 − C 2
1212
(cid:16) C 1
(cid:16) 3
(cid:17)
(cid:17)(cid:16)(cid:16)
1212
,
A3
12111 =
A1
12111 =
18
72
η(A3,3
(cid:16) C 2
(cid:16) C 2
(cid:16) C 1
η
η(A1,3
12111 = − 1
B3
18
(cid:16) C 2
(cid:16) C 1
(cid:16) C 1
(cid:16) C 1
(cid:16) C 1
22222 = − 1
A3
18
22222 = − 1
A1
18
B3
22222 =
B1
22222 =
1
18
1
18
1212
(cid:17)2
(cid:17)2
(cid:17)(cid:16)
(cid:16) C 2
(cid:17)2
(cid:17)(cid:16) C 2
(cid:17)(cid:16)− 1
(cid:17)(cid:16)
(cid:17)(cid:16)(cid:16) C 1
4
12111η3 + A3,2
1212 η + C 1
12111η2 + A3,1
12111η1 + A3,0
C 1
C 1
2222 (η + 1)3
C 2
12111)
2222
1212
1212
12111η3 + A1,2
1212 η + C 1
12111η2 + A1,1
12111η1 + A1,0
C 1
C 1
2222 (η + 1)3
C 2
12111)
1212
2222
1212
,
,
1212 − C 2
C 2
1111
2222 − C 2
C 2
1122
2 − C 2
1122
C 2
1212
(cid:17) C 2
1212 η + C 1
1212
C 2
1212
2222
(cid:17)
,
B1
12111 =
η(B1,3
12111η3 + B1,2
1212 η + C 1
1212
72
12111η2 + B1,1
12111η1 + B1,0
(η + 1)3 C 2
12111)
C 1
2222
,
2222 − C 2
2222
2222 η3 +
2222 + 2 C 2
2222
(cid:16)
η2 +
(cid:17)
η + C 1
2222
2 C 1
2222 + C 2
2222
(cid:17)
(cid:16)− 3
2
(cid:17)
(cid:17)
η
2222 + C 2
C 1
2222
C 2
2222
η2 +
,
,
(cid:17)
η + C 1
2222
(cid:17)
η
,
(cid:17)
C 2
2222
η − C1
2222
4
(cid:17)
η
.
2222 η + C 1
2222
(η + 1)2
(cid:17)
(cid:17)2
C 2
4
2222
1212
2222 η3 +
(cid:16) C 1
(cid:16) C 2
(cid:16) 3
(cid:16) C 2
(cid:16) C 1
(η + 1)2(cid:16) C 2
(cid:17)
(η + 1)2(cid:16) C 2
C 2
2222
2
2
2222
(cid:17)2
2222 − 3
C 1
(cid:17)
(cid:16)− 3
2222
η2 +
2222 η + C 1
2222
C 1
2222 + 3
4
2
(cid:17)2
2222 η + C 1
2222
2222 η + C 1
(η + 1)2
(cid:17)
(cid:16)
(cid:17)2
44
2222 − C 2
2222
C 2
2222 − C 2
2222
2222 − 3
2222 − C 2
2222
η3 C 2
2222 +
2222 + 2 C 2
η2 +
2 C 1
2222 + C 2
2222
(178)
(179)
(180)
(181)
(182)
(183)
(184)
(185)
(186)
(187)
(188)
(189)
Appendix C.5. Perturbation functions N (3,2)
hpq
The perturbation functions N (3,2)i
hpq
, i = {1, 2}, obtained by performing the cell problem (41) are:
N (3,2)1
211 = A3
211ξ3
2 + A1
211ξ2, N (3,2)2
211 = B3
211ξ3
2 + B1
211ξ2,
N (3,2)1
121 = A3
121ξ3
2 + A1
121ξ2, N (3,2)2
121 = B3
121ξ3
2 + B1
121ξ2,
N (3,2)1
112 = A3
112ξ3
2 + A1
112ξ2, N (3,2)2
112 = B3
112ξ3
2 + B1
112ξ2,
N (3,2)1
222 = A3
222ξ3
2 + A1
222ξ2, N (3,2)2
222 = B3
222ξ3
2 + B1
222ξ2,
(190)
(191)
(192)
(193)
where the constants A3
211, A1
211, B3
211, B1
211, A3
121, A1
121, B3
121, B1
121, A3
112, A1
112, B3
112, B1
112, B1
122, A3
222,
(cid:17) C 1
(cid:17)2
(cid:17) C 2
(cid:17)2
211η + A3,0
A3,1
2222 η + C 1
2222
211
,
C 1
2222
1212
211η3 + A1,2
2222 η + C 1
2222
211η2 + A1,1
211η + A1,0
211)
C 2
C 1
C 1
2222
1212
1212
211η + B3,1
211)
η(B3,2
2222 η + C 1
2222
,
C 2
2222
1212
211η3 + A1,2
2222 η + C 1
2222
211η2 + A1,1
211η + A1,0
211)
C 1
C 2
C 2
2222
1212
1212
121 + A3,0
1212 η + C 1
121
1212
(cid:17)
,
2
C 1
1212
121η3 + A1,2
1212 η + C 1
121η + A1,0
121η2 + A1,1
121)
2
C 2
1212 C 1
1212
1212
(cid:17)2
(cid:17)2
(cid:17) ,
121η + B3,1
121)
1212 η + C 1
1212
121η3 + B1,2
1212 η + C 1
121η + B1,0
121η2 + B1,1
121)
C 2
2 C 1
1212
1212
1212
,
,
(194)
(195)
(196)
(197)
(198)
(199)
(200)
(201)
(202)
,
,
A1
222, B3
222 and B1
222 are
1
6
1
24
1
6
1
24
(η + 1)
(η + 1)
η(A1,4
η(A1,4
211η4 + A1,3
211η4 + A1,3
(cid:16) C 2
(η + 1)4(cid:16) C 2
(cid:16) C 2
(η + 1)4(cid:16) C 2
(cid:16) C 2
24 (η + 1)4(cid:16) C 2
2(cid:16) C 2
24 (η + 1)4(cid:16) C 2
(cid:16) C 2
121η4 + B1,3
121η4 + A1,3
η(B3,2
ηA3,1
(cid:17)
1212
6 (η + 1) C 2
6 (η + 1)
η(B1,4
η(A1,4
A3
211 =
A1
211 =
B3
211 =
B1
211 =
A3
121 =
A1
121 =
B3
121 =
B1
121 =
A3
112 =
1
6
(− (η + 3) ρ1 + 2 ρ2) C 1
1212 η + C 1
1212
(η + 1) C 1
1212
1212
,
45
η(A1,4
112η4 + A1,3
112η3 + A1,2
A1
112 =
1212 η + C 1
112η2 + A1,1
(η + 1)4 C 2
112η + A1,0
112)
,
C 1
1212
1212
24
(cid:16) C 2
(cid:16)(cid:16)(cid:0)− 3
(cid:17)2
1212
(cid:1) C 2
B3
112 =
1
3
2 ρ2 + ρ1
1212 + 1
(cid:16) C 2
η − 1
(cid:17)
1212
2 ρ2 C 1
1212 η + C 1
1212
1212 + C 1
1212
2 ρ2 C 2
(η + 1) C 2
1212
(cid:0)− 1
2 ρ2 + ρ1
(cid:1)(cid:17)
η
,
B1
112 =
η(B1,4
112η4 + B1,3
24 (η + 1)4(cid:16) C 2
112η3 + B1,2
1212 η + C 1
112η2 + B1,1
112η + B1,0
112)
C 2
C 1
1212
1212
1212
A3
222 =
1
6
(cid:16) C 2
(− (η + 3) ρ1 + 2 ρ2) C 1
2222 η + C 1
2222
(cid:17)
2222 + 2 C 2
(η + 1) C 1
2222
2222 η ρ2
,
,
,
η(A1,4
222η4 + A1,3
24 (η + 1)4(cid:16) C 2
(cid:16)(cid:16)(cid:0) 3
(cid:1) C 2
2 ρ2 + ρ1
A1
222 =
B3
222 =
1
3
222η3 + A1,2
2222 η + C 1
222η2 + A1,1
222η + A1,0
222)
C 1
C 2
2222
2222
2222
2222 + 1
2 ρ2 C 1
2222
η − 1
2 ρ2 C 2
2 ρ2 + δ22 ρ1
2222 +(cid:0)− 1
(cid:17)
2222 η + C 1
2222
(η + 1) C 2
2222
(cid:17)
(cid:1) C 1
2222
η
,
(cid:17)
(cid:17)
(cid:17)2
(cid:17)2
(cid:16) C 2
(cid:17)2
(203)
(204)
(205)
(206)
(207)
(208)
(209)
B1
222 =
η(B1,4
222η4 + B1,3
24 (η + 1)4(cid:16) C 2
222η3 + B1,2
2222 η + C 1
222η + B1,0
222η2 + B1,1
222)
C 1
C 2
2222
2222
2222
.
The constants that appear in the perturbation functions Eq. (194)-Eq. (209) depend on the geometric and
mechanical properties of the phases 1 and 2 and for sake of simplcity are not reported here.
Appendix D. Constants appearing in the perturbation functions
The constants that characterize the perturbation functions in appendix D.2, A2,0
1111, A2,1
1111, A0,3
1111, A0,2
1111,
A0,1
1111, A0,0
1111, B2,1
1111, B2,2
1111, B0,2
2222 and B0,1
(cid:17) C 1
2222, assume the form
1212 − C 2
C 1
1122)2 − C 2
1122
2222
1122 − ( C 2
A2,1
1111 =
1122 + C 1
1212
1111 − C 2
1111
A2,0
1111 = C 1
1122
2
1212 − C 2
1122
1122 − C 2
1122
1212 − C 1
C 1
2222
1111 − C 2
1111
(cid:16) C 1
(cid:16) C 1
(cid:17)
,
,
(cid:17)
(cid:16) C 1
(cid:16) C 2
(cid:16)
+
(cid:16) C 1
(cid:16)(cid:16) C 2
(cid:17) C 1
(cid:17) C 1
1122
(cid:17) C 1
A0,3
1111 = C 2
1212
1122 − C 1
1212 − C 1
1122
C 2
1122 + ( C 2
2
1122)
+ C 2
2222
1111 − C 2
1111
A0,2
1111 =
4 C 2
1122 − 4 C 1
1122
1212 − C 1
1122
+ ( C 1
2222 + 3 C 2
2222)( C 1
1111 − C 2
1111) C 2
1212,
2 − 2 C 1
1122
C 2
1122 + 3 ( C 2
2
1122)
+
46
(cid:17)(cid:17)
,
(210)
(211)
(212)
(213)
A0,1
1111 =
5 C 2
1122 − 5 C 1
1122
1212 − 3 C 1
1122
2
+ 3 C 1
1122
C 2
1122 + 3 C 1
2222
+ 2 C 1
1212
( C 2
1122)
2 − C 1
C 2
1122 + C 2
2222
1122
1111 − C 2
1111
(cid:16) C 1
(cid:17)(cid:17)
,
1122 − C 1
1122
1212 − C 1
1122
2
+ C 1
1122
C 2
1122 + C 1
2222
1212 − C 1
1122
1122 − C 2
1212
1111 − C 2
(cid:16) C 1
(cid:16) C 1
C 1
1122 + C 2
2222
1212 + C 1
C 2
2222
1122
1122 − C 2
1212
1122 + C 2
1122
1111 − C 2
1111
,
(cid:17) C 1
(cid:17) C 2
(cid:17) C 2
(cid:17) C 2
(cid:17) C 1
(cid:16)(cid:16)
(cid:16)
(cid:16)(cid:16) C 2
(cid:16) C 2
(cid:16) C 2
+
+
2
2
A0,0
1111 = 2 C 1
1212
B2,2
1111 = ( C 2
1122)
1111 = −C 1
B2,1
(cid:16)(cid:16)
(cid:16)−C 1
1111 =
− 2
B0,2
1122
2
(cid:18)
+ C 1
1122
C 2
1122 + C 1
2222
(cid:18) 2
3
(cid:16) C 2
(cid:17) C 1
− 3( C 1
1111 − C 2
1111)
2222
3
+ C 1
2222
1212.
(cid:17)(cid:17) C 2
1212
(cid:17)(cid:17)
,
(cid:16) C 1
1111 − C 2
1111
(cid:16) C 1
1111
1111 − C 2
(cid:17)
(cid:17)
1111
,
(cid:16) C 1
(cid:17)(cid:17) C 1
(214)
(215)
(216)
(217)
(cid:19)
+
(219)
5 C 1
1122 − 5 C 2
1122
1212 + 3 C 1
1122
1122 − 3 ( C 2
C 2
1122)
2 − 3 C 2
2222
1111 − C 2
1111
1212+ (218)
(cid:17)(cid:17) C 2
1111
1212,
(cid:16) C 1
1111 − C 2
(cid:19)
1111 = −3
B0,1
−C 1
1122
2
+
( C 2
1122) − 4
3
C 2
1212
C 1
1122 +
2
( C 2
1122)
+
1
3
4
3
( C 2
1122) C 2
1212
Appendix E. Overall inertial terms and overall consitutive tensors
By taking into account the method I, the transformed inertial tensors appearing in Eq. (122) assume the
form:
I2121 = ε2(cid:104)ρm(N (1,0)
121 N (1,0)
121 ) − 2ρmN (2,0)
2211 (cid:105) 1
ρ
I2222 = ε2(cid:104)ρm(N (1,0)
222 N (1,0)
222 ) − 2ρmN (2,0)
2222 (cid:105) 1
ρ
I22 = ε2(cid:104)ρm(N (2,2)
22 + N (2,2)
22
)(cid:105) 1
ρ
,
I1212 = ε2(cid:104)ρm(N (1,0)
112 N (1,0)
112 ) − 2ρmN (2,0)
1122 (cid:105) 1
ρ
I1111 = ε2(cid:104)ρm(N (1,0)
211 N (1,0)
211 ) − 2ρmN (2,0)
1111 (cid:105) 1
ρ
,
,
,
,
I11 = ε2(cid:104)ρm(N (2,2)
11 + N (2,2)
11
)(cid:105) 1
ρ
.
47
(220)
(221)
(222)
(223)
(224)
(225)
Referred to the Eq. (122), the constitutive tensors in the Laplace space are
J11 = ε2(cid:104) G1212B(2,2)
121 (cid:105),
121 B(2,2)
J22 = ε2(cid:104) G2222B(2,2)
222 (cid:105),
222 B(2,2)
G1111 = (cid:104) G1111B(1,0)
1111 B(1,0)
1111 + G2222B(1,0)
2211 B(1,0)
2211 + 2 G1122B(1,0)
1111(cid:105),
2211 B(1,0)
G2121 = (cid:104) G2222B(1,0)
2221 B(1,0)
2221(cid:105) = G1212,
G2222 = (cid:104) G2222B(1,0)
2222(cid:105),
2222 B(1,0)
2121 = −ε2(cid:104)2 G1122B(2,0)
J 1
11211B(2,2)
222 − 2 G1212B(1,0)
1221 B(3,2)
1221 + 2 G2222B(2,0)
22211B(2,2)
222 − 2 G1212B(1,0)
2121 B(3,2)
2121 +
− 2 G1212B(1,0)
2121 B(3,2)
1221 − 2 G1212B(1,0)
2121(cid:105),
2121 B(3,2)
1212 = −ε2(cid:104)−2 G1212B(1,0)
J 1
1212 B(3,2)
1212 + 2 G1212B(2,0)
121 (cid:105),
12122B(2,2)
2222 = ε2(cid:104)−2 G2222B(1,0)
J 1
2222 B(3,2)
2222 + 2 G2222B(2,0)
222 (cid:105),
22222B(2,2)
111111 = ε2(cid:104)−2 G1111B(1,0)
S1
1111 B(3,0)
111111 + G1212B(2,0)
12111B(2,0)
12111 + G1212B(2,0)
21111B(2,0)
21111+
+ 2 G1212B(2,0)
12111B(2,0)
21111 − 2 G2222B(1,0)
2211 B(3,0)
221111 − 2 G1122B(1,0)
2211 B(3,0)
111111 − 2 G2211B(1,0)
1111 B(3,0)
221111(cid:105),
211211 = ε2(cid:104) G1111B(2,0)
S1
11211B(2,0)
11211 + 2 G1122B(2,0)
11211B(2,0)
22211 + G2222B(2,0)
22211B(2,0)
22211 − 2 G1212B(1,0)
1212 B(3,0)
12111+
(226)
(227)
(228)
(229)
(230)
(231)
(232)
(233)
(234)
(235)
− 2 G1212B(1,0)
1221 B(3,0)
122111 − 2 G1212B(1,0)
2121 B(3,0)
122111 − 2 G1212B(1,0)
1221 B(3,0)
212111 − 2 G1212B(1,0)
2121 B(3,0)
212111(cid:105).
The inertial and constitutive tensors derived from the method 2 coincide with those ones from (220) to (230),
whereas the ones from (231) to (235) are slightly different because the localization tensors B(3,0)
and
B(3,2)
hkpq1
1111 = −ε2(cid:104)2 G1212B(2,0)
J 2
disappear in the formulation and they assume the form:
121 + 2 G1212B(2,0)
121 (cid:105)
21111B(2,2)
12111B(2,2)
hkpq1q2q3
(236)
2121 = −ε2(cid:104)2 G1122B(2,0)
J 2
11211B(2,2)
222 + 2 G2222B(2,0)
222 (cid:105)
22211B(2,2)
1212 = −ε2(cid:104)2 G1212B(2,0)
J 2
121 (cid:105),
12122B(2,2)
2222 = −ε2(cid:104)2 G2222B(2,0)
J 2
222 (cid:105),
22222B(2,2)
111111 = ε2(cid:104) G1212B(2,0)
S2
12111B(2,0)
12111 + G1212B(2,0)
21111B(2,0)
21111 + 2 G1212B(2,0)
12111B(2,0)
211211 = ε2(cid:104) G1111B(2,0)
S2
11211B(2,0)
11211 + 2 G1122B(2,0)
11211B(2,0)
22211 + G2222B(2,0)
22211B(2,0)
21111(cid:105),
22211(cid:105).
48
(237)
(238)
(239)
(240)
(241)
|
1802.01206 | 1 | 1802 | 2018-02-04T22:17:44 | In-Situ Self-Monitoring of Real-Time Photovoltaic Degradation Only Using Maximum Power Point: the Suns-Vmp Method | [
"physics.app-ph"
] | The uncertainties associated with technology- and geography-specific degradation rates make it difficult to calculate the levelized cost of energy (LCOE), and thus the economic viability of solar energy. In this regard, millions of fielded photovoltaic (PV) modules may serve as a global testbed, where we can interpret the routinely collected maximum power point (MPP) time-series data to assess the time-dependent "health" thereof. The existing characterization methods, however, cannot effectively mine/decode these datasets to identify various degradation pathways of the corresponding solar modules. In this paper, we propose a new methodology, i.e., the Suns-Vmp method, which offers a simple and powerful approach to monitoring and diagnosing time-dependent degradation of solar modules by physically mining the MPP data. The algorithm reconstructs "IV" curves by using the natural illumination- and temperature-dependent daily MPP characteristics as constraints to fit the physics-based compact model. These synthetic IV characteristics are then used to determine the time-dependent evolution of circuit parameters (e.g., series resistance) which in-turn allows one to deduce the dominant degradation mode (e.g., corrosion) for the modules. The proposed method has been applied to analyze the MPP data from a test facility at the National Renewable Energy Laboratory (NREL). Our analysis indicates that the solar modules degraded at a rate of 0.7 %/year due to discoloration and weakened solder bonds. These conclusions are independently validated by outdoor IV measurement and on-site imaging characterization. Integrated with physics-based degradation models or machine learning algorithms, the method can also serve to predict the lifetime of PV systems. | physics.app-ph | physics | In-Situ Self-Monitoring of Real-Time Photovoltaic
Degradation Only Using Maximum Power Point – the
Suns-Vmp Method
Xingshu Sun,1 Raghu Vamsi Krishna Chavali,1 and Muhammad Ashraful Alam1,*
1 Network of Photovoltaic Technology, Purdue University, West Lafayette, IN, 47907, USA
*Corresponding author: [email protected]
Abstract - The uncertainties associated with technology-
and geography-specific degradation rates make it difficult
to calculate the levelized cost of energy (LCOE), and thus
the economic viability of solar energy. In this regard,
millions of fielded photovoltaic (PV) modules may serve as
a global testbed, where we can interpret the routinely
collected maximum power point (MPP) time-series data to
assess the time-dependent "health" thereof. The existing
characterization methods, however, cannot effectively
mine/decode these datasets to identify various degradation
pathways of the corresponding solar modules. In this paper,
we propose a new methodology, i.e., the Suns-Vmp method,
which offers a simple and powerful approach to monitoring
and diagnosing time-dependent degradation of solar
modules by physically mining the MPP data. The algorithm
reconstructs "IV" curves by using the natural illumination-
and temperature-dependent daily MPP characteristics as
constraints to fit the physics-based compact model. These
synthetic IV characteristics are then used to determine the
time-dependent evolution of circuit parameters (e.g., series
resistance) which in-turn allows one to deduce the dominant
degradation mode (e.g., corrosion) for the modules. The
proposed method has been applied to analyze the MPP data
from a test facility at the National Renewable Energy
Laboratory (NREL). Our analysis indicates that the solar
modules degraded at a rate of ~0.7 %/year due to
discoloration and weakened
solder bonds. These
conclusions are independently validated by outdoor IV
measurement and on-site
imaging characterization.
Integrated with physics-based degradation models or
machine learning algorithms, the method can also serve to
predict the lifetime of PV systems.
I. INTRODUCTION
As an alternative renewable energy resource, photovoltaics
(PV) has experienced exponential growth over the last several
decades. For investors, an important metric to benchmark the
financial viability of PV against other energy resources is the
levelized cost of electricity (LCOE). However, the current
estimates of the LCOE for PV often rely on the presumption of
a linear performance degradation over time. Unfortunately, this
presumption leads to an inaccurate LCOE because PV
degradations are inherently nonlinear [1]. Also, the rate and
magnitude of PV degradation depends sensitively on cell
technology and vary substantially across geographic locations
[2], [3]. Hence, a "technology-agonistic" in-situ monitoring
method – that characterizes the temporal PV degradation in real
time while taking the meteorological information into account
– can improve our understanding of the technology- and
location-specific degradation rates. This will improve LCOE
estimates and suggest opportunities for reliability-aware
technology improvement.
literature,
There have been many studies on PV reliability reported in
characterization
the
methodologies. These methodologies can be roughly divided
into two groups: off-line and on-line techniques.
different
based
on
the
examine
techniques
Typical off-line
temporal
degradation by periodically and temporarily disconnecting the
solar modules for a detailed characterization. For instance,
Jordan et al. [4] and Sutterlueti et al. [5] inspected the
degradation mechanisms of PV systems by interpreting IV
curves based on the physically-defined five parameter model
and the empirical loss factors model (LFM), respectively.
Additional sophisticated characterization techniques (e.g.,
electroluminescence and infrared imaging) can even yield the
spatial-resolved degradation analysis for fielded solar modules
[6], [7]. Indeed, these off-line methods are incredibly powerful
for degradation characterization; however,
they require
interrupting the normal operation of solar modules at the
maximum power point, hence not suitable for continuous
monitoring.
On-line techniques, on the other hand, rely on information
routinely collected from solar modules. For example, Refs. [8],
[9] have analyzed the on-line temporal evolution of PV
degradation by continuously examining three time-series
performance metrics: (a) DC/GPOA, the ratio of DC power over
the plane-of-array irradiance [10], (b) the performance ratio
(PR), a number between 0 and 1 (under STC conditions) equal
to the ratio between actual energy yield and nameplate rating
[11], (c) the regression PVUSE method that empirically
translates on-site output power to the standard test condition
(STC) [12]. These methods have the advantage that the
modules are not disconnected/interrupted for characterization.
The understanding of the degradation pathways, which is
critical to establishing the fundamental physics of degradation
and promoting reliability-aware design, is still missing from
these analysis.
the method can monitor the reliability of solar modules in real
time.
Another on-line characterization approach involves PV data
analyzed by statistical machine learning algorithms [13]–[15].
Machine Learning has been proved to be a potent tool to
analyze massive data and generate useful insights for different
applications. Nonetheless, the weight functions in these
algorithms are not physically defined, and it can be difficult to
correlate the weights to specific degradation mechanisms.
Moreover, network training necessitates a tremendous amount
of field data spanning across different geographic locations and
technologies as training sets, which are not easily accessible.
Therefore, an on-line methodology that can physically and
continuously track the degradation of PV systems in real time
by interpreting the available field data (and providing insights
obtainable only by off-line techniques) can be a transformative
tool for the PV community.
Inspired by the well-known Suns-Voc method, in this paper
we have developed a simple and powerful strategy to mine the
time-series field data to yield a deep understanding of PV
reliability and identify various degradation pathways. The
Suns-Voc method [16], where one monitors the open-circuit
voltage by manually varying illumination intensity of a solar
simulator (see. Fig. 1), has been demonstrated to be a useful
characterization tool during module development. Obviously, it
cannot apply directly to field data composed exclusively of
maximum power point (MPP) current (𝐼𝑚𝑝) and voltage (𝑉𝑚𝑝)
information. Hence, we propose the Suns-Vmp method that,
by taking advantage of the natural daily variation of sunlight,
can deduce circuit parameters as a function time by fitting the
reconstructed MPP "IV" throughout the day, see Fig. 1. By
systematically and physically mining the streaming MPP data,
Fig. 1 A schematic illustration to explain the working
principles of the Suns-Voc and Suns-Vmp method.
In this paper, we begin by introducing the detailed
methodology of the Suns-Vmp method in Sec. II. In Sec. III,
the Suns-Vmp method is applied to an NREL test facility to
extract the degradation rate and the dominant degradation
modes. Sec. IV discusses the implication of the Suns-Vmp
method on the prediction and design of PV reliability and the
limitation herein. Finally, we summarize the paper in Sec. V.
II. THE SUNS-VMP METHOD
In this section, we will discuss the Suns-Vmp algorithm, as
summarized in Fig. 2. The algorithm has the following four
steps: 1) develop the physics-based equivalent circuit model for
a specific technology; 2) extract pristine (time-zero) circuit
parameters
IV
characteristics; 3) preprocess MPP data to reconstruct IV
characteristics synthetically, and 4) finally, analyze the time-
degradation of circuit parameters for insights regarding the
dominant degradation modes.
datasheet/pre-installation
based
on
A. Step 1: Development and Choice of the Equivalent Circuit
(Compact Model)
Mainstream PV technologies can be categorized into three
groups: 1) p-n homojunction (e.g., c-Si and GaAs), 2) p-i-n
junction (e.g., a-Si and perovskites), and 3) p-n heterojunction
(e.g., CIGS and CdTe). Depending on a particular technology,
we select the corresponding equivalent circuit in the Suns-Vmp
method, see for example, [17] for CIGS, [18] for perovskites,
[19] for silicon heterojunction. Since a solar cell is exposed to
varying illumination intensity and temperature, the equivalent
circuit must be capable of describing the illumination- and
temperature-dependent IV curves.
Fig. 2 The flowchart of the Suns-Vmp method. The analytical
formulation of the five-parameter model is from [20], [33] and
summarized in the supplementary material.
IlluminationNoonMorningEveningSolar SimulatorEveningNoonMorningSuns-Voc(Indoor)Suns-Vmp(Outdoor)Datasheet Fitting Reconstruct IVMPP dataEnvironmental DataVMPTCIMPILPhysics-Based Equivalent CircuitInitial Guessonlyat t = 0Extract Circuit ParametersInitial Guess at ti+1Next Time Step:ti+1 = ti+ Δt(3) Data Preprocessing(2) Pristine Circuit Parameters(4) IV Fitting(1) Equivalent Circuit
Initial fitting to the datasheet (Siemens M55 [30]) for
Fig. 3
time-zero circuit parameters. The extracted circuit parameters
are summarized in the supplementary material.
In this paper, we will demonstrate the Suns-Vmp method on
a c-Si PV system, therefore we make use of the well-known five
parameter model for Si solar modules [20], which explicitly
accounts for the illumination- and temperature-dependencies of
circuit parameters, namely, JPH, J01, J02, RSH, and RS, see Fig. 2.
The complete set of equations and parameter descriptions for
the five parameters is summarized in the supplementary
material (SI). If needed, the five parameter model can be
generalized to include nonlinear shunt resistance [21] and
temperature- and illumination-dependent series resistance [22],
[23].
B. Step 2: Extracting Pristine Module Parameters
Next, we extract the pristine (time-zero) module parameters
(before the module is fielded) as robust initial guesses for the
Suns-Vmp method. We do so by fitting the complete
illumination- and temperature-dependent IV measurements
available from the datasheet or pre-installation measurement.
With the robust initial guesses, we can eliminate multiple
solutions in the sequential IV fitting process, see Fig. 3. Typical
datasheet usually provides a set of full IV measurement under
various
temperature conditions which
guarantee the uniqueness of the extracted circuit parameters and
consequently the robustness of the initial guess.
C. Step 3: Preprocessing MPP Data
illumination and
the
After obtaining
time-zero circuit parameters, we
(a) A schematic showing different three light components
Fig. 1
construct – at any time during the onsite operation – a synthetic
(direct, diffuse, and Albedo light) and an elevated bifacial solar panel
IV curve by sampling MPP data over a given period (typically
with self-shading. (b) Hourly intensity of GHI, DNI, and DHI on June.
10th at Washington DC. (c) Hourly output power of east-west facing
2-3 days, referred as measurement window hereafter). Recall
vertical bifacial (Vert. bifacial), conventional south facing bifacial
that in the Suns-Voc measurement [24], [25], one traces the
(Con. bifacial) titled at 43o, and monofacial solar panels titled at 37o.
open circuit voltage of solar cells, through deliberately varying
All of them have an elevation of 0.5 m above the ground.
the intensity of the solar simulator, to construct the IV curve in
the absence of series resistance. In the Suns-Vmp method,
however, we take advantage of the natural temporal variation
of the sunlight (the plane-of-array irradiance: 𝐺𝑃𝑂𝐴) and the cell
temperature (𝑇𝐶 ) to track the maximum power point. Hence,
due to the changing GPOA and TC, the module output 𝐼𝑚𝑝 and
𝑉𝑚𝑝 (operating current and voltage at the maximum power
point, respectively) increase from morning to noon then
decrease from noon to evening, see Fig. 4(a). For example, if
the data is recorded every 10 minutes of 8 diurnal hours over a
3-day measurement window, then 144 data points of four
variables (i.e. 𝐺𝑃𝑂𝐴, 𝑇𝐶, 𝐼𝑚𝑝, 𝑉𝑚𝑝) are available to calculate the
circuit parameters of the compact model, namely, calibrating
the circuit parameters until the MPP IV is reproduced as shown
in Fig. 4(b). Note that Suns-Vmp method does not interrupt the
normal module operation by disconnecting solar modules for
IV sweeps or deviating them from the MPP bias [16], [26]; thus
the technique empowers characterization of solar modules in
real-time operation.
In the Suns-Vmp methodology, to reduce uncertainties in the
extraction, we also explicitly preprocess the data to account for
1) cell-to-module temperature difference, 2) spectral mismatch
between pyranometer and solar modules, and 3) reflection loss
as a function of time. The specific steps are summarized in the
SI. Also, while the basic algorithm is easy to understand, it is
important to realize that the (𝐺𝑃𝑂𝐴, 𝑇𝐶, 𝐼𝑚𝑝, 𝑉𝑚𝑝) may involve
noisy or corrupted data. In this case, the window duration must
be choosen judiciaously and the corrupted data must be
rejected, for a robust parameter extraction of the compact
Fig. 4
(a) Three-day MPP and environmental data (circles)
from 11/09/2002 to 11/11/2002 of the test facility in Sec. III.
The fitting results of the MPP data (solid lines) using the Suns-
Vmp method
illustration of
is also present. (b) An
reconstructing "IV" from the MPP data in (a).
11/09/200211/10/200211/11/200211/12/2002DateModelReal-time data(a)Environmental DataMPP DataReconstructed "IV" from 11/09/2002 to 11/11/2002(b)GPOA(W/m2)TC(K)1000 W/m225 oC1000 W/m247 oC500 W/m225 oC after removing voltage stress for potential induced degradation
[28], [29]), such recovery is expected to be negligible due to
constant environmental stress (e.g., thermal cycling, moisture
exposure) applied on the operating solar modules.
For any inverse algorithm such as the Suns-Vmp method, one
must ensure that the uniqueness of the degradation analysis.
Hence, we present a sensitivity analysis of these two algorithm
parameters,
i.e., measurement window and maximum
degradation rate of circuit parameters, on the final extraction of
degradation rates, see Fig. 5. Our results show that moderate
change in the algorithm parameters in the Suns-Vmp method
does not interfere with the final results – the deduced
degradation rates of performance metric remain unique.
In the next section, we will demonstrate the Suns-Vmp
method on an NREL test facility with recorded field data to
analyze the degradation of solar modules in real time. The
analysis will reveal the possible root causes of power losses by
physically
circuit
parameters.
time-dependent
interoperating
the
III. APPLICATION TO FIELD DATA
A. Introduction to Field Data
The studied PV system (No: NREL x-Si #7) perches at the
west side of the Solar Energy Research Facility (SERF)
building at the National Renewable Energy Laboratory
(NREL), Golden, CO, USA. It comprises two arrays with
negative and positive monopole, each of which consists of five
strings with 14 x-Si Siemens M55 solar modules [30] totaling
to around 7.42 kW capacity. In 2007, a negatively grounded
inverter replaced the previous bipolar inverter, but we maintain
the bipolar naming convention (negative versus positive) in this
paper. The modules are 45o tilted and oriented 22o east of south.
All the onsite MPP and environmental data (illumination and
module temperature) including the metadata were retrieved
from the publicly accessible NREL PV Data Acquisition
(PVDAQ) database [31] with time resolution spanning from 1
min to 15 min. The analyzed field data is from 05/13/1994 to
12/31/2014. Three measurments of module temperature were
initially recorded by thermocouples attached to the backsheets
but significantly inconsistency was found after the eighth year.
Therefore, we applied the calibrated Faiman model [32] to
obtain module temperature. In addition to continuous MPP
data, outdoor IV measurements were also carried out at the
array level using a portable Daystar I–V tracer. These IV data
sets help us validate the analysis obtained from the Suns-Vmp
method. More details on this PV systems can be found in SI.
model. Hence, we have developed a physics-based self-filtering
algorithm to preprocess the data as follows before fitting (see
the SI material for additional details).
it
is
to
long
enough
contain
The measurement window of MPP data must be chosen such
that
sufficient
illumination/temperature variations, but short enough such that
the module does not degrade significantly within the window.
The time-scale of degradation processes is slow [13], thereby
the circuit parameters can be assumed to be constant over the
course of a few days. Hence, the recommended measurement
window of MPP data can be up several days (e.g., three days in
Fig. 4), as long as there exists sufficient variation in
illumination and temperature to reconstruct the MPP IV. In the
case of catastrophic degradation (such as partial shading
degradation in thin-film solar modules [7]), the extracted circuit
parameters become the average value of pre- and post-
degradation values over time.
and
After
(PSO)
D. Step 4: MPP IV Fitting Algorithm:
IV
reconstructing MPP
preprocessing
environmental data, we proceed with using rigorous fitting
algorithms to model the measured MPP data and extract circuit
parameters. In this paper, we have used the nonlinear least-
squares fitting algorithm and bio-inspired particle swarm
optimization
("lsqcurvefit" and "particleswarm"
functions in Matlab® [27], respectively), both of which have
been found to give identical results. Note that both fitting
algorithms require a lower and upper bound of each circuit
parameter at each time step. In our analysis, circuit parameters
are assumed to degrade monotonically as a function of time (i.e.,
no recovery) with a maximum degradation rate of 1 %/day,
except for the short-circuit current JPh. Hence, given the used
length of measurement window, the upper and lower bound can
be determined. Since the short-circuit current may fluctuate
abruptly due to soiling and precipitation, the upper and lower
bound thereof are set to be the datasheet short-circuit current
and zero, respectively. Even though recovery of certain
degradation pathways is possible (e.g., output power recovers
Fig. 5 Degradation rate of performance metrics of the
Fig. 8 Temporal STC efficiencies calculated by the Suns-
negative array as a function of different settings (i.e.,
Vmp and DC/GPOA methods for the arrays with a negative (a)
measurement window and maximum degradation rate of circuit
and positive monopole (b), respectively.
parameters) in the Suns-Vmp method.
DC/GPOASuns-Vmp(a)(b)NegPos3 Days & 1%3 Days & 2%4 Days & 1%2 Days & 1%
to mitigate
(a) 20-year data of IMP and VMP of the negative
Fig. 6
monopole. (b) One-day data exhibits the existence of corrupted
outlier points.
Figure 6 displays the example data while highlighting the
two major challenges of analyzing this field data – 1) several
gaps even up to 5 years of absent field data and 2) corrupted
data with outliers possibly due to instrumentation error, inverter
clipping, weather condition, etc. First,
the
uncertainty in deducing the circuit parameters induced by
missing data, the Suns-Vmp method makes use of the results
from the previous time step as initial guesses and establishes
the upper/lower bounds with a preset maximum change rate
when fitting the MPP IV. Second, we need a self-consistent
scheme to detect and remove these outliers. Toward this goal,
we have created a continuous self-filtering algorithm as
summarized in the SI. Enabled by these techniques, the Suns-
Vmp method can retain excellent error control, i.e., the mean
absolute percentage error (MAPE) is less than 5% for both
Vmp and Imp throughout the entire 20-year analysis.
B. Results and Validation
Figure 7 summarizes the extracted circuit parameters of the
negative array by fitting the five-parameter model (see Fig. 2)
in [20], [33] to the MPP data with a three-day measurement
window over a span of 20 years (from 1994 to 2014). The
positive array also shows a very similar result, therefore not
(a) A schematic showing different three light components
Fig. 1
(direct, diffuse, and Albedo light) and an elevated bifacial solar panel
with self-shading. (b) Hourly intensity of GHI, DNI, and DHI on June.
10th at Washington DC. (c) Hourly output power of east-west facing
vertical bifacial (Vert. bifacial), conventional south facing bifacial
(Con. bifacial) titled at 43o, and monofacial solar panels titled at 37o.
Fig. 7 The extracted circuit parameters under standard test
condition of the five-parameter model for the negative array as
a function of time. Notations: JPH is the maximum photocurrent
density; J01 is the reverse saturation current density with ideality
factor of 1; J02 is the reverse saturation current density with
ideality factor of 2; RSH is the shunt resistance; RS is the series
resistance. JPH is corrected so that it monotonically decreases
with time (red dashed line).
included here. The maximum photocurrent (JPH) fluctuates
possibly due to the accumulation of dust/snow [34] or
recalibration of the pyranometer during 20 years. However, it
is expected that this fluctuation in JPH does not disturb the
extraction of other parameters, since the five-parameter model
assumes voltage-independent JPH and therefore the fluctuation
will just shift the IV in Fig. 4 but not change the underlying IV
characteristics (shape). Remarkably, it appears that all the
circuit parameters in Fig. 7 were degrading (e.g., shunt
resistance (RSH) reduces, and series resistance (RS) increases).
To quantify the degradation rate, we calculate the efficiency at
standard test condition (STC) at each time step, see Fig. 8.
to DC/GPOA method.
Remarkably, the extracted STC efficiency by the Suns-Vmp
method compares well with that of the conventional DC/GPOA
method [35], showing both the negative and positive arrays near
their warranty lifetime (80% of initial efficiency). However, the
result obtained from the DC/GPOA method shows greater
fluctuation than the Suns-Vmp method due to 1) the empirical
approaches to filtering outliers and 2) linear temperature-
correction of real-time output power to STC by a constant
temperature coefficient (which changes over time). Because the
Suns-Vmp method uses a physics-based equivalent circuit for
1: Comparison
Validation
Fig. 1
(a) A schematic showing different three light components
(direct, diffuse, and Albedo light) and an elevated bifacial solar panel
with self-shading. (b) Hourly intensity of GHI, DNI, and DHI on June.
10th at Washington DC. (c) Hourly output power of east-west facing
TC(K)(a)(b)Missing Data
loss
is recoverable) when calculating
Fig. 8 Temporal STC efficiencies calculated by the Suns-
Vmp and DC/GPOA methods for the arrays with a negative (a)
and positive monopole (b), respectively.
outlier filtering and temperature correction, the fluctuation is
substantially reduced. Note that, for the Suns-Vmp method, we
correct JPH so that it monotonically decreases with time (i.e.,
soiling
the STC
efficiency, see Fig. 7.
Validation 2: Outdoor IV Measurement. To further validate
the Suns-Vmp method, we benchmark the obtained results
against those characterized by the periodic outdoor IV
measurement through 20 years. Figure 9 shows the comparison
between real-time (not STC) PV performance metrics
calculated by circuit parameters deduced by the Suns-Vmp and
direct outdoor IV measurements. Indeed, we find great
consistencies (less than 4% MAPE) between these two
methods, which corroborates the accuracy of the extraction by
the Suns-Vmp method.
Validation 3: Parameter degradation Rates. Besides the
performance metric, we also benchmark the rate of change of
the performance metrics estimated from the Suns-Vmp method
against outdoor IV from [36] in Fig. 10 (top), which again are
Fig. 10 Rate of change of the performance metrics (top) and
series resistance RS (bottom) of the analyzed PV systems via the
Suns-Vmp method and outdoor IV measurement.
in good agreement. The error bars are calculated within 95%
confidence interval. The degradation rate of the efficiencies for
both the negative and positive arrays are around 0.7%/Year. It
is noteworthy that the efficiency degradation may be primarily
attributed to the reduction in fill factor (-0.6 to -0.4 %/Year),
while Voc and Isc only worsen slightly.
We attribute this degradation to the increased series
resistance, which erodes fill factor without substantially
affecting Voc and Isc. Both the Suns-Vmp and outdoor IV
measurement reveal the rapid increment of series resistance at
the rate of 5 – 10%/year as shown in Fig. 10 (bottom), which
confirms our conjecture of series-resistance induced efficiency
degradation.
Validation 4: Onsite inspection. Next, we will deconvolve
and quantify the power losses ascribed to each circuit parameter
to identify the predominant physical degradation pathways. As
shown in Fig. 11 (a), we deconvolve the power losses
associated with each parameter for the negative array. The key
observations are threefold:
1) At the end of 20 years, Fig. 11 (a) elucidates that the
increased series resistance is the dominant contributor to
efficiency reduction for both the negative and positive
polarities. Remarkably, the on-site infrared image in Fig. 11 (b)
exhibits localized hot spots caused by solder bond failure, in
accord with our deconvolution analysis of increasing series
resistance. It is generally known that the failure of solder bonds
Fig. 1
(a) A schematic showing different three light components
(direct, diffuse, and Albedo light) and an elevated bifacial solar panel
is because of thermal stress induced by the different thermal
with self-shading. (b) Hourly intensity of GHI, DNI, and DHI on June.
expansion coefficients of solder joints and components during
10th at Washington DC. (c) Hourly output power of east-west facing
repeated thermal cycles [37], [38]. Therefore, solder bonds fail
vertical bifacial (Vert. bifacial), conventional south facing bifacial
(Con. bifacial) titled at 43o, and monofacial solar panels titled at 37o.
All of them have an elevation of 0.5 m above the ground.
Fig. 1
(a) A schematic showing different three light components
(direct, diffuse, and Albedo light) and an elevated bifacial solar panel
with self-shading. (b) Hourly intensity of GHI, DNI, and DHI on June.
10th at Washington DC. (c) Hourly output power of east-west facing
vertical bifacial (Vert. bifacial), conventional south facing bifacial
(Con. bifacial) titled at 43o, and monofacial solar panels titled at 37o.
All of them have an elevation of 0.5 m above the ground.
Fig. 9 Comparison of performance metric generated by the
Suns-Vmp method and outdoor array IV measurement for the
negative array. The mean absolute percentage errors (MAPE)
are also labelled in each plot.
Eff.FFVOCISCRate of Change (%/Year)Suns-VmpOutdoor IVSuns-VmpOutdoor IVNegPosRate of Change (%/Year)MAPE=3.5%MAPE=2.8%MAPE=3.1%MAPE=1.2%Outdoor IVSuns-Vmp(a)(b)(c)(d)DC/GPOASuns-Vmp(a)(b)NegPos
Fig. 11 (a) Temporal degradation deconvolution with respect
to circuit parameters for the negative polarity. (b) IR image
shows a hot spot caused by solder bond failure. (c) Picture
shows that most cells suffer from discoloration in the center.
*(b) and (c) are obtained from [36].
(crack) at the cycle of failure in a stepwise fashion [39]. Indeed,
the incremental time signature of the series resistance is
stepwise in the Suns-Vmp analysis, see Fig. 11 (a).
2) Discoloration of the encapsulants can be expected because
of the relatively high ultraviolet light concentration at Denver
(altitude of ~1800 m) [40]. Indeed, a photograph of the solar
modules in the field shows that the majority of the solar cells
suffer from discoloration, see Fig. 11 (c). Meanwhile,
notwithstanding the JPH fluctuation shown in Fig. 7, our
deconvolution results also manifests a symmetric decrease of
JPH and ascribes a significant amount of power loss (~4%) to
JPH reduction, an indicator of discoloration. This agreement
again confirms the PV degradation diagnosed by the Suns-Vmp
method. It is noteworthy that the photocurrent reduction due to
discoloration has occurred within the first year of installation.
Another study has also found early advent of discoloration, i.e.,
discoloration has been seen in 50% of the solar module less than
five years old [41].
3) The operating voltage of the modules is only around 200
V; therefore, the efficiency degradation by potential-induced
degradation (PID) is expected to be insignificant [42]. Indeed,
our result confirms this conjecture by showing that only ~3%
power loss is due to shunting (RSH) and increased recombination
currents (J01 and J02), both of which are effective indicators for
PID [43], [44].
As demonstrated here,
the Suns-Vmp allows us
to
(a) A schematic showing different three light components
Fig. 1
quantitatively and qualitatively diagnose the pathology of
(direct, diffuse, and Albedo light) and an elevated bifacial solar panel
with self-shading. (b) Hourly intensity of GHI, DNI, and DHI on June.
degraded solar modules exposed in the field by analyzing and
10th at Washington DC. (c) Hourly output power of east-west facing
interpreting the time signature of individual circuit parameters.
vertical bifacial (Vert. bifacial), conventional south facing bifacial
All the results have been validated by both outdoor IV
(Con. bifacial) titled at 43o, and monofacial solar panels titled at 37o.
measurement and on-site characterization.
All of them have an elevation of 0.5 m above the ground.
IV. DISCUSSION
In the previous section, we have applied the Suns-Vmp
method to an NREL test facility and demonstrated its capability
of analyzing the degradation of solar modules in real time.
Next, we discuss the potential use of the time-dependent
parameters obtained through the analysis and limitations of the
approach.
A. Geography and
technology-specific reliability-aware
design
The underlying physical degradation mechanisms of PV are
strongly contingent on local meteorological factors and
different technologies, e.g., solar modules exposed in humid
regions are more susceptible to contact corrosion [45], and
monolithic thin-film solar modules are vulnerable to partial
shading degradation [7]. Similarly, modules more likely to
suffer from PID should adopt Corning® Willow™ Glass to
impede ion migration [46]. Therefore, ideally, module design
ought to be geography- and technology-dependent. However,
solar modules are often overdesigned for reliability (perhaps at
a considerable cost) so that they can survive a broad range of
weather conditions. This is due to the lack of comprehensive
understanding of local degradation. The Suns-Vmp method
offers an opportunity to efficiently diagnose the degradation
pathways of fielded solar modules of different technologies
across the entire world. The results can be ultimately collected
in a global database, allowing the manufacturers to design and
produce the next generation reliable-aware PV with maximized
durability.
B. More accurate long-term reliability prediction
Accurate prediction of long-term energy production by PV
systems is crucial to evaluating the bankability thereof. Various
degradation pathways depend nonlinearly on stress time and
local stress factors (irradiance, voltage, moisture, temperature).
Therefore, it is difficult to predict future energy yields based on
empirical linear degradation models [2]. In this regard, the
Suns-Vmp method can facilitate accurate reliability prediction.
Recently, several physics-based degradation models have been
developed that can directly map various PV degradation modes
(e.g., corrosion, PID, yellowing) to the temporal behavior of
circuit parameters [47], [48]. The extracted circuit parameters
by the Suns-Vmp method can be used to calibrate these
degradation models (e.g., moisture diffusion coefficient for
corrosion). Integrated with the weather forecast, the calibrated
degradation models will predict the lifespan of solar modules.
Alternatively, the time-dependent circuit parameters can train
machine learning algorithms; the trained machine learning
algorithms [13] can predict PV lifetime. The validity of these
predictive approaches, however, remains an interesting open
question and requires more rigorous research efforts.
C. Guidance for collection of field data
The Suns-Vmp methodology highlights the importance of
physics-based modeling in creating databases. For example, we
(b)(c)(a) have seen fitting of the pristine module characteristics requires
temperature- and illumination-dependent IV measurement to
ensure a robust and unique initial guess. Second, we have noted
that weather data may be corrupted or missing. Thus it is
important for PV databases
to contain complementary
information from multiple sources [49]. Finally, compact
model parameters offer an important recipe for improving data
compression and computational efficiency;
the model
parameters can diagnose the module by only deciphering the
stored Vmp-Imp information (a byproduct data of normal
operation at maximum power point) for the entire duration.This
eliminates the need for deliberate measurement of massive IV
data [15] and time-consuming collection of field data [41].
D. Intra-string variability
Process-induced variability can lead to performance variation
in the cell, module, or array levels [19], [50], [51], especially
for the thin-film PV where binning is not possible. Similarly,
various degradation modes introduce local variability as well.
For example, non-uniform degradation (e.g., cells adjacent to
module edges are more prone to contact corrosion than those
located away from the edges [52]; solar modules close to the
negative array are more susceptible to PID [53]), etc. As
implemented, the Suns-Vmp method uses a single equivalent
circuit to analyze a string consisting of multiple modules and
thus accounts for "average" variability/degradation. As a result,
it is critical to investigate how performance variability can
potentially affect the accuracy of the Suns-Vmp method.
Therefore, we have tested Suns-Vmp under various scenarios
of performance variability, and the results are listed in the SI.
Remarkably, our findings highlight that the circuit parameters
extracted by the Suns-Vmp method are still valid to interpret
PV degradation with moderate non-uniformity. Affected by
severe non-uniformity, however, the Suns-Vmp method may
not be able to identify the primary circuit parameters
contributing to power losses. For instance, the Suns-Vmp
method could attribute the predominant degradation to the
increased recombination current (J01 and J02), and series
resistance RS , whereas the actual degradation is due to reduced
shunt resistance RSH. For these cases, it will be important to
represent the string by a few equivalent circuit models. Despite
the
the following
considerations are expected to simplify the calibration process:
1) availability of time-zero information of each module, 2) the
large amount of data available within the measurement
window, and 3) several degradation modes (e.g., yellowing) are
expected to affect all the modules uniformly, while others (e.g.,
PID) are dominated by a few modules. Ability to account for
non-uniform degradation will be an important direction of
future research on this topic.
the parameter number,
increase
in
V. CONCLUSION
To summarize, we have presented a novel method, i.e., the
Suns-Vmp method, for analyzing the PV degradation:
1. The Suns-Vmp method enables in-situ monitoring and
diagnosis of PV reliability in real time by systemically and
physically mining the time-series MPP data. The method
can extract physically defined circuit parameters by fitting
IVs consisting of
the varying MPP data over a
characterization window. The extracted circuit parameters
can be used to estimate the STC efficiency, quantitively
deconvolute PV degradation, and identify the dominant
degradation pathways.
2. We have demonstrated the Suns-Vmp method by analyzing
MPP data from an NREL test facility, where physics-based
circuit parameters and efficiency of the solar modules have
been extracted as a function of time. Independent outdoor
IV measurements have systemically validated our results.
Our analysis suggests that the PV system degrades at a rate
of 0.7%/Year, primarily due to reduced short-circuit
current and increased series resistance most likely caused
by discoloration and weakened solder bond, respectively.
The on-site optical photograph and IR image indeed
substantiate our interpretation of the physical degradation
pathways, i.e., discoloration and solder bond failure.
3. The analysis of deconvoluting the underlying degradation
pathways by the Suns-Vmp method can deepen the current
understanding of technology- and geographic-dependent
degradation, and inspire more robust environment-specific
designs for the next-generation "reliability-aware" solar
modules. The Suns-Vmp method can be used to calibrate
physics-based degradation models as well as train machine
learning algorithms, both of which can then predict power
degradation of PV and improve the evaluation of
"bankability."
ACKNOWLEDGMENT
This work was supported by the US-India Partnership to
Advance Clean Energy-Research (PACE-R) for the Solar
Energy Research Institute for India and the United States
(SERIIUS), and the DEEDS program by the National Science
Foundation under award #1724728. The authors would like to
thank Haejun Chung, Reza Asadpour, and Dr. Mohammad
Ryyan Khan for helpful discussion, Dr. Chris Deline and Dr.
Dirck Jordan for providing IV measurement, as well as Prof.
Mark S. Lundstrom and Prof. Peter Bermel for kind guidance.
REFERENCES
[1]
[2]
D. C. Jordan, T. J. Silverman, B. Sekulic, and S. R.
Kurtz, "PV degradation curves: non-linearities and
failure modes," Prog. Photovoltaics Res. Appl., vol.
15, no. February 2013, pp. 659–676, 2016.
D. C. Jordan, S. R. Kurtz, K. VanSant, and J.
[3]
[4]
[5]
[6]
[7]
[8]
[9]
Newmiller, "Compendium of photovoltaic
degradation rates," Prog. Photovoltaics Res. Appl.,
vol. 24, no. 7, pp. 978–989, Jul. 2016.
A. Ndiaye, A. Charki, A. Kobi, C. M. F. Kébé, P. A.
Ndiaye, and V. Sambou, "Degradations of silicon
photovoltaic modules: A literature review," Sol.
Energy, vol. 96, pp. 140–151, Oct. 2013.
D. C. Jordan, C. Deline, S. Johnston, S. R. Rummel,
B. Sekulic, P. Hacke, S. R. Kurtz, K. O. Davis, E. J.
Schneller, X. Sun, M. A. Alam, and R. A. Sinton,
"Silicon Heterojunction System Field Performance,"
IEEE J. Photovoltaics, pp. 1–6, 2017.
J. Sutterlueti, S. Ransome, J. Stein, and J. Scholz,
"Improved PV performance modelling by combining
the PV_LIB toolbox with the Loss Factors Model
(LFM)," in 2015 IEEE 42nd Photovoltaic Specialist
Conference (PVSC), 2015, pp. 1–6.
T. Potthoff, K. Bothe, U. Eitner, D. Hinken, and M.
Köntges, "Detection of the voltage distribution in
photovoltaic modules by electroluminescence
imaging," Prog. Photovoltaics Res. Appl., vol. 18, no.
2, pp. 100–106, Mar. 2010.
T. J. Silverman, M. G. Deceglie, X. Sun, R. L. Garris,
M. A. Alam, C. Deline, and S. Kurtz, "Thermal and
Electrical Effects of Partial Shade in Monolithic Thin-
Film Photovoltaic Modules," IEEE J. Photovoltaics,
vol. 5, no. 6, pp. 1742–1747, Nov. 2015.
G. Belluardo, P. Ingenhoven, W. Sparber, J. Wagner,
P. Weihs, and D. Moser, "Novel method for the
improvement in the evaluation of outdoor
performance loss rate in different PV technologies and
comparison with two other methods," Sol. Energy,
vol. 117, pp. 139–152, Jul. 2015.
D. C. Jordan and S. R. Kurtz, "The dark horse of
evaluating long-term field performance-Data
filtering," IEEE J. Photovoltaics, vol. 4, no. 1, pp.
317–323, 2014.
[10] D. C. Jordan and S. Kurtz, "PV degradation risk," in
the World Renewable Energy Forum, 2012.
[11] H. Haeberlin and C. Beutler, "Normalized
Representation of Energy and Power for Analysis of
Performance and On-line Error Detection in PV-
Systems," in European Photovoltaic Solar Energy
Conference and Exhibition, 17th, 1995.
[12] C. Jennings, "PV module performance at PG&E," in
Conference Record of the Twentieth IEEE
Photovoltaic Specialists Conference, 1988, pp. 1225–
1229 vol.2.
[13] R. H. French, R. Podgornik, T. J. Peshek, L. S.
Bruckman, Y. Xu, N. R. Wheeler, A. Gok, Y. Hu, M.
A. Hossain, D. A. Gordon, P. Zhao, J. Sun, and G.-Q.
Zhang, "Degradation science: Mesoscopic evolution
and temporal analytics of photovoltaic energy
materials," Curr. Opin. Solid State Mater. Sci., vol.
19, no. 4, pp. 212–226, Aug. 2015.
T. J. Peshek, J. S. Fada, Y. Hu, Y. Xu, M. A. Elsaeiti,
E. Schnabel, M. Köhl, and R. H. French, "Insights into
[14]
metastability of photovoltaic materials at the
mesoscale through massive I–V analytics," J. Vac.
Sci. Technol. B, Nanotechnol. Microelectron. Mater.
Process. Meas. Phenom., vol. 34, no. 5, p. 50801, Sep.
2016.
[15] Y. Hu, V. Y. Gunapati, P. Zhao, D. Gordon, N. R.
Wheeler, M. A. Hossain, T. J. Peshek, L. S.
Bruckman, G.-Q. Zhang, and R. H. French, "A
Nonrelational Data Warehouse for the Analysis of
Field and Laboratory Data From Multiple
Heterogeneous Photovoltaic Test Sites," IEEE J.
Photovoltaics, vol. 7, no. 1, pp. 230–236, Jan. 2017.
[16] M. J. J. Kerr, A. Cuevas, and R. A. A. Sinton,
"Generalized analysis of quasi-steady-state and
transient decay open circuit voltage measurements," J.
Appl. Phys., vol. 91, no. 1, p. 399, 2002.
[17] X. Sun, T. Silverman, R. Garris, C. Deline, and M. A.
Alam, "An Illumination- and Temperature-Dependent
Analytical Model for Copper Indium Gallium
Diselenide (CIGS) Solar Cells," IEEE J.
Photovoltaics, vol. 1, pp. 1–10, 2016.
[18] X. Sun, R. Asadpour, W. Nie, A. D. Mohite, and M.
A. Alam, "A Physics-Based Analytical Model for
Perovskite Solar Cells," IEEE J. Photovoltaics, vol. 5,
no. 5, pp. 1389–1394, Sep. 2015.
[19] R. V. K. Chavali, E. C. Johlin, J. L. Gray, T.
Buonassisi, and M. A. Alam, "A Framework for
Process-to-Module Modeling of a-Si/c-Si (HIT)
Heterojunction Solar Cells to Investigate the Cell-to-
Module Efficiency Gap," IEEE J. Photovoltaics, vol.
6, no. 4, pp. 1–13, Jul. 2016.
[20] M. Hejri, H. Mokhtari, M. R. Azizian, M. Ghandhari,
and L. Soder, "On the Parameter Extraction of a Five-
Parameter Double-Diode Model of Photovoltaic Cells
and Modules," IEEE J. Photovoltaics, vol. 4, no. 3,
pp. 915–923, May 2014.
S. Dongaonkar, J. D. Servaites, G. M. Ford, S. Loser,
J. Moore, R. M. Gelfand, H. Mohseni, H. W.
Hillhouse, R. Agrawal, M. A. Ratner, T. J. Marks, M.
S. Lundstrom, and M. A. Alam, "Universality of non-
Ohmic shunt leakage in thin-film solar cells," J. Appl.
Phys., vol. 108, no. 12, p. 124509, Dec. 2010.
[21]
[22] K. Lee, "Improving the PV Module Single-Diode
Model Accuracy with Temperature Dependence of the
Series Resistance," in IEEE 44th Photovoltaic
Specialist Conference (PVSC), 2017.
[23] R. V. K. Chavali, J. E. Moore, X. Wang, M. A. Alam,
M. S. Lundstrom, and J. L. Gray, "The Frozen
Potential Approach to Separate the Photocurrent and
Diode Injection Current in Solar Cells," IEEE J.
Photovoltaics, vol. 5, no. 3, pp. 865–873, May 2015.
[24] M. Wolf and H. Rauschenbach, "Series resistance
effects on solar cell measurements," Adv. Energy
Convers., vol. 3, no. 2, pp. 455–479, Apr. 1963.
[25] R. V. K. Chavali, J. V. Li, C. Battaglia, S. De Wolf, J.
L. Gray, and M. A. Alam, "A Generalized Theory
Explains the Anomalous Suns–VOC Response of Si
Heterojunction Solar Cells," IEEE J. Photovoltaics,
vol. 7, no. 1, pp. 169–176, Jan. 2017.
Components and Technology Conference, 2008, pp.
139–145.
[26] M. G. G. Deceglie, T. J. J. Silverman, B. Marion, and
S. R. R. Kurtz, "Real-Time Series Resistance
Monitoring in PV Systems Without the Need for I – V
Curves," IEEE J. Photovoltaics, vol. 5, no. 6, pp.
1706–1709, 2015.
"Matlab2016a." The MathWorks Inc, Natick, MA,
2014.
P. Hacke, S. Spataru, S. Johnston, K. Terwilliger, K.
VanSant, M. Kempe, J. Wohlgemuth, S. Kurtz, A.
Olsson, and M. Propst, "Elucidating PID Degradation
Mechanisms and In Situ Dark I–V Monitoring for
Modeling Degradation Rate in CdTe Thin-Film
Modules," IEEE J. Photovoltaics, vol. 6, no. 6, pp.
1635–1640, Nov. 2016.
P. Hacke, S. Spataru, K. Terwilliger, G. Perrin, S.
Glick, S. Kurtz, and J. Wohlgemuth, "Accelerated
Testing and Modeling of Potential-Induced
Degradation as a Function of Temperature and
Relative Humidity," IEEE J. Photovoltaics, vol. 5, no.
6, pp. 1549–1553, Nov. 2015.
"Siemans Solar Panels Installation Guide." [Online].
Available:
http://iodlabs.ucsd.edu/dja/codered/engineering/proce
dures/solarPower/siemens solar panels.pdf.
"PVDAQ (PV Data Acquisition)." [Online].
Available: http://developer.nrel.gov/docs/solar/pvdaq-
v3/.
[27]
[28]
[29]
[30]
[31]
[32] D. Faiman, "Assessing the outdoor operating
temperature of photovoltaic modules," Prog.
Photovoltaics Res. Appl., vol. 16, no. 4, pp. 307–315,
Jun. 2008.
[33] M. T. Boyd, S. a. Klein, D. T. Reindl, and B. P.
Dougherty, "Evaluation and Validation of Equivalent
Circuit Photovoltaic Solar Cell Performance Models,"
J. Sol. Energy Eng., vol. 133, no. 2, p. 21005, 2011.
[34] M. G. Deceglie, M. Muller, Z. Defreitas, and S. Kurtz,
"A Scalable Method for Extracting Soiling Rates from
PV Production Data," Pvsc, no. June, 2016.
[35] D. C. Jordan and S. R. Kurtz, "2012 PV degradation
risk," in Technical Report in World Renewable Energy
Forum.
[36] D. C. Jordan, B. Sekulic, B. Marion, and S. R. Kurtz,
"Performance and Aging of a 20-Year-Old Silicon PV
System," IEEE J. Photovoltaics, vol. 5, no. 3, pp.
744–751, May 2015.
[37] U. Itoh, M. Yoshida, H. Tokuhisa, K. Takeuchi, and
Y. Takemura, "Solder joint failure modes in the
conventional crystalline si module," Energy Procedia,
vol. 55, no. 3, pp. 464–468, 2014.
[38] Y.-L. Shen, "Numerical Study of Solder Bond Failure
[41]
[40] M. D. Kempe, "Ultraviolet light test and evaluation
methods for encapsulants of photovoltaic modules,"
Sol. Energy Mater. Sol. Cells, vol. 94, no. 2, pp. 246–
253, Feb. 2010.
S. Chattopadhyay, R. Dubey, V. Kuthanazhi, J. J.
John, C. S. Solanki, A. Kottantharayil, B. M. Arora,
K. L. Narasimhan, V. Kuber, J. Vasi, A. Kumar, and
O. S. Sastry, "Visual Degradation in Field-Aged
Crystalline Silicon PV Modules in India and
Correlation With Electrical Degradation," IEEE J.
Photovoltaics, vol. 4, no. 6, pp. 1470–1476, Nov.
2014.
J. Hattendorf., R. Löw, W.-M. Gnehr, L. Wulff, M. C.
Koekten, D. Koshnicharov, A. Blauaermel, and J. A.
Esquivel, "Potential Induced Degradation in Mono-
Crystalline Silicon Based Modules: An Acceleration
Model," in 27th EU PVSEC, 2012, pp. 3405–3410.
[42]
[43] D. Lausch, V. Naumann, O. Breitenstein, J. Bauer, A.
Graff, J. Bagdahn, and C. Hagendorf, "Potential-
induced degradation (PID): Introduction of a novel
test approach and explanation of increased depletion
region recombination," IEEE J. Photovoltaics, vol. 4,
no. 3, pp. 834–840, 2014.
[44] V. Naumann, D. Lausch, A. Hähnel, J. Bauer, O.
Breitenstein, A. Graff, M. Werner, S. Swatek, S.
Grosser, J. Bagdahn, and C. Hagendorf, "Explanation
of potential-induced degradation of the shunting type
by Na decoration of stacking faults in Si solar cells,"
Sol. Energy Mater. Sol. Cells, vol. 120, no. PART A,
pp. 383–389, 2014.
"All-India India Survey of Photovoltaic Module
Degradation : 2013."
J. Oh, G. TamizhMani, S. Bowden, and S. Garner,
"Surface Disruption Method With Flexible Glass to
Prevent Potential-Induced Degradation of the
Shunting Type in PV Modules," IEEE J.
Photovoltaics, no. 2156, pp. 1–6, 2016.
[45]
[46]
[47] R. Asadpour, R. V. K. V. K. Chavali, and M. A. A.
Alam, "Physics-Based computational modeling of
moisture ingress in solar modules: Location-specific
corrosion and delamination," in 2016 IEEE 43rd
Photovoltaic Specialists Conference (PVSC), 2016,
pp. 0840–0843.
P. Bermel, R. Asadpour, C. Zhou, and M. A. Alam,
"A modeling framework for potential induced
degradation in PV modules," in SPIE, 2015, p.
95630C.
[48]
[49] B. Zhao, X. Sun, M. A. Alam, and M. R. Khan,
in Photovoltaic Modules," Procedia Eng., vol. 139,
pp. 93–100, 2016.
[50]
[39] V. Vasudevan and Xuejun Fan, "An acceleration
model for lead-free (SAC) solder joint reliability
under thermal cycling," in 2008 58th Electronic
"Purdue University Meteorological Tool." [Online].
Available: https://nanohub.org/resources/pumet.
S. Dongaonkar, S. Loser, E. J. Sheets, K.
Zaunbrecher, R. Agrawal, T. J. Marks, and M. A.
Alam, "Universal statistics of parasitic shunt
formation in solar cells, and its implications for cell to
module efficiency gap," Energy Environ. Sci., vol. 6,
[51]
no. 3, p. 782, 2013.
E. S. Mungan, Y. Wang, S. Dongaonkar, D. R. Ely, R.
E. García, and M. A. Alam, "From process to
modules: End-to-end modeling of CSS-deposited
CdTe solar cells," IEEE J. Photovoltaics, vol. 4, no. 3,
pp. 954–961, 2014.
[52] M. D. Kempe, "Modeling of rates of moisture ingress
[53]
into photovoltaic modules," Sol. Energy Mater. Sol.
Cells, vol. 90, no. 16, pp. 2720–2738, 2006.
S. Pingel, O. Frank, M. Winkler, S. Daryan, T. Geipel,
H. Hoehne, and J. Berghold, "Potential Induced
Degradation of solar cells and panels," in 2010 35th
IEEE Photovoltaic Specialists Conference, 2010, pp.
002817–002822.
In-Situ Self-Monitoring of Real-Time Photovoltaic Degradation Only Using
Maximum Power Point – the Suns-Vmp Method
Xingshu Sun,1 Raghu Vamsi Krishna Chavali,1 and Muhammad Ashraful Alam1
1Purdue University School of Electrical and Computer Engineering, West Lafayette, IN, 47907, USA.
Supplementary Information
1. Preprocess Environmental Data
The Suns-Vmp method relies on environment data, i.e., cell temperature and irradiance. The
weather information is used as inputs to the equivalent circuit to fit the reconstructed MPP IV. The
raw data can contain seasonal irradiance variation and temperature correction. Hence, it is
important to preprocess the raw data so that the parameters extracted are accurate and robust.
Below, we discuss this issue of data preprocessing in detail.
Cell Temperature. Module temperature is typically measured by attaching thermal sensors to the
back side of solar modules. The actual cell temperature can be higher than the measured back-side
module temperature regardless of convective and radiative heat transfer at the module surfaces.
Ref. [1] has developed an empirical equation to predict cell temperature (𝑇𝐶) based on illumination
intensity (GPOA) and module temperature (𝑇𝑀), which is used in this paper.
Irradiance Data. In addition to thermal information, we also need the illumination data to perform
the Suns-Vmp method. The on-site illumination data is typically measured by pyranometers
orientated as same as solar modules to collect the plane-of-array irradiance GPOA. However,
directly applying the raw GPOA data to the Suns-Vmp method can cause inaccuracy in extracting
short-circuit current because of 1) air mass dependent spectral mismatch between field and
standard test condition (STC) and 2) reflection loss of flat-plate solar modules. Thus, one must
preprocess GPOA data to eliminate the above non-idealities, as discussed below.
Spectral Mismatch. The spectral profile of GPOA under which MPP data is generated can differ
from the AM1.5G spectrum used in the STC for initial rating. Because the extracted circuits from
the Suns-Vmp method are eventually corrected to their STC values, the spectral mismatch between
real-time field irradiance and STC can contaminate the fitting results primarily for the short-circuit
current. Fortunately, the Sandia PV Array Performance Model (SAPM) has developed a
polynomial equation to empirically describe the spectral content of solar irradiance as a function
of air mass (AM) [1]. In this paper, we use the SAPM to correct the real-time GPOA to its STC
values, where AM is calculated by the Sandia PV modeling library [2] and the Direct Normal
Incidence (DNI) is retrieved from [3] at the installation location.
Reflection Loss. Pyranometers can accept irradiance coming from a highly oblique angle of
incidence (AOI) thanks to the doom-shaped glass cover, while flat-plane solar modules are
susceptible to reflection loss at high AOI. Consequently, one must also adjust GPOA measured by
pyranometers to account for reflection loss. In this paper, we also utilize the SAPM module [1] to
correct for reflection loss of the direct normal incidence, given the tilt and azimuth angles of the
analyzed solar modules.
1
Although the metrological information is often available from the on-site weather station, this may
not be the always the case. In this case, meteorological databases, such as Ref. [3] can be
alternative sources for reproducing illumination [2] and temperature information [4].)
2. Physics-Based Filtering Algorithm
Outlier data points due to instrumentation error, inverter clipping, weather condition, etc., can exist
in the field data [5]. For example, the Imp data point at around 9 am in Fig. S1 shows substantial
inconsistency with GPOA. The inclusion of these outliers in the Suns-Vmp method can induce
significant uncertainties in extracting circuit parameters. Therefore, it is necessary to develop a
self-consistent scheme to detect and then remove these outliers. Toward this goal, we have created
a continuous self-filtering algorithm to eliminate outlier data points, see Fig. S2. The steps are as
follows:
1) Fit the MPP data with non-zero POA irradiance using the equivalent circuit (MPP data
with zero irradiance always yields zero current and voltage, thereby irrelevant). Note that this
fitting step is confined to the MPP data only within the measurement window at a single time
step.
2) Calculate the relative error of fitting each MPP data point. If the error is greater than 50%,
the corresponding data point is treated as an outlier and discarded.
3) Examine the number of the remaining data points after step 2. If the remaining still consists
of more than 80% of the raw data points, proceed to step 4. Otherwise, the corresponding time
step is considered as an outlier as a whole (i.e., remove all the data points at this time step),
and will not be analyzed further. Rather, the Suns-Vmp method will directly move to the next
time-window. The entire measurement window may consist of corrupted data if temporary
instrumentations malfunctions for more than a few days.
4) Fit the filtered MPP data by the equivalent circuit and extract the circuit parameters.
5) Move to next time step.
Note that our continuous self-filtering algorithm in this paper has comprehensively accounted for
outliers caused by various non-idealities (e.g., cloud brightening,
inverter clipping,
temperature/illumination stability, pyranometer error); thus, there is no need to create individual
data filters as in [5] for the Suns-Vmp method. Moreover, the percentage thresholds in steps 2 and
3 (i.e., 50% for relative error and 80% for the number of remaining data points) is found to work
well for analyzing field data, and we do not expect a moderate adjustment of the percentage
thresholds will significantly impact the outcome. Enabled by our filtering algorithm, excellent
error control has been achieved, i.e., the relative error is less than 5% for both Vmp and Imp
through the entire 20-year analysis.
2
Fig. S1 Raw MPP data with outliers, filtered MPP data, and the environmental data on 05/16/1994 of the
NREL test facility.
Fig. S2 Flowchart of our self-filtering algorithm to identify and eliminate outlier data points.
3. Variability Test of the Suns-Vmp Method
3
06121824ModelReal-time dataOutlierEnvironmental DataMPP DataGPOA(W/m2)Fit MPP Data with GPOA> 0 W/m2within the measurement windowDiscard data points with relative fitting error > 50% Fit the filteredMPP DataMove to the next time stepYesNo* and are the numbers of remaining and total data points, respectively.
We have tested the Suns-Vmp method under various scenarios of variability using synthetic
weather data in Fig. S3. Non-uniform degradation of solar cells in the field can occur due to
different degradation pathways and have different levels of non-uniformity. Hence, we have
emulated four cases of performance variability: 1) 6 out of 36 cells degrades due to contact
corrosion (RS increases tenfold); 2) 6 out of 36 cells have encapsulant delamination (only retain
80% of initial short-circuit current); 3) 6 out of 36 cells suffer from moderate potential-induced
degradation (shunt resistance decrease by one order); 4) 6 out of 36 cells suffer from server
potential-induced degradation (shunt resistance decrease by two orders). All the tests of
performance variability are summarized in Figs. S4 to S7.
Fig. S3 Synthetic weather data containing hourly illumination and module temperature is used
to test the Suns-Vmp method.
As shown in Figs. S3 – S7, the Suns-Vmp method is still capable of diagnosing the pathology of
solar modules with non-uniform degradation. For example, the Suns-Vmp method has attributed
efficiency degradation to the increased series resistance in Fig. S4. This result, however, is not
surprising since series resistance can essentially be aggregated into one single resistance in a series-
connected circuit in Fig. S4(a). Remarkably, the Suns-Vmp method is still valid even for non-
uniform delamination- and PID-induced degradation where simple superstition of either short-
circuit current and shunt resistance of "good" and degraded cells does not hold, see Figs. S5 and
S6. The Suns-Vmp, however, cannot correctly extract the degraded circuit parameter by only one
single equivalent circuit under severe performance variability, see Fig. S7. Hence, it is
recommended to utilize multiple equivalent circuits in the Suns-Vmp method to analyze solar
modules with substantial performance variability.
4
Time (h)
(d) Extracted parameter by the Suns-Vmp method
(d) Extracted parameter by the Suns-Vmp method
Default (30
Degraded (6
Extraction
Default (30
Degraded (6
Extraction
cells)
𝐽𝑃𝐻,𝑆𝑇𝐶
282 A/m2
cells)
282 A/m2
282 A/m2
𝐽𝑃𝐻,𝑆𝑇𝐶
282 A/m2
cells)
cells)
225 A/m2
244 A/m2
𝐽01,𝑆𝑇𝐶 1.3 x 10-8 A/m2
𝐽02,𝑆𝑇𝐶 4.6 x 10-4 A/m2
𝑅𝑆ℎ,𝑆𝑇𝐶
0.12 Ω.m2
1.3 x 10-8 A/m2
4.6 x 10-4 A/m2
1.3 x 10-8 A/m2
4.6 x 10-4 A/m2
0.12 Ω.m2
0.12 Ω.m2
𝐽01,𝑆𝑇𝐶 1.3 x 10-8 A/m2
𝐽02,𝑆𝑇𝐶 4.6 x 10-4 A/m2
𝑅𝑆ℎ,𝑆𝑇𝐶
0.12 Ω.m2
1.3 x 10-8 A/m2
4.6 x 10-4 A/m2
1.3 x 10-8 A/m2
4.6 x 10-4 A/m2
0.12 Ω.m2
0.12 Ω.m2
𝑅𝑆
1.7 x 10-4 Ω.m2
1.7 x 10-3 Ω.m2
4.2 x 10-4 Ω.m2
𝑅𝑆
1.7 x 10-4 Ω.m2
1.7 x 10-4 Ω.m2
1.7 x 10-4 Ω.m2
Fig. S4
(a) A schematic of the simulated 36-cell solar module
including 6 cells degraded due to contact corrosion. The degraded
circuit elements are also highlighted. (b,c) Vmp and Imp of the solar
panel using the synthetic weather data in Fig. A1. Circles are simulated
data and solid lines are fitting data using the Suns-Vmp method. (d)
Table summarizes input parameters (both default and degraded) and
extracted parameter set using the Suns-Vmp method (affected
parameters are in bold).
Fig. S5
(a) A schematic of the simulated 36-cell solar module
including 6 cells degraded due to delamination. The degraded circuit
elements are also highlighted. (b,c) Vmp and Imp of the solar panel
using the synthetic weather data in Fig. A1. Circles are simulated data
and solid lines are fitting data using the Suns-Vmp method. (d) Table
summarizes input parameters (both default and degraded) and
extracted parameter set using the Suns-Vmp method (affected
parameters are in bold).
5
-….+36 CellsCorrosion-Degraded(6 cells)(a)FittedData(b)(c)….+36 CellsDelamination-Degraded (6 cells)(a)FittedData(b)(c)
(d) Extracted parameter by the Suns-Vmp method
(d) Extracted parameter by the Suns-Vmp method
Default (30
PID- Degraded
Extraction
Default (30
PID- Degraded
Extraction
cells)
𝐽𝑃𝐻,𝑆𝑇𝐶
282 A/m2
(6 cells)
282 A/m2
282 A/m2
𝐽𝑃𝐻,𝑆𝑇𝐶
282 A/m2
cells)
(6 cells)
282 A/m2
282 A/m2
𝐽01,𝑆𝑇𝐶 1.3 x 10-8 A/m2
𝐽02,𝑆𝑇𝐶 4.6 x 10-4 A/m2
𝑅𝑆ℎ,𝑆𝑇𝐶
0.12 Ω.m2
1.3 x 10-8 A/m2
4.6 x 10-4 A/m2
1.3 x 10-8 A/m2
4.6 x 10-4 A/m2
0.012 Ω.m2
0.026 Ω.m2
𝑅𝑆
1.7 x 10-4 Ω.m2
1.7 x 10-4 Ω.m2
1.7 x 10-4 Ω.m2
𝐽01,𝑆𝑇𝐶 1.3 x 10-8 A/m2
𝐽02,𝑆𝑇𝐶 4.6 x 10-4 A/m2
𝑅𝑆ℎ,𝑆𝑇𝐶
0.12 Ω.m2
𝑅𝑆
1.7 x 10-4 Ω.m2
1.3 x 10-8 A/m2
4.6 x 10-4 A/m2
0.0012 Ω.m2
1.7 x 10-4 Ω.m2
1.5 x 10-7 A/m2
2.1 x 10-2 A/m2
0.12 Ω.m2
1.8 x 10-4 Ω.m2
Fig. S6
(a) A schematic of the simulated 36-cell solar module
including 6 cells degraded due to moderate potential induced
degradation. The degraded circuit elements are also highlighted. (b,c)
Vmp and Imp of the solar panel using the synthetic weather data in
Fig. A1. Circles are simulated data and solid lines are fitting data using
the Suns-Vmp method. (d) Table summarizes input parameters (both
default and degraded) and extracted parameter set using the Suns-Vmp
method (affected parameters are in bold).
Fig. S7
(a) A schematic of the simulated 36-cell solar module
including 6 cells degraded due to severe potential induced degradation.
The degraded circuit elements are also highlighted. (b,c) Vmp and Imp
of the solar panel using the synthetic weather data in Fig. A1. Circles
are simulated data and solid lines are fitting data using the Suns-Vmp
method. (d) Table summarizes input parameters (both default and
degraded) and extracted parameter set using the Suns-Vmp method
(affected parameters are in bold).
4. Equations of the Five Parameter Model for Si Solar Modules
Here, we will present the analytical formulation of the five-parameter model [28] used in this paper
(see Fig. 2) and the temperature- and illumination- dependency of each parameter in Table S1.
Also, detailed description and initial STC value for Siemens M55 [25] of each parameter is listed
in Table A2. Note that GSTC = 1000 W/m2 and TSTC = 25 oC for standard test condition in for
standard test condition in Table S2.
6
-….+36 CellsPID-Degraded(6 cells)(a)FittedData(b)(c)-….+36 CellsPID-Affected(6 cells)(a)FittedData(b)(c)
TABLE S1. The equations of the five-parameter model
Analytical equations for I-V characteristics
𝐽𝐷1 = 𝐽01 (𝑒
𝑞(𝑉−𝐽𝑅𝑆)
𝑘𝑇 − 1)
𝐽𝐷2 = 𝐽01(𝑒
𝑞(𝑉−𝐽𝑅𝑆)
2𝑘𝑇 − 1)
𝐽𝑆ℎ𝑢𝑛𝑡 =
(𝑉 − 𝐽𝑅𝑆)
𝑅𝑆ℎ𝑢𝑛𝑡
𝐽 = 𝐽𝑃𝐻 + 𝐽𝐷1 + 𝐽𝐷2 + 𝐽𝑆ℎ𝑢𝑛𝑡
Illumination and temperature dependencies of the parameters
𝐽𝑃𝐻 =
𝐺
𝐺𝑆𝑇𝐶
× 𝐽𝑃𝐻,𝑆𝑇𝐶 × (1 + 𝛽 × (𝑇 − 𝑇𝑆𝑇𝐶))
𝐽01 = 𝐽01,𝑆𝑇𝐶 × (
𝐽02 = 𝐽02,𝑆𝑇𝐶 × (
𝑇
𝑇𝑆𝑇𝐶
𝑇
𝑇𝑆𝑇𝐶
1
)3 × exp(
𝑘
2
)2.5 × exp(
𝑘
𝐸𝐺,𝑆𝑇𝐶
(
𝑇𝑆𝑇𝐶
𝐸𝐺,𝑆𝑇𝐶
(
𝑇𝑆𝑇𝐶
𝐺
−
))
−
))
𝐸𝐺
𝑇
𝐸𝐺
𝑇
𝑅𝑆ℎ𝑢𝑛𝑡 = 𝑅𝑆ℎ𝑢𝑛𝑡,𝑆𝑇𝐶 ×
𝐺𝑆𝑇𝐶
𝐸𝐺 = 𝐸𝐺,𝑆𝑇𝐶 + 𝛼(𝑇 − 𝑇𝑆𝑇𝐶)
(A.1)
(A.2)
(A.3)
(A.4)
(A.5)
(A.6)
(A.7)
(A.8)
(A.9)
𝐽𝑃𝐻
𝐽01
𝐽02
𝑅𝑆ℎ𝑢𝑛𝑡
𝐸𝐺
Short-circuit current
TABLE S2. Parameter description and their initial STC values for Siemens M55 [6]
𝐽𝑆𝐶,𝑆𝑇𝐶
𝐽01,𝑆𝑇𝐶
𝐽02,𝑆𝑇𝐶
𝑅𝑆ℎ,𝑆𝑇𝐶
Diode recombination current with ideality factor of 1
Diode recombination current with ideality factor of 2
Shunt resistance
282 A/m2
1.3 x 10-8 A/m2
4.6 x 10-4 A/m2
0.12 Ω.m2
1.7 x 10-4 Ω.m2
0.49 %/K
Series resistance
temperature coefficient of short-circuit current
Bandgap of Si absorber
temperature coefficient of Si bandgap
1.12 eV
-6 x 10-4 eV/K
𝑅𝑆
𝛽
𝐸𝐺
𝛼
References:
[1] Kratochvil JA, Boyson WE, King DL. Photovoltaic array performance model,
Albuquerque, NM, and Livermore, CA, 2004.
PV_LIB Toolbox. [Online]. Available: https://pvpmc.sandia.gov/applications/pv_lib-
toolbox/.
[2]
[3] NREL. National Solar Radiation Data Base, 2010. [Online]. Available:
http://rredc.nrel.gov/solar/old_data/nsrdb/.
Faiman D. Assessing the outdoor operating temperature of photovoltaic modules, Prog.
7
[4]
Photovoltaics Res. Appl., 2008, 16, 307–315.
Jordan DC, Kurtz SR. The dark horse of evaluating long-term field performance-Data
filtering, IEEE J. Photovoltaics, 2014, 4, 317–323.
Siemans Solar Panels Installation Guide. [Online]. Available:
http://iodlabs.ucsd.edu/dja/codered/engineering/procedures/solarPower/siemens solar
panels.pdf.
[5]
[6]
8
|
1708.05000 | 1 | 1708 | 2017-07-28T19:46:28 | Focus-Induced Photoresponse: a novel optoelectronic distance measurement technique | [
"physics.app-ph",
"physics.ins-det",
"physics.optics"
] | We present the Focus-Induced Photoresponse (FIP) technique, a novel approach to optical distance measurement. It takes advantage of a widely-observed phenomenon in photodetector devices: a nonlinear, irradiance-dependent photoresponse. This means that the output from a sensor is dependent on the total number of photons incident and the size of the area in which they fall. With a certain arrangement of sensor and lens, this phenomenon will cause the output of the sensor to change based on how far in or out of focus an object is. We call this the FIP effect. Here we demonstrate how to use the FIP effect for distance measurements. We show that this technique works with different sensor materials, device types, as well as visible and near infrared light. In principle, any sensor exhibiting a photoresponse that depends nonlinearly on irradiance could be used with the FIP technique. It is our belief that the FIP technique can become an important method for measuring distance. | physics.app-ph | physics | Focus-Induced Photoresponse: a novel optoelectronic distance measurement
technique
Authors: Oili Pekkola1, Christoph Lungenschmied1,*, Peter Fejes1, Anke Handreck1, Wilfried
Hermes1, Stephan Irle2, Christian Lennartz1, Christian Schildknecht1, Peter Schillen1, Patrick
Schindler1, Robert Send1, Sebastian Valouch1, Erwin Thiel2, Ingmar Bruder1
Affiliations:
1trinamiX GmbH – a subsidiary of BASF SE, Industriestr. 35, 67063 Ludwigshafen, Germany.
2ERT Optik Dr. Thiel GmbH, Donnersbergweg 1, 67059 Ludwigshafen, Germany.
*Correspondence to: [email protected]
Abstract: We present the Focus-Induced Photoresponse (FIP) technique, a novel approach to
optical distance measurement. It takes advantage of a widely-observed phenomenon in
photodetector devices: a nonlinear, irradiance-dependent photoresponse. This means that the
output from a sensor is dependent on the total number of photons incident and the size of the area
in which they fall. With a certain arrangement of sensor and lens, this phenomenon will cause the
output of the sensor to change based on how far in or out of focus an object is. We call this the
FIP effect. Here we demonstrate how to use the FIP effect for distance measurements. We show
that this technique works with different sensor materials, device types, as well as visible and near
infrared light. In principle, any sensor exhibiting a photoresponse that depends nonlinearly on
irradiance could be used with the FIP technique. It is our belief that the FIP technique can become
an important method for measuring distance.
One Sentence Summary: We introduce a novel distance measurement technique, which utilizes
detectors with an irradiance-dependent photoresponse, demonstrate how to apply it to
measurement challenges using two types of photodetectors in the visible and the IR regime, and
model the irradiance dependence of these detectors.
INTRODUCTION
Optical distance measurement is already key to diverse applications throughout a wide range of
industries. In the coming years, it is expected to gain even more importance due to the emergence
of disruptive technologies such as machine vision and autonomous driving. These technologies
have the power to revolutionize the world around us. However, in order to do so, further advances
in optical depth sensing are required (1, 2).
Technologies like time-of-flight (ToF) and image-based depth sensing rely on advanced
manufacturing techniques such as highly developed lithography processes and CMOS
technologies (3). Since much of this development has been focused on silicon, it has become the
de facto standard in optical sensors. However, the optical characteristics of silicon are somewhat
limited. It is only sensitive to the visible and near-infrared (IR) region of light ranging from 350
– 1,100 nm (4). The optimal performance is limited to the visible range, whereas NIR absorption
is rather weak. Improved sensitivity in the IR regime would offer significant advantages in terms
of eye safety (5), night vision, and visibility in foggy conditions or through smoke (6). Materials
that feature narrower absorption bands could make measurements more robust against stray light
and background illumination. In many cases, even if adapting other materials than silicon to the
manufacturing processes is technically possible, it may not be economically viable. If these
processes were not required, materials and device types could be adopted to suit the sensing
application, rather than the other way around.
Here we introduce Focus-Induced Photoresponse (FIP), a novel method to measure distances. In a
FIP-based system, distance is determined by using the analog photoresponse of a single pixel
sensor. This means that the advantages of high-density pixelation and high-speed response are not
necessary or even relevant for the FIP technique. High resolution can be achieved without the
limitations of pixel size, and detectors selected for a FIP system can be orders of magnitude slower
than those required by ToF based ones. A system based on FIP does not require advanced sensor
manufacturing processes to function, making adoption of unusual sensors more economically
feasible.
Irradiance-dependent photoresponse of photodetectors
In the FIP technique, a light source is imaged onto the photodetector by a lens. The size of its
image depends on the position of the detector with respect to the focused image plane. FIP exploits
the nonlinearly irradiance-dependent photoresponse of semiconductor devices. This means that the
signal of a photodetector not only depends on the incident radiant power, but also on its density
on the sensor area, the irradiance. This phenomenon will cause the output of the detector to change
when the same amount of light is focused or defocused on it. This is what we call the FIP effect.
In most reports on nonlinear photoresponse, a change in irradiance has been achieved by varying
the incident radiant power. However, an irradiance variation over several orders of magnitude as
well as irradiances of over 1 sun can also be realized by focusing the light in a smaller spot.
Irradiance-dependence is a commonly-observed phenomenon among various photodetector
technologies operating from the UV to the IR regime, making them suitable for the FIP technique.
For many thin-film solar cell technologies, a nonlinear photoresponse at low irradiance levels has
been reported. Trapping of charge carriers as well as photoconductivity have been identified to
decrease the responsivity at low light levels (7, 8). On the other hand, high light intensities
exceeding 1 sun can reduce the current collection in solar cells, decreasing the photoresponse of
these devices at high irradiances (9, 10). The reduced photovoltaic performance has been attributed
to a change in recombination mechanisms and a change in series resistance with irradiance (9). In
organic solar cells, recombination has been found to depend on the charge carrier concentration
(11-13). In an experimental setup similar to that of the FIP technique, a reduced photoresponse in
thin-film solar cells as well as PbS and HgCdTe photoconductors has been reported when only a
small area of the device receives high-intensity illumination (10, 14-16).
RESULTS
The FIP effect in dye-sensitized solar cells
In dye-sensitized solar cells (DSSC), the dependence of the photovoltaic performance on light
intensity is published in detail (17-19). A nonlinear photoresponse to modulated light is reported
for low light intensities. DSSC contain a mesoporous (mp) TiO2 layer sensitized with dye
molecules. mp-TiO2 acts as the electron transporting material, and its pores are filled with a hole
transporter. In the experiments presented below, the hole transporting layer is made of a solid film
of the organic material spiro-MeOTAD (20); the device is hence referred to as a solid-state DSSC
(sDSSC). Charge transport in the mp-TiO2 structure is strongly impeded by localized states in the
band gap. These traps are occupied by photogenerated electrons relaxing into these states. The
electron diffusion coefficient in the mp-TiO2 is found to increase with the electron concentration
(17, 18, 21, 22). A higher density of absorbed photons will yield more and therefore faster electrons
per unit area.
The effect can be seen in the photocurrent transients shown in Fig. 1a. The sDSSC sample was
illuminated through a lens by a square wave modulated 530 nm LED. While the distance between
the LED and the lens was kept constant, the spot diameter was varied between 0.1 mm² and 19
mm2 by moving the cell along the optical axis around the focused image plane. Assuming a
uniform light distribution, this corresponds to irradiances between 25 – 6,000 W/m² (for details
about the spot size and irradiance calculations, see Supplementary Information S1 and S2). We
observe that increasing the irradiance strongly shortens the rise time of the photocurrent. At the
two highest irradiance levels shown in Fig. 1a, the identical equilibrium photocurrent is reached
within the pulse period and both currents reduce to zero within the dark period. All measured
transients at lower irradiance levels are already too slow to reach equilibrium or decay to zero
within the pulse duration. We interpret this behavior as a direct consequence of the reported
increase of the diffusion coefficient of the electrons in TiO2 with irradiance.
Figure 1b shows the alternating photocurrent density of the sDSSC as a function of the irradiance
at modulation frequencies between 75 Hz and 975 Hz. Irradiances between 10 and 10,000 W/m²
are covered by moving the sDSSC behind the lens in 0.1 mm steps and thereby changing the size
of the image on the sensor. The scatter points represent the photocurrent densities measured at
each position. The areas of the light spots were calculated with paraxial optics and assumed to be
uniformly illuminated (see Supplementary Information S1). To test the validity of this assumption,
we also obtained beam profiles by raytracing and used a segmentation approach to parametrize the
photocurrent densities as a function of the irradiance. The model is described in detail in
Supplementary Information S3. The resulting photocurrent densities are plotted as lines in Fig. 1b.
We find good agreement with the data obtained by assuming uniformly illuminated light spots.
The dashed line with a slope of unity represents the linear detector response at high intensities and
is added as a guide for the eye. At low modulation frequencies, the photocurrents are close to the
steady-state level. With increasing frequency, the deviation from linearity at low irradiance levels
increases. The alternating photocurrent induced by a modulated light source can thus be larger at
higher photon densities even though the total amount of light on the sample stays constant: the
solar cell works more efficiently at high irradiance. This is the signature of the FIP effect in sDSSC.
Figure 1. Nonlinear photoresponse of sDSSC. a, Transient photocurrent response of a sDSSC to
pulsed illumination through a lens. The device is placed at various positions on the optical axis,
hence the same incident radiant power is distributed over different surface areas. The relative
variation of the light spot sizes is illustrated by the circles. b, Alternating photocurrent of a sDSSC
as a function of the irradiance for different modulation frequencies. The scatter points assume
uniform illumination of the light spot. The solid lines are based on a raytracing model and the
segmentation of the light spot. The dashed line with a slope of 1 acts as guide to the eye.
Measuring distances with the FIP technique
We demonstrate how the FIP effect can be utilized for distance measurements with a setup
schematically depicted in Fig. 2. Its components are a modulated LED1, a consumer grade camera
lens, two photodetectors, and a signal processing unit. The lens collects and focuses the light of
the LED. The semitransparent sensors are placed behind the lens near its focal plane. The
modulated light of the LED generates an alternating photoresponse, which is then amplified and
recorded using lock-in or Fourier transform techniques. The spot size on the sensor changes with
the distance between the light source and the lens as the position of the focused image plane shifts.
The active area must therefore be large enough to accommodate the maximum spot size within the
desired measurement range.
Figure 2. A typical setup for measuring distances with the FIP technique. The size of the light spot
and therefore the irradiance depends on the sensor position behind the lens. The graph shows the
illuminated area around the position of the focused image plane for the experiments presented in
Fig. 1.
There are two main contributions to the sensor output. The first is due to the FIP effect: when the
modulated LED is positioned at a certain distance from the lens, the measured photocurrent
depends of the irradiance on the sensor, i.e. how well the light is focused. Secondly, the
photocurrent is impacted by the total amount of light collected by the lens. When the LED is moved
1 Instead of LEDS any light source, either actively emitting or reflecting light may be used. Alternatively to the
camera lens, any optical element that captures the light and focuses it, such as lenses or mirrors, may be used.
away from the lens, this contribution decreases. If the radiant power of the light source is unknown,
the photoresponse of a single detector at any given position behind the lens does not allow for an
unambiguous distance determination. Whether the LED is distant and bright or close and dim
cannot be distinguished. To solve this problem, we use two detectors in the beam path. By
calculating the ratio of the two photoresponses, the distance dependence as well as fluctuations in
the output of the light source cancel out. Due to the FIP effect, the quotient changes with the
distance, yielding a unique signature for each LED position.
Demonstrating the FIP technique with sDSSC
We have realized such a setup by using two semitransparent sDSSC as sensors and a modulated
green LED. The results are summarized in Fig. 3. Figure 3a and 3b show the responsivity of the
sDSSC as a function of their position behind the lens. The responsivity was calculated by
normalizing the short circuit current to the total radiant power incident on the device. The sensor
closer to the lens is referred to as the first sensor, and the one further from the lens as the second
sensor. Due to the FIP effect, the maximum photocurrent is recorded when the sensor is positioned
in the focused image plane. In this case, the LED light is focused on the sensor. A reduction in the
current is observed as the light spot widens symmetrically. With increasing LED distance, the
focused image plane moves towards the focal plane of the lens, leading to a shift of the
photocurrent peaks. As shown in Fig. 3, we observe that the responsivity curves intersect at the
focal plane of the lens. In this point, the irradiance in the light spot is independent of the LED
distance, and only the diameter of the light spot changes (for the derivation, see Supplementary
Information Section S4). This observation is consistent with our assumption that the FIP effect is
induced by the irradiance-dependent sensor response.
Figure 3. distance measurement with sDSSC. a,b, Responsivity of the first (a) and second (b)
sDSSC sensor for a range of positions behind the lens at various LED distances. The second sensor
is illuminated through the first. c, The absolute photocurrent of both sensors and the radiant power
on the first sensor as a function of LED distance. d, The quotient of the photocurrents as a function
of LED distance.
The positions of the sensors in the beam path may be adjusted to the specific measurement
problem. In the presented case, we measure distances in the range between 20 cm and 1.8 m (Fig.
3c and 3d). The first sensor is positioned 33.3 mm behind the lens, between the lens and its focal
plane. The second sensor is placed 37.6 mm behind the lens. With increasing LED distance, the
responsivity of the first sensor decreases, while it increases for the second sensor. Therefore, a
large change in the ratio of their photocurrents is achieved over the measurement range. The
absolute photocurrents of both sensors as well as the incident radiant power are plotted in Fig. 3c.
The photocurrents are dominated by the inverse dependence of the radiant power on LED distance.
The deviation from linearity with irradiance is visible as a difference in slopes. The resulting
quotient increases monotonically over the entire measurement range (Fig. 3d), assigning a single
value to each object distance. With this calibration curve, the LED distance can be directly
determined by simply measuring the individual photocurrents of both cells and calculating their
ratio.
The FIP effect in PbS photoconductors
The low-irradiance nonlinearity observed in DSSC at sufficiently large frequencies causes a
maximum photoresponse when the device is placed in focus. In this situation, the irradiance and
consequently the efficiency of the device are maximized. For other thin-film solar cell and
photoconductor technologies, a reduced photoresponse has been reported when only a small area
of the device receives high intensity illumination (10, 14, 15). We therefore expect the
photoresponse of these devices to reach a minimum in focus, as the efficiency decreases with
irradiance. PbS photoconductors used as IR sensors are an example of such behavior. In contrast
to photovoltaic detectors, photoconductors do not generate a photocurrent. Instead, their resistance
changes upon illumination. Even though the functionality differs fundamentally from DSSC, these
devices can be used for distance measurements with the FIP technique. PbS detectors are opaque
and therefore cannot be used in sensor stacks like the sDSSC presented above. Instead, a beam
splitter may be used to position both sensors at appropriate distances from the lens.
Figure 4a shows the responsivity of a PbS detector to modulated 1,550 nm illumination as a
function of its position behind the lens for various LED distances. The photoresponse is recorded
as a voltage using an amplifier circuit. We determine the responsivity by normalizing the
photoresponse to the radiant power reaching the detector. A minimum in the photoresponse is
observed when the light spot is focused on the detector.
Figure 4. Distance measurement with PbS. a, responsivity of a PbS photoconductor for a range
of positions behind the lens at various LED distances. The device is modeled as a network of
infinitesimally small photoconductors with a linear response to irradiance. The lines represent
best fits to the experimental data. b, the quotient of the photoresponse of the PbS device at 40.4
mm and at 45 mm as a function of LED distance. c, detail of the quotient curve with an indication
of a corridor of ± 0.1 % of the distance to illustrate the obtained resolution.
We have modeled the responsivity of the tested PbS photoconductor to light of various spot sizes
and irradiance levels. The device is simulated as a two-dimensional grid of connected
infinitesimally small photoconductor elements. By performing a limiting process, we can model
the grid with an elliptic partial differential equation, i.e. ∇ ⋅ (
∇𝜑
𝑅(𝑥,𝑦)
) = ∇ ⋅ (σ(x, y)∇𝜑) = 0, where
φ(x,y) is the electric potential, R(x,y) the local resistance and σ(x,y) the conductance (more details
in Supplementary Information S5). This model can be solved using the Finite Elements Method.
The dark conductance 𝜎𝑑 of every point of the PbS photoconductor is modified by illumination.
The conductance of an illuminated point is described as 𝜎 = 𝜎𝑑 + Δ𝜎. Δσ is assumed to be directly
proportional to the irradiance (𝐸~
𝑚2), i.e. Δ𝜎 = 𝑝 ⋅ 𝐸. The irradiance is calculated using paraxial
optics and assumes uniformly illuminated light spots (see S1). We have fitted the measured data
𝑊
plotted in Fig. 6a using Δ𝜎 = 0.298
𝑚𝑚
𝑉2 ⋅ 𝐸.
The resulting curves are shown as lines in Fig. 4a. We interpret the good agreement between the
measured data and our model as a confirmation of the assumed linear dependence of Δσ on
irradiance for the tested PbS photoconductor within the studied irradiance regime from 0.3 W/m²
to 5,000 W/m². Even if a photoconductor reacts perfectly linearly to the irradiance, a FIP effect is
observed when the active area is only partially illuminated. This behavior is consistent with
experimental data (15) and shown formally in Supplementary Information S5.
The FIP effect in PbS photoconductors is used for distance measurements by assigning a
photoresponse quotient to measured distances as shown in Fig. 4b. The ratio of the photoresponses
was determined for sensor positions at 45 mm and 40.4 mm behind the lens. The resulting quotient
increases over the studied measurement range, enabling accurate distance measurements between
45 – 165 cm. Figure 4c depicts a small range to illustrate the resolution of distance measurements
with the FIP technique. LED positions as close together as 500 µm can be distinguished by the
photoresponse quotient at a distance of 52 cm, corresponding to a depth resolution of better than
0.1 %.
CONCLUSIONS AND OUTLOOK
We have demonstrated that the FIP technique is a new and versatile method for measuring distance.
The differences between it and traditional methods like ToF and triangulation open the door for
different types of measurements. FIP sensors can be extremely simple. They do not need to be
arrayed or operate at high speed. Many applications could benefit from using wavelengths outside
the bounds of traditional distance measurement sensors; with FIP this is possible. The only
requirement is that the sensor displays the FIP effect, which many materials do. We have observed
the FIP effect in various thin-film photovoltaic device technologies such as DSSC, amorphous
silicon, CdTe, CIGS, CIS, CZTS, as well as in organic solar cells (23) and PbS photoconductors.
The samples we tested were either purchased or produced using standard techniques. In this
article, we have shown resolution of below 500 µm at a distance of 50 cm. In our supplementary
information, distance measurements up to 70 m can be found (Section S6). We believe that
research and device optimization will further improve these results.
The technique of FIP can be combined with other technologies, to create systems with even more
functionality. A device sensitive to the x, y and z coordinates (24) of a light spot can be created
by using commercially available position sensitive devices (PSD) as the sensors. Simultaneous
tracking of multiple light spots is possible if they have different modulation frequencies. It is also
possible to utilize projected light spots instead of actively emitting ones. This allows lasers to be
used as the light sources instead of the LEDs we have presented here. An advantage of this method
is that the position of the laser's origin does not impact the measurement. The FIP technique only
measures the distance to the light spot. With improved understanding and further development of
the technique, FIP can become an important distance measurement technique.
References and Notes:
1
2
3
International Roadmap for Devices and Systems 2016 Edition. (IEEE 2016).
International Technology Roadmap for Semiconductors, http://www.itrs2.net
Theuwissen, A. J. P. CMOS image sensors: State-of-the-art. Solid-State Electronics 52,
1401-1406 (2008).
4
Yokogawa, S. et al. IR sensitivity enhancement of CMOS Image Sensor with diffractive
light trapping pixels. Scientific Reports 7, 3832 (2017).
5
International Standard IEC 60825-1:2014 Safety of laser products – Part I: Equipment
classification and requirements (2014).
6
Goossens, S. et al. Broadband image sensor array based on graphene-CMOS integration.
Nature Photonics 11, 366-371 (2017).
7
Hegedus, S. S. & Shafarman, W. N. Thin-film solar cells: device measurements and
analysis. Progress in Photovoltaics: Research and Applications 12, 155-176 (2004).
8
Dalal, V. L. & Rothwarf, A. Comment on ''A simple measurement of absolute solar cell
efficiency''. Journal of Applied Physics 50, 2980-2981 (1979).
9
Hohl-Ebinger, J., Siefer, G. & Warta, W. Non-linearity of solar cells in spectral response
measurements. 22nd European Photovoltaic Solar Energy Conference and Exhibition
(Milan, Italy, 2007).
10 McMahon, T. J. & von Roedern, B. Effect of Light Intensity on Current Collection in Thin-
Film Solar Cells. 26th IEEE Photovoltaic Specialists Conference (Anaheim, CA, 1997).
11
Clarke, T. M., Lungenschmied, C., Peet, J., Drolet, N. & Mozer, A. J. A Comparison of
Five Experimental Techniques to Measure Charge Carrier Lifetime in Polymer/Fullerene
Solar Cells. Advanced Energy Materials 5, 1401345 (2015).
12
Kirchartz, T. & Nelson, J. Meaning of reaction orders in polymer:fullerene solar cells.
Physical Review B 86, 165201 (2012).
13
Deledalle, F., Tuladhar, P. S., Nelson, J., Durrant, J. R. & Kirchartz, T. Understanding the
Apparent Charge Density Dependence of Mobility and Lifetime in Organic Bulk
Heterojunction Solar Cells. The Journal of Physical Chemistry C 118, 8837-8842 (2014).
14
Theocharous, E. Absolute linearity measurements on a PbS detector in the infrared. Applied
Optics 45, 2381-2386 (2006).
15
Theocharous, E., Ishii, J. & Fox, N. P. Absolute linearity measurements on HgCdTe
detectors in the infrared region. Applied Optics 43, 4182-4188 (2004).
16
Bartoli, F., Allen, R., Esterowitz, L. & Kruer, M. Auger‐limited carrier lifetimes in
HgCdTe at high excess carrier concentrations. Journal of Applied Physics 45, 2150-2154
(1974).
17
Cao, F., Oskam, G., Meyer, G. J. & Searson, P. C. Electron Transport in Porous
Nanocrystalline TiO2 Photoelectrochemical Cells. The Journal of Physical Chemistry 100,
17021-17027 (1996).
18
van der Zanden, B. & Goossens, A. The Nature of Electron Migration in Dye-Sensitized
Nanostructured TiO2. The Journal of Physical Chemistry B 104, 7171-7178 (2000).
19
O'Regan, B., Moser, J., Anderson, M. & Graetzel, M. Vectorial electron injection into
transparent semiconductor membranes and electric field effects on the dynamics of light-
induced charge separation. The Journal of Physical Chemistry 94, 8720-8726 (1990).
20
Bach, U. et al. Solid-state dye-sensitized mesoporous TiO2 solar cells with high photon-
to-electron conversion efficiencies. Nature 395, 583-585 (1998).
21
Peter, L. M. & Wijayantha, K. G. U. Intensity dependence of the electron diffusion length
in dye-sensitized nanocrystalline TiO2 photovoltaic
cells. Electrochemistry
Communications 1, 576-580 (1999).
22
Hagfeldt, A., Boschloo, G., Sun, L., Kloo, L. & Pettersson, H. Dye-Sensitized Solar Cells.
Chemical Reviews 110, 6595-6663 (2010).
23
Valouch, S. et al. Detector for an optical detection of at least one object. Patent application
number PCT/EP2016/051817.
24 Martins, R. & Fortunato, E. Thin Film Position Sensitive Detectors: From 1D to 3D
Applications. Technology and Applications of Amorphous Silicon (ed Robert A. Street)
(Springer-Verlag, 2010).
Acknowledgments: We acknowledge the help of John Dowell and Ines Kühn in improving the
overall quality of the manuscript. We wish to thank Peter Haring Bolivar from the
University of Siegen for discussions and the scientific support, as well as Peter Erk, Karl
Hahn, and Harald Lauke from BASF SE for their continued support.
Methods
Fabrication of sDSSC. The FTO substrates (Pilkington glass) were first cleaned with a glass
detergent, then rinsed with water and cleaned with acetone and isopropanol. Subsequently, the
substrates were ozone treated for 30 min (Novascan PSD Series Digital UV Ozone System). After
that, a TiO2 blocking layer was deposited on the substrates via spray pyrolysis. 9.72 g titanium
diisopropoxide bis(acetylacetonate) was dissolved in 100 ml ethanol. 25 spray cycles were
performed at 350 °C. After the pyrolysis, the samples were annealed at 350 °C for 30 min. For the
deposition of a mesoporous TiO2 layer, transparent titania paste (Dyesol, average particle size 20
nm) was mixed with ethanol in a ratio of 1:3. The solution was spin coated at 3700 RPM for 30 s,
and the films were sintered subsequently at 450 °C for 30 min. The TiO2 films were immersed in
a 5 mM dye (N-Carboxymethyl-9-(7-(bis(9,9-dimethyl-fluoren-2-yl)amino)-9,9-dihexyl-fluoren-
2-yl)perylene-3,4-dicarboximide) solution in toluene for 1 h. After that the samples were rinsed
with water and dried with nitrogen. 100 mg/ml hole conductor 2,2',7,7'-Tetrakis[N,N-di(4-
methoxyphenyl)amino]-9,9'-spirobifluorene (Spiro-MeOTAD) in chlorobenzene was mixed with
20 mM bis(trifluoromethane) sulfonamide lithium salt in cyclohexanone and 2.5 mg/ml vanadium
pentoxide, and oxidized in air for 1 h. Subsequently, vanadium pentoxide was removed by filtering
the solution through a 0.2 µm PTFE filter. The solution was spin coated at 2000 RPM for 30 s and
the samples were left to dry for 30 min. We used PEDOT:PSS (Clevios F HC Solar) as the counter
electrode. The dispersion was filtered with a 0.45 µm PTFE filter and spin coated at 2000 RPM
for 30 s. After that, the samples were dried on a hot plate at 90 °C. Finally, 200 nm Ag contacts
were evaporated on top of the PEDOT:PSS film using a custom-made Creavac thermal evaporator.
Transient photocurrent. The sDSSC was illuminated with a 530 nm LED (Thorlabs M530L3). The
LED was modulated at 375 Hz with square wave pulses and a duty cycle of 50 %. The light was
focused with an aspheric lens (Thorlabs AL2520-A) that was positioned at 35.2 mm from the LED.
The sensor was mounted on a translational stage that allowed movements along the optical axis.
The light power on the sensor was 465 µW, and the size of the image on the sensor was changed
by moving the sensor on the optical axis. The transient photocurrent was amplified with a Femto
DLPCA transimpedance amplifier (gain 104 V/A) and recorded with a National Instruments PXIe-
4492 measurement card.
Photocurrent as a function of photon density and modulation frequency. The measurement setup
was identical to that of transient photocurrent as described above. The alternating photocurrent
was amplified with a Femto DLPCA transimpedance amplifier (gain 104 V/A) and recorded with
a Behringer U-Phoria UMC202HD sound card. The LED was modulated at 75, 175, 375, 575, 775,
and 975 Hz, and the light power on the sensor was 855 µW.
FIP distance measurements with DSSC. A stack of two semitransparent sDSSC was illuminated
with a square wave pulsed LED (Thorlabs M530L3) at 530 nm. The LED was modulated with
square wave pulses at 475 Hz. The light was focused with a Nikkor 50 mm f/1.2 lens. The distance
between the sensors in the stack was 4.3 mm. The stack was mounted on a translational stage and
the LED on a rail that allowed movements along the optical axis. The radiant power on the first
sensor at LED distance of 13.5 cm was 790 µW. The photocurrents were amplified with two Femto
DLPCA-200 transimpedance amplifiers (gain 104 V/A) and recorded with two lock-ins (SR 850,
Stanford Research Systems).
FIP distance measurements with PbS and the photoconductor model. A commercial PbS
photoconductor (HertzstückTM, active area 1 cm x 1cm) was illuminated with an LED at 1,550 nm
(Thorlabs M1550L3). The LED was modulated with square wave pulses at 606 Hz and its light
was focused with a Nikkor 50 mm f/1.2 lens. The photoresponse was measured using a voltage
divider including a 2 MΩ resistor to match the dark resistance of the PbS device of similar
resistance. A voltage of 100 V was applied to the photoconductor and the 2 MΩ resistor, hence an
electric field of 50 V/cm was present across the active area of the photodetector. The photoresponse
to the modulated LED was then determined with a Behringer U-Phoria UMC202HD sound card
connected via a unity gain buffer. The experimental data are obtained using an FFT with a
bandwith of 1 Hz. This can be interpreted as a smoothing process. In order to remove numerical
perturbations, the simulated results were smoothed by using a moving average filter as well. The
PbS photoconductor was mounted on a translational stage and the LED on a rail that allowed their
movements along the optical axis. The radiant power on the sensor at an LED distance of 13.5 cm
was 35.1 µW.
Supplementary information
S1 – Modelling an optical image with paraxial optics
We use paraxial optics to estimate the size of an optical image on the FIP sensor. The optical
setup is described by the paraxial approximation, i.e. the model is based on the thin lens equation
1
𝑓
=
1
𝑧
+
1
𝑏
(𝑆1 − 1)
where 𝑓 is the focal length of the lens, 𝑧 the distance between the light source and the lens and 𝑏
the distance between the lens and the focused image of the light source. It should be noted that the
model is an approximation. The irradiance distribution within the image or the properties of the
lens are not considered.
Fig. S1.1
Optical setup
The disk-shaped light source has a radius 𝑟𝐿𝐼𝐺𝐻𝑇. The sensor is placed at a distance 𝑧𝑠 behind the
lens. The image of the light source on the sensor has a radius 𝑟𝐼𝑀𝐺. It consists of the image of the
light source without blur, 𝑟𝑂𝐵𝐽, and the circle of confusion 𝑐𝑜𝑐:
𝑟𝐼𝑀𝐺 = 𝑟𝑂𝐵𝐽 + 𝑐𝑜𝑐
(𝑆1 − 2)
The circle of confusion is determined by the intercept theorem, i.e.
Combining equations S1 – 1 and S1 – 3, the circle of confusion yields to
𝑐𝑜𝑐
𝑏 − 𝑧𝑠
=
𝑟𝐿𝐸𝑁𝑆
𝑏
.
𝑐𝑜𝑐 = 𝑟𝐿𝐸𝑁𝑆 (1 −
𝑧𝑠
𝑏
1
) = 𝑟𝐿𝐸𝑁𝑆 (1 − 𝑧𝑠 ⋅ (
𝑓
−
1
𝑧
)) = 𝑟𝐿𝐸𝑁𝑆 (1 − 𝑧𝑠 ⋅ (
𝑧 − 𝑓
𝑓𝑧
(𝑆1 − 3)
)) .
(𝑆1 − 4)
The object size on the image 𝑟𝑂𝐵𝐽 is also determined by the intercept theorem, i.e.
𝑟𝑂𝐵𝐽
𝑧𝑠
=
𝑟𝐿𝐼𝐺𝐻𝑇
𝑧
.
The radius of the optical image 𝑟𝐼𝑀𝐺 is thus
𝑟𝐼𝑀𝐺 = 𝑟𝐿𝐸𝑁𝑆 1 − 𝑧𝑠 ⋅ (
𝑧 − 𝑓
𝑓𝑧
) +
𝑟𝐿𝐼𝐺𝐻𝑇
𝑧
𝑧𝑠.
This formula can be extended easily by using the diameter 𝑑𝐿𝐸𝑁𝑆 instead of the radius:
𝑑𝐼𝑀𝐺 = 𝑑𝐿𝐸𝑁𝑆 1 − 𝑧𝑠 ⋅ (
𝑧 − 𝑓
𝑓𝑧
) +
𝑑𝐿𝐼𝐺𝐻𝑇
𝑧
𝑧𝑠.
(𝑆1 − 5)
(𝑆1 − 6)
(𝑆1 − 7)
S2 – Image sizes and photon densities in transient photocurrent measurements (Fig. 1a)
Distance of the sensor from the focused image plane: 0.2 – 5 mm, power of the light source: 465
µW
Sizes of the light spot were calculated with paraxial optics (Eq. S1-7) with following parameters:
• Distance of the light source 35.2 mm
• Diameter of the light source 2 mm
• Focal length of the lens
20 mm
• Working F#
0.979877
Distance from
Spot area (mm²)
Irradiance (W/m²)
focused image plane
(mm)
0.2
0.5
1
2
3
4
5
0.08
0.28
0.91
3.25
66.18
12.24
18.90
6087.73
1650.56
510.09
143.08
66.18
37.99
24.61
S3 – Simulation of beam profiles with ray tracing
Although paraxial optics allows for a qualitative understanding of the imaging process, a ray
tracing model accounting for actual lens systems is needed. Since the sensor signal depends on
irradiance, special care must be taken to account for the spatial distribution of photons over the
sensor area. It is not possible to assign a single irradiance value to a specific beam profile in a
FIP measurement, since each profile on the sensor consists of a characteristic distribution of
local irradiances. To deduce the specific relation between irradiance and the current density of a
certain sensor type, we have chosen the following ansatz:
We discretize the sensor area with a rectangular grid. At each sensor position, every pixel is
assumed to behave as a local sensor that is exposed to a discrete irradiance 𝐸𝑖,𝑠. The discrete
irradiance depends on the radiant power distribution Φ𝑙𝑜𝑐𝑎𝑙 at that pixel and sensor position:
𝐸𝑖,𝑠 =
Φ𝑙𝑜𝑐𝑎𝑙(𝑥𝑖, 𝑦𝑖, 𝑧𝑠)
𝐴𝑙𝑜𝑐𝑎𝑙
,
(𝑆3 − 1)
where 𝑥𝑖 and 𝑦𝑖 are the pixel coordinates, 𝑧𝑠 the sensor position with respect to the lens and
𝐴𝑙𝑜𝑐𝑎𝑙 the area of each pixel. We now define the local response function 𝑓𝑙𝑜𝑐𝑎𝑙 as
𝑓𝑙𝑜𝑐𝑎𝑙 = 𝑝1 ∙ 𝐸𝑖,𝑠 − 𝑝2 ∙ 𝐸𝑖,𝑠 ∙ exp(−𝑝3 ∙ 𝐸𝑖,𝑠
𝑝4) ,
(𝑆3 − 2)
where 𝑝1 − 𝑝4 are simulation parameters. The local response function gives the local current 𝐼𝑖,𝑠
originating from the local irradiance:
By summing over all local currents, we obtain the overall sensor response current 𝐼𝑠 for each
𝐼𝑖,𝑠 = 𝑓𝑙𝑜𝑐𝑎𝑙(𝐸𝑖,𝑠)
(𝑆3 − 3)
sensor position 𝑧𝑠:
𝐼𝑠 = ∑ 𝑓𝑙𝑜𝑐𝑎𝑙(𝐸𝑖,𝑠)
𝑖
(𝑆3 − 4)
To obtain the parameters of the local response function, we used the simulated irradiation
profiles at different sensor positions in combination with the experimentally measured
photocurrents and performed a least-squares fit to obtain the best matching local response
function.
Specifically, we used the following parameters:
• Lens: Thorlabs AL2520M-A, Mounted Asphere, Ø25.0mm, EFL = 20.0mm, NA=0.54, -A
Coating
• LED position with respect to the lens: 32.5 cm
• Wavelength: 530 nm
• Width of the discretized sensor: 7 mm
• Number of pixels per line/column: 1,001
• Number of simulation rays: 5,000,000
Parameters for different modulation frequencies:
Frequency
p1
p2
p3
p4
975 Hz
775 Hz
575 Hz
375 Hz
175 Hz
75 Hz
0.001077475772490 0.001090632227783 0.026971714492125 0.531302606764043
0.001112257999756 0.001151022565467 0.034556007165795 0.523592469700944
0.001149372503804 0.001279720508382 0.058938000077514 0.479743037095579
0.001181978599514 0.001493075487758 0.109307540253059 0.430996445237502
0.001221120774188 0.002960866589844 0.477235010400842 0.281457998864533
0.001242651802542 0.012728796621985 1.658609785709937 0.166377333205729
S4 - Modelling the irradiance of a photodetector through a lens
Assumptions:
• The light source is infinitesimally small (point light source) and emits uniformly in all
directions
• The optical setup is described by the paraxial approximation, observing the thin lens
equation S1-1
• The sensor is larger than the image of the light source
Lens
Sensor
𝑟𝐿𝐸𝑁𝑆
𝑟𝑜𝑏𝑗
𝑧
𝑧𝑠
𝑏
Figure S4.1: Optical Setup
In consideration of these assumptions, the optical image on the sensor is a circular disk. Its radius
(𝑟𝑜𝑏𝑗) is given by
𝑟𝑜𝑏𝑗 = 𝑟𝐿𝐸𝑁𝑆 (1 − 𝑧𝑠
𝑧 − 𝑓
𝑧𝑓
) ,
(𝑆4 − 1)
where 𝑟𝐿𝐸𝑁𝑆 is the radius of the lens and 𝑧𝑠 is the distance between lens and sensor.
The amount of light shining on the sensor is modeled by the radiant power 𝐿. For sufficiently
large 𝑧, it decreases with the square of the distance between light source and lens. The parameter
𝜆0 characterizes the emitted light and the transmission properties of the lens.
𝐿(𝑧) =
𝜆0
𝑧2
(𝑆4 − 2)
The irradiance 𝐸 of the sensor is given by the distribution of the radiant power on the sensor
𝐸(𝑧) = {
𝜋𝑟𝑜𝑏𝑗
0
1
2 𝐿(𝑧) ‖𝑥‖ ≤ 𝑟𝑜𝑏𝑗
‖𝑥‖ > 𝑟𝑜𝑏𝑗
(𝑆4 − 3)
The nonlinear sensor response function to the irradiance is defined by 𝐹. The overall sensor
response is the spatial integral over the irradiance, i.e.
𝐼(𝑧) = ∫ 𝐹(𝐸(𝑥))𝑑𝑥 = 𝜋𝑟𝑜𝑏𝑗
2
The normalized sensor response 𝐼𝑛𝑜𝑟𝑚 is defined by
⋅ 𝐹 (
1
2
𝜋𝑟𝑜𝑏𝑗
⋅ 𝐿(𝑧)) .
(𝑆4 − 4)
𝐼𝑛𝑜𝑟𝑚 =
𝐼(𝑧)
𝐿(𝑧)
.
(𝑆4 − 5)
Iso-FIP theorem:
Let the sensor position set to 𝑧𝑠 = 𝑓. Then the following result is valid: For any sensor response
function 𝐹 the normalized sensor response does not depend on the distance 𝑧.
Proof:
Let 𝑧𝑠 = 𝑓, the radius of the illuminated disk reduces to
𝑟𝐿𝐸𝑁𝑆 ⋅ 𝑓
𝑟𝑜𝑏𝑗 =
.
𝑧
Then the normalized sensor response 𝐼𝑛𝑜𝑟𝑚 yields to
𝐼𝑛𝑜𝑟𝑚 =
𝑓2
2
𝜋𝑟𝐿𝐸𝑁𝑆
𝑧2
𝐹 (
𝑧2
2
𝜋𝑟𝐿𝐸𝑁𝑆
𝑓2 𝐿(𝑧)) 𝐿(𝑧)−1.
Plug in the irradiance function on the sensor 𝐿:
𝐼𝑛𝑜𝑟𝑚 =
𝑓2
2
𝜋𝑟𝐿𝐸𝑁𝑆
𝑧2
𝐹 (
𝑧2
2
𝜋𝑟𝐿𝐸𝑁𝑆
𝜆0
𝑧2)
𝑧2
𝜆0
=
𝑓2
2
𝜋𝑟𝐿𝐸𝑁𝑆
𝜆0
𝑓2
𝐹 (
𝜆0
2
𝜋𝑟𝐿𝐸𝑁𝑆
𝑓2) .
(𝑆4 − 6)
(𝑆4 − 7)
(𝑆4 − 8)
For the normalized sensor response, 𝑧 cancels out, hence the response does not depend on the
distance 𝑧.
Explanation:
When the distance between light source and sensor (𝑧) increases, the amount of light impinging
on the sensor decreases. At the same time, the size of the optical image decreases. Both trends
contribute to the amount of light per unit area, the irradiance of the sensor.
Assuming the characteristics of a point source and the validity of the thin lens approximation, we
show that if the sensor is positioned in the focal plane of the lens (at 𝑧𝑠 = 𝑓), the area of the
optical image is inversely proportional to the square of the distance between light source and
sensor. This yields that the reduction in the amount of light on the sensor cancels out the
reduction in image size. Hence the irradiance of the image remains constant when 𝑧 is varied.
We assume that the quantum efficiency, as expressed by the sensor response function 𝐹, depends
on the irradiance. Under the above assumptions, the irradiance is constant over of area of the
image. Thus, the normalized sensor response of the detector positioned in the focal plane is
identical for any distance 𝑧.
Even though the conditions in the actual experiments deviate significantly from the assumptions
defined above, we find that responsivity curves at various distances for sDSSC (Fig. 3a), PbS
photoconductors (Fig. 4a) and amorphous silicon (Fig. S6.1) intersect close to the focal length of
the used lens.
S5 - Modelling the photoresponse of partially illuminated photoconductors
We model a photoconductor device as an infinitesimally fine network of light dependent
resistors. The current flow is governed by the Ohm's law, i.e. 𝑈 = 𝑅 ⋅ 𝐼.
The goal is to derive a continuous model of the network by a limit process for mesh sizes Δ𝑥
approaching zero (Fig. S5.1). The result of the limit process is an elliptic partial differential
equation that describes the current flow within a photoconductor.
Resistor
Δ𝑥
Fig. S5.1
Visualization of a limit process of electricity network. The photoconductor is
Δ𝑥
→ 0
modeled as a coarse electricity network.
The continuous model can be derived as follows:
We can interpret the solution of the resistor network as the solution of a two-dimensional finite-
volume scheme. Each arc of the network should have the same length Δ𝑥. We assume that the
values of a node correspond the value of a cell (see Fig. S5.2). For more details see: Peter
Schillen - Modelling and Control of Balance Laws with Applications to Networks – Dr. Hut -
ISBN 9783843922159 – Section 3.6. The voltage is given by the finite difference of the potential
𝜑, e.g. 𝑈𝑁 = (𝜑𝑖,𝑗+1 − 𝜑𝑖,𝑗).
𝜑𝑖,𝑗−1
𝑅𝑁
𝜑𝑖−1,𝑗
𝑅𝑊
𝑅𝐸
𝜑𝑖+1,𝑗
𝑅𝑆
𝜑𝑖,𝑗+1
Fig. S5.2
Scheme of a node in the resistor network
Kirchhoff's current laws and Ohm's law yield the following equation
0 =
𝐼𝑆−𝐼𝑁
Δ𝑥
+
𝐼𝐸−𝐼𝑊
Δ𝑥
=
(𝜑𝑖,𝑗+1−𝜑𝑖,𝑗)𝑅𝑁−(𝜑𝑖,𝑗−𝜑𝑖,𝑗−1)𝑅𝑆
Δ𝑥2𝑅𝑆𝑅𝑁
+
(𝜑𝑖,𝑗+1−𝜑𝑖,𝑗)𝑅𝑊−(𝜑𝑖,𝑗−𝜑𝑖,𝑗−1)𝑅𝐸
Δ𝑥2𝑅𝐸𝑅𝑊
(𝑆5 − 1)
For Δ𝑥 → 0 the finite differences coincide with its derivatives.
0 =
𝜕𝑥𝑥𝜑𝑅(𝑥, 𝑦) − 𝜕𝑥𝜑𝜕𝑥𝑅(𝑥, 𝑦)
𝑅(𝑥, 𝑦)2
+
𝜕𝑦𝑦𝜑𝑅(𝑥, 𝑦) − 𝜕𝑦𝜑𝜕𝑦𝑅(𝑥, 𝑦)
𝑅(𝑥, 𝑦)2
(𝑆5 − 2)
It is necessary to assume that 𝑅 is weakly differentiable.
Finally, the limit yields a continuous model based on a partial differential equation
∇ ⋅ (
∇𝜑
𝑅(𝑥, 𝑦)
) = 0,
(𝑆5 − 3)
for all 𝑥 ∈ Ω, where Ω is the cell domain. Note that Kirchhoff's voltage law is already fulfilled
by the fundamental theorem of calculus.
Additionally, boundary conditions are required for unique solutions: The first case is that the
boundary of a cell is connected to a voltage source. That means that the potential on the
boundaries is described by a known function 𝜑0, i.e.
𝜑(𝑥) = φ0(𝑥) ∀𝑥 ∈ 𝜕Ω.
(𝑆5 − 4)
The boundary of the cell domain Ω is denoted by 𝜕Ω.
The second case is that there is no electrical connection between the cell boundary and a voltage
source, e.g. isolation by air. Then the boundary condition is given by
∇𝜑(𝑥) ⋅ 𝑛 = 0 ∀𝑥 ∈ 𝜕Ω,
(𝑆5 − 5)
where 𝑛 is the outgoing normal of the domain Ω. Note that these two cases of boundary
conditions can be mixed. An example is given in Fig. S5.3. The cell has two electrodes with a
voltage of 𝑈0 and two isolated edges.
𝜑(𝑥) = 𝑈0
𝜕𝛺
𝛺
𝑛
𝛻𝜑(𝑥) ⋅ 𝑛 = 0
𝜑(𝑥) = 0
Fig. S5.3
Example of a cell domain 𝛺. The continuous edge (top and bottom) corresponds
to the cell electrodes with a voltage of 𝑈0. The dashed line (left and right) indicates the part
without any voltage source.
Analytical solution for the one-dimensional case:
For the one-dimensional case 𝜕𝑥 (
𝜕𝑥
𝑅(𝑥)
) = 0 we can solve the equation analytically. First, we
express the resistance in the form of conductivity, i.e. 𝑅(𝑥) =
1
𝜎(𝑥)
. By integrating over 𝑥 we get:
𝑥
(𝑥) = 𝑐0 ∫ 𝑅(𝑠)
𝑑𝑠 + 𝑐1
0
(𝑆5 − 6)
We consider an area of length 𝑏, with boundary values (0) = 0, (𝑏) = 𝑈0 , hence yielding a
voltage of 𝑈0. With the boundary values, it follows that, 𝑐1 = 0, 𝑐0 = 𝑈0
1
𝑏
∫ 𝑅(𝑠)
0
𝑑𝑠
. As one can
see, 𝑐0 is the current of the system. We assume that the conductivity of a photoconductor device
changes with the light intensity and the geometry of the cell.
We consider a one-dimensional photoconductor. If the length is fixed, the conductivity only
depends on the light intensity. The dark conductivity 𝜎𝑑 changes upon illumination by Δ𝜎 = 𝑝 ⋅
𝐸, Δ𝜎 hence depending linearly on the irradiance 𝐸. In Fig. S6.4 a one-dimensional
photoconductor is shown schematically.
𝑏
𝑎
𝑟
𝑎
illuminated part
Fig. S5.4
scheme of a partially illuminated one-dimensional photoconductor.
Now we can compute ∫ 𝑅(𝑠)
𝑏
0
𝑏
𝑑𝑠 = ∫
0
1
𝜎(𝑠)
𝑎
𝑑𝑠 = ∫
0
1
𝜎𝑑
𝑎+2𝑟
𝑑𝑠 + ∫
𝑎
1
𝜎𝑑+𝑝𝐸
𝑑𝑠
+ ∫
2(𝑎+𝑟)
𝑎+2𝑟
1
𝜎𝑑
𝑑𝑠
=
2𝑟
1
𝜎𝑑+𝑝𝐸
+ 2𝑎
1
𝜎𝑑
=
2𝑟𝜎𝑑+2𝑎(𝜎𝑑+𝑝𝐸)
𝜎𝑑(𝜎𝑑+𝑝𝐸)
. Now we have 𝑐0 =
𝜎𝑑(𝜎𝑑+𝑝𝐸)
2𝑟𝜎𝑑+2𝑎(𝜎𝑑+𝑝𝐸)
𝑈0.
Theorem: c0(E) is affine linear for 𝑎 = 0 (i.e. the sensor is fully and uniformly illuminated) and
non-linear for 𝑎 > 0.
Proof: We derive 𝑐0(𝐸) =
𝜎𝑑(𝜎𝑑+𝑝𝐸)
2𝑟𝜎𝑑+2𝑎(𝜎𝑑+𝑝𝐸)
𝑈0 w.r.t. 𝐸 and get
𝑑
𝑑𝐸
𝑐0(𝐸) =
2𝑟
𝑝𝜎𝑑
2(𝑎(𝑝𝐸+𝜎𝑑)+𝜎𝑑𝑟)2 𝑈0.
Trivially
𝑑
𝑑𝐸
𝑐0(𝐸) does not depend on 𝐸 for 𝑎 = 0, otherwise for 𝑎 ≠ 0 it does. This proves that
𝑐0(𝐸) is affine linear for 𝑎 = 0.
q.e.d.
S6 – Long-range measurements over 72 m using amorphous silicon-based FIP detectors
Long-range measurements up to 72 m are performed using an amorphous silicon based FIP
detector. The distance to a modulated LED emitting at 660 nm2 is determined using the FIP
technique. The LED light is collected by a 95 mm diameter commercial lens3 which directs the
converging beam towards the sensors. A stack of two semitransparent amorphous silicon solar
cells4 operating at short circuit is used as detector. To cover distances beyond 36 m, a 20 x 30 cm
large mirror is used as reflector, allowing for measurements up to 72 m.
2 Thorlabs model M660L3. 700 mW emitted optical power. 50% duty cycle, 2225 Hz,
https://www.thorlabs.com/thorproduct.cfm?partnumber=M660L3
3 Walimex
mount/produkt/walimex-pro-50063-dslr-mirror-c-mount-white.html)
4 40 x 40 mm active area, 500 nm i-layer thickness. Custom-made by Solems S. A.
(http://www.walimexpro.de/en/video/video-lens/c-
lens
Pro
500mm
Fig. S6.1
Short circuit photocurrent of the first sensor at various positions behind the lens
for several LED distances ranging from 12 m to 72 m. The dashed lines indicate the sensor
position used for the quotient determination.
Fig. S6.2
Short circuit photocurrent of the second sensor at various positions behind the
lens for several LED distances ranging from 12 m to 72 m. The light passes through the first
sensor before impinging on the second. The dashed lines indicate the sensor position used for the
quotient determination.
Figures S6.1 and S6.2 show the signals of both sensors in the detector stack. The data is collected
for each LED distance by moving the stack behind the lens on the optical axis. The photocurrent
in the nano Ampere regime is measured using a Fourier tranform technique5.
In order to perform distance measurements in the range between 12 m and 72 m, the first sensor
is positioned 76 mm and the second 110 mm from the back of the lens. Calculating the ratio of
5 The signal is amplified with a Femto DLPCA and recorded with a Behringer U-Phoria UMC202HD sound card
the first sensor signal to the second sensor at these positions for various LED distances yields the
calibration curve depicted in Fig. S6.3. The quotient of the sensor signals is shown to
monotonically increase over the entire measurement range, making it possible to assign a single
quotient value to any distance up to 72 m.
Fig. S6.3
Calibration curve for distance measurements generated by calculating the
quotient of the two sensor signals at 12m, 24 m, 36 m, 48 m, 60 m and 72 m. The line acts as
guide to the eye.
|
1711.02815 | 1 | 1711 | 2017-11-08T03:43:20 | Electric-Field Guided Precision Manipulation of Catalytic Nanomotors for Cargo Delivery and Powering Nanoelectromechanical Devices | [
"physics.app-ph"
] | We report a controllable and precision approach in manipulating catalytic nanomotors by strategically applied electric (E-) fields in three dimensions (3-D). With the high controllability, the catalytic nanomotors have demonstrated new versa-tility in capturing, delivering, and releasing of cargos to designated locations as well as in-situ integration with nanome-chanical devices (NEMS) to chemically power the actuation. With combined AC and DC E-fields, catalytic nanomotors can be accurately aligned by the AC E-fields and instantly change their speeds by the DC E-fields. Within the 3-D orthog-onal microelectrode sets, the in-plane transport of catalytic nanomotors can be swiftly turned on and off, and these cata-lytic nanomotors can also move in the vertical direction. The interplaying nanoforces that govern the propulsion and alignment are investigated. The modeling of catalytic nanomotors proposed in previous works has been confirmed quan-titatively here. Finally, the prowess of the precision manipulation of catalytic nanomotors by E-fields is demonstrated in two applications: the capture, transport, and release of cargos to pre-patterned microdocks, and the assembly of catalytic nanomotors on NEMS to power the continuous rotation. The innovative concepts and approaches reported in this work could further advance ideal applications of catalytic nanomotors, e.g. for assembling and powering nanomachines, nano-robots, and complex NEMS devices. | physics.app-ph | physics | Electric-Field Guided Precision Manipulation of Catalytic Nanomotors
for Cargo Delivery and Powering Nanoelectromechanical Devices
Jianhe Guo1, Jeremie June Gallegos2, Ashley Robyn Tom2, and Donglei Fan1, 2*
1 Materials Science and Engineering Program, the University of Texas at Austin, Austin, TX 78712, USA
2 Department of Mechanical Engineering, the University of Texas at Austin, Austin, TX 78712, USA
KEYWORDS: Nanomotors, Catalytic Nanomotors, Electric Tweezers, Cargo Delivery, NEMS, rotary NEMS, nano-
robotics
ABSTRACT: We report a controllable and precision approach in manipulating catalytic nanomotors by strategically applied
electric (E-) fields in three dimensions (3-D). With the high controllability, the catalytic nanomotors have demonstrated
new versatility in capturing, delivering, and releasing of cargos to designated locations as well as in-situ integration with
nanomechanical devices (NEMS) to chemically power the actuation. With combined AC and DC E-fields, catalytic nanomo-
tors can be accurately aligned by the AC E-fields and instantly change their speeds by the DC E-fields. Within the 3-D
orthogonal microelectrode sets, the in-plane transport of catalytic nanomotors can be swiftly turned on and off, and these
catalytic nanomotors can also move in the vertical direction. The interplaying nanoforces that govern the propulsion and
alignment are investigated. The modeling of catalytic nanomotors proposed in previous works has been confirmed quanti-
tatively here. Finally, the prowess of the precision manipulation of catalytic nanomotors by E-fields is demonstrated in two
applications: the capture, transport, and release of cargos to pre-patterned microdocks, and the assembly of catalytic na-
nomotors on NEMS to power the continuous rotation. The innovative concepts and approaches reported in this work could
further advance ideal applications of catalytic nanomotors, e.g. for assembling and powering nanomachines, nanorobots,
and complex NEMS devices.
INTRODUCTION
The integration of autonomous inorganic micro/nanomo-
tors as components of micro/nanomachines with high pre-
cision and versatility to power their operations is a critical
step in realizing the ideal nanofactories and nanorobots,
which could revolutionize modern lives.1-9 Catalytic mi-
cro/nanomotors that convert chemical energy into me-
chanical motions, are one of the most widely exploited au-
tonomous nanomotors.10-18 Billions of catalytic nanomotors
can be facilely fabricated by using a variety of techniques,
such as electrodeposition into nanoporous templates and
electron beam deposition on monolayer nanospheres.1-2
They self-propel by harvesting chemical energies from
fuels in suspension, such as hydrogen peroxide.18-19 Re-
cently, substantial research efforts have been focused on
strategically designing and fabricating catalytic nanomo-
tors with an array of compositions and geometries, such as
bimetallic nanorods,18-20 catalytic microtubes,21-23 and Janus
particles24-27. The efforts lead towards dramatic improve-
ment of propulsion speeds up to hundreds of μm/sec (or
100 body lengths per second),19-21 and readiness in harness-
ing energy from a variety of fuels, such as hydrazine,28
urea29-30 and even pure water13, 26. More importantly, vast
applications of catalytic nanomotors have been demon-
strated, such as on-chip cargo transport,24, 31 drug deliv-
ery,32-33 microchip repair,34 nanolithography,35 biomolecu-
lar sensing and in-vivo disease treatment.36
However, it remains challenging to align catalytic na-
nomotors with high precision and modulate their moving
speeds facilely and instantly. The ability to achieve this
could open unprecedented opportunities. Innovatively,
magnetic fields have been exploited in guiding catalytic na-
nomotors, however, this strategy requires the integration
of magnetic elements in the nanomotors and precise align-
ment of magnetic moments.24, 31 Also, to generate magnetic
forces, bulky electromagnets are often employed, which
could be the bottleneck when developing portable na-
nomotor based devices. Acoustic tweezers have been used
in guiding catalytic nanomotors to aggregate and dis-
perse,37 while, the resolution in manipulation is restricted
by the large wavelength of acoustic waves. Besides control-
ling the orientation of catalytic nanomotors, it is of para-
mount importance to facilely tune their speed. Several
unique approaches have been exploited to control locomo-
tion speed of catalytic nanomotors. With the strong de-
pendence on catalytic reactivity, the speed of catalytic mi-
cro/nanomotors can be tuned by localized stimuli, includ-
ing fuel concentration,18 temperature,38 and light illumina-
tion.26, 39-41 Furthermore, by applying electrical potentials to
create chemical gradient,42 or generating ultrasonic
waves,37, 43 the speed of micro/nanomotors can also be
modulated. However, it remains difficult to realize both
the guiding and speed tuning of catalytic nanomotors with
high accuracy, facileness, and in an all-on-chip manner.
Scanning electron microscopy (SEM) and energy-disper-
sive X-ray spectroscopy (EDS) in Fig. 1(b)-(c) confirm the
uniform cylindrical morphology, controlled size and com-
position of the Pt-Au catalytic nanomotors.
A software interfaced 3-D orthogonal microelectrode
setup is designed and constructed for guiding the catalytic
nanomotors as shown in Scheme Fig. 2. The set-up in-
cludes in-plane quadruple microelectrodes patterned on a
glass slide and a pair of indium tin oxide (ITO) parallel
electrodes assembled in the vertical direction. The bottom
ITO electrode is fabricated on the opposite side of the
quadruple microelectrodes on the glass (1 mm in thick-
ness). A suspension reservoir is formed on top of the quad-
ruple microelectrodes by a piece of polydimethylsiloxane
(PDMS) elastomer (~1 mm in thickness) with a well of ~ 4
mm in diameter. The top of the well is sealed with a second
ITO electrode for providing E-field in the vertical direction.
In this work, we report a unique approach for manipu-
lating catalytic nanomotors with high precision and facile-
ness. The work is based on strategically combined AC and
DC E-fields, the so-called electric tweezers, applied via a 3-
D orthogonal microelectrode setup.44 Here, the DC E-field
tunes the transport speed via electrophoretic and elec-
troosmosis effects. The AC E-field guides the alignment in-
dependently via electric torques on the induced dipoles of
nanomotors. By applying the combined AC and DC E-fields
in 3-D, catalytic nanomotors can instantly align, transport
along defined directions, start and stop, and change speeds
on demand. The involved various nanoforces governing
the motions are investigated. Leveraging the high precision
in the alignment, the linear dependence of speed on the
inverse of size of nanomotors (1/l) down to submicrome-
ters, is experimentally determined, confirming previous
theoretical predictions.14 Finally, the manipulation of cata-
lytic nanomotors by E-fields is demonstrated for two appli-
cations: the dynamic loading, transport, and unloading of
micro-targets to pre-patterned microdocks; and assem-
bling and integration of a catalytic nanomotor on a rotary
NEMS to power its continuous operation.
Figure 2. Scheme of 3-D orthogonal microelectrode setup.
Applying an AC E-field on the in-plane quadruple micro-
electrodes, the catalytic nanomotors can be instantly
aligned along the direction of the AC E-field and move au-
tonomously in the direction with the Pt segment as the
front, as shown in the schematic diagram in Fig. 1(a), over-
lapped images in Fig. 1(d) and movie S1 (Supporting Infor-
mation). We found that an AC peak-to-peak voltage of 10
V is sufficient to align the catalytic nanomotors and to
guide their motions. When superimposing a DC E-field on
the AC E-field, the speed of the nanomotors can increase
or decrease instantaneously, and even reverse the moving
direction. We characterize the manipulation in detail in
the following to unveil the fundamental interactions be-
tween the nanomotors and E-fields.
First, we investigate the dependence of speed of the cat-
alytic nanomotors on the concentration of hydrogen per-
oxide fuel (H2O2), both with and without E-fields. The av-
erage speed of the catalytic nanomotors is determined sta-
tistically from the behaviors of 10 nanomotors for 10 sec-
onds (details in the Supporting Information). As shown in
Fig. 3(a), regardless whether the E-field is applied or not,
the speed of the nanomotors increases with the concentra-
tion of H2O2 and reaches a plateau, which can be attributed
to the saturation of catalytic active sites on the nanomotors
in high concentration H2O2 fuels.18 Consistently, we find
that the speed of these catalytic nanomotors under AC E-
Figure 1. (a) Schematic diagram of 3D manipulations of Pt-
Au catalytic nanomotors in H2O2 fuel with AC E-fields. (b)
Scanning electron microscopy (SEM) and (c) Energy-dis-
persive X-ray spectroscopy (EDX) images of Pt-Au catalytic
nanomotors (250 nm in diameter, 5 μm in length; consist-
ing of 2-μm Pt segment and 3-μm Au segment). (d) Over-
lapped snapshots of a catalytic nanomotor guided by AC E-
fields.
RESULTS AND DISCUSSION
Guide, Start and Stop, and Modulate Speed of Cata-
lytic Nanomotors
The demonstrations of transport guidance and speed mod-
ulation of catalytic nanomotors by E-fields are carried out
by using the classical Pt-based bimetallic nanorod motors
as a model system [Fig. 1(a)]. Arrays of multi-segmented Pt-
Au nanorod motors are synthesized with controlled
lengths and diameters by electrodeposition into nanopo-
rous templates in a three-electrode setup.18, 45 The fabrica-
tion details are provided in the Supporting Information.
fields is always higher than those without AC E-fields. Fur-
thermore, the speed of all nanomotors, regardless of their
moving directions, increases when applying a uniform AC
E-field. Neither the electroosmosis flows nor the induced
dielectrophoretic forces could result in the observed be-
havior, so the dominating factor is the reduction of rota-
tional Brownian motions of nanomotors due to the align-
ment by AC E-fields. Our experimental results and analysis
support this understanding. First, we determine the speed
of nanomotors as a function of AC frequency and voltage.
It is found that the average speed of nanomotors monoton-
ically increases with AC frequency before reaching a con-
stant at around 500 KHz and 20V. At a fixed AC frequency,
i.e. 500 KHz, the average speed also increases with AC volt-
age amplitude until reaching a constant at 30 V as shown
in Fig. 3(b)-(c). The voltage dependence can be readily un-
derstood from the increase of electric torque (𝜏𝑒) with ap-
plied E-field (E), which counters the rotational Brownian
motions and thus enhances the degree of alignment of a
nanomotor, given by:46
𝜏𝑒 =
8𝜋𝑙𝑟2
3
∙ 𝑃 ∙ 𝐸2 ∙ 𝑠𝑖𝑛2𝜃
(1)
where l and r are the length and radius of nanomotor, re-
spectively; θ is the angle between the long axis of nanomo-
tor and the E-field; P is a value determined by the permit-
tivity and conductivity of the medium and nanomotor as
well as the AC frequency. We observe a leveling off of the
moving speed when the voltage is adequately high, i.e.
above 30 V, as shown in Fig. 3(b). It could be understood
that when the electric torque is sufficiently high, the rota-
tional Brownian motion is suppressed to an extent that the
increase of speed with voltages is too small to be deter-
mined compared to the statistic distributions of speed of
tested nanomotors.
By analyzing electric torques, we can also attribute the
observed dependence of nanomotor speed on AC fre-
quency [Fig. 3(c)] to the suppressed rotational Brownian
motions. We experimentally determine electric torques as
a function of angular positions when aligning nanomotors
at 5 KHz to 50 MHz with Eq. (1). The 𝜏𝑒 exerted on the cat-
alytic nanomotors at an angle of θ can be readily obtained
from the angular velocity (𝜔) versus angle (θ) [Fig. S1, sup-
porting information], since the viscous torque ( 𝜏𝜂 ) in-
stantly counters the electric torque (𝜏𝑒) as given by 𝜏𝑒 = 𝜏𝜂
in low Reynolds number environment. Here, the viscous
drag torque 𝜏𝜂 on a rotating nanorod is calculated as fol-
lows:47
𝜏𝜂 =
1
3
𝜔𝜋𝜂𝑙3
𝑁3−𝑁
𝑁3(ln
𝑙
𝑁𝑟
+0.5)
= 𝐾1 𝜔 (𝑁 𝑚)
(2)
where η is the viscous coefficient of suspension medium
and N is the number of nanorod segments, taken as 2 in
this calculation. For the nanomotors used in our experi-
ments, the constant 𝐾1 is determined as 2.50×10−20. Now,
with 𝜏𝑒 determined at different θ from Eq. (1) and (2), we
can readily obtain the coefficient of 𝜏𝑒 at different AC fre-
quencies as shown in Fig. 3(c). It can be readily found that
the dependence of nanomotor speed on AC frequency well
matches the dependence of electric torque applied on na-
nomotors versus AC frequency. It well supports the key
contribution of alignment by AC E-fields to the observed
enhanced speed of nanomotors.
Figure 3. (a) Speed of catalytic nanomotors versus concentration of H2O2 with/without AC E-fields (5 MHz, 15 V). (b) Speed versus
peak-to-peak voltage of AC E-fields (500KHz) in 7.5 wt% H2O2 solution. (c) Speed versus frequency of AC E-fields (20 V) in 7.5
wt% H2O2 solution. Inset: electric torque versus frequency of AC E-fields. (d) Trace of catalytic nanomotors of different length in
2 seconds. The lengths of nanomotors (from top to bottom) are 0.82, 2.50, 4.78, 6.68 and 8.96 μm, respectively. The green dashed
line shows the starting position. (e) Log-log plot of speed versus length of nanomotors. The slope is -0.98. (f) Speed of Au nanorods
and Pt-Au catalytic nanomotors of same dimensions versus DC E-fields in different suspension mediums.
With the uniform AC E-fields, we not only improve the
alignment and speed of nanomotors as discussed as above,
but also successfully guided catalytic nanomotors in the
vertical direction in a facile and all-on-chip manner. Na-
nomotors can move along prescribed trajectories in the 2-
D X-Y plane, start and stop on demand, and even move
vertically as shown in Fig. 1(a), Movie S1 and S2 in Support-
ing Information. For nanomotors moving in the X-Y 2-D
plane, it is known that the catalytic driving force balances
with the drag force, given by:48
𝐹𝑐𝑎𝑡𝑙𝑦𝑡𝑖𝑐 𝑑𝑟𝑖𝑣𝑒 = 𝐹𝑑𝑟𝑎𝑔 =
2𝜋𝜂𝑙
𝑙
−0.5
𝑟
𝑙𝑛
𝜈 = 𝐾2 𝜈 (𝑁)
(3)
where v is the velocity of catalytic nanomotors and the ge-
ometric factor 𝐾2= 8.77×10−9 for the tested nanomotors (5
μm in length and 250 nm in diameter). Note that the value
of viscous coefficient η of suspension medium is estimated
from that of pure water at room temperature, knowing that
the change of viscous coefficient is less than 3% when the
concentration of H2O2 is less than 10 wt%.49 In the vertical
direction, a catalytic nanorod made of Pt(2-μm)-Au(3-μm)
experiences gravitational and buoyant forces of 4.85×
10−14𝑁 and 0.24×10−14𝑁 , respectively. Therefore, from
Eq. (3) and the experimental results of velocity versus fuel
concentration (CH2O2) shown in Fig. 3(a), we can readily de-
termine that the catalytic driving force of nanomotors is
sufficiently high to realize the propulsion of the nanomo-
tor in the vertical direction when the concentration of fuel
(CH2O2) is above 10 wt%. This analysis is validated by exper-
imental study as shown in movie S2 in the Supporting In-
formation. When CH2O2 is 12.5 wt%, where a driving force
of 5.69×10−14𝑁 is determined by calculation, the catalytic
nanomotors indeed aligned and transported vertically,
overcoming gravitation forces, and gradually disappeared
from the view. While at a lower fuel concentration, e.g.
when CH2O2 is 7.5 wt%, a vertically applied AC E-field can
instantly align nanomotors vertically and stop the motions
in 2-D planes. Here the driving forces due to catalytic reac-
tions, calculated as 3.17×10−14𝑁, is not sufficient to propel
the motors vertically. Both the demonstrations of 3-D ma-
nipulation at high fuel concentrations and "on/off" control
of 2-D motion at low fuel concentrations offer considerable
promise for versatile operations of these catalytic nanomo-
tors, opening many opportunities for applications. Such a
strategy could also be applied for controlling photocata-
lytic nanomotors26, 40 and self-propelled enzyme nanomo-
tors12, 30.
The high precision in the alignment of nanomotors by
AC E-fields offers new opportunities in studying and un-
derstanding the working mechanism of the self-propelled
catalytic nanomotors. In a recent model of a catalytic na-
nomotor made of bimetallic nanorods, the velocity (𝑣) due
to self-electrophoresis is given by:14
𝑣~
𝜁𝑝𝜀𝜀0
𝜂
𝐸𝑖𝑛𝑡~
𝜁𝑝𝜀𝜀0𝛥𝜙
𝜂𝑙
(4)
where ε is the relative dielectric constant of the suspension
medium and 𝜀0 is the dielectric permittivity of free space.
The self-generated E-field 𝐸𝑖𝑛𝑡 is approximated as the po-
tential drop 𝛥𝜙 divided by the length of the nanorods l and
𝛥𝜙 is determined by the chemical potential of the two
metal segments of a nanomotor, which is considered as a
constant with different lengths; 𝜁𝑝 is Zeta potential of a na-
nomotor, independent of the size of the nanomotor (Fig.
S3). Therefore, we can readily find that the self-electropho-
retic velocity (𝑣) of a nanomotor should be inversely pro-
portional to its length (l) according to Eq. (4) given by the
model. Although in previous studies, it has been shown
that the smaller the catalytic nanomotors, the higher the
moving speed, quantitative study has yet been done on the
catalytic nanomotors as predicted by the model.14, 50 To ex-
perimentally determine the dependence of speed on size of
nanomotors, the key is to precisely align nanomotors with
controlled moving trajectories and to suppress noises due
to Brownian motion. As shown in the above studies, AC E-
field alignment provides a facile and effective tool, where
the alignment is determined by the overall shape anisot-
ropy of a nanomotor. Indeed, with AC E-fields, we can fac-
ilely synchronously align catalytic nanomotors with sizes
ranging from 0.82 to 8.96 μm (details in Fig. S4). When
suspending in H2O2 solutions, the shorter nanomotors ex-
hibit higher moving speeds compared to longer ones [Fig.
3(d)]. The speed and length of nanomotors follows an in-
verse proportional relationship as shown in the log-log plot
with a slope of -0.98±0.04 in Fig. 3(e). It indicates that the
power law dependence of speed on size of nanomotors is
approximately -1, which provides quantitative proof of the
working mechanism proposed previously for nanomotors
in the size range of 0.82 to 8.96 μm. This result points to-
wards the great advantages of catalytic nanomotors in
achieving ultrahigh propulsion speed when made into ul-
tra-small dimensions.23, 51-52
Next, we exploit the effect of DC E-fields on nanomotors.
The speed of the Pt-Au catalytic nanomotors due to DC E-
fields are tested in pure water and 5-10 wt% H2O2 fuels.
Gold (Au) nanorods of the same dimension are fabricated
and tested at the same conditions for control experiments.
AC E-fields are superimposed to align the nanomotors in
their long axis direction when applying DC voltages of -1 to
1 V on 500 µm-gapped microelectrodes. The speed of cata-
lytic nanomotors linearly depends on DC E-fields and can
be controlled to instantly increase, decrease and even re-
verse directions, depending on the magnitude and direc-
tion of DC E-fields as shown in Fig. 3(f) and movie S3. A DC
voltage as low as 1 V can lead to a velocity change of ~20
μm/sec of the catalytic nanomotors, which is effective in
modulating speed of nanomotors for various applications
[Fig. 3(f)].
We also observe that the speeds of both catalytic na-
nomotors in fuel solutions and those in control experi-
ments always increase in the direction of DC E-fields.
While the Zeta potentials of both Pt-Au nanomotors and
Au nanorods are negative as shown in the measurements
in Fig. S2. Furthermore, although the speed of the catalytic
nanomotors show much stronger responses to the DC E-
field, the value of their Zeta potential in fuel solutions is
significantly lower than those of control samples, i.e. for
Pt-Au nanomotors in 5% H2O2 fuel, the Zeta potential of -
8.76 mV is around 1/4 of those of control samples.
The above phenomena cannot be explained simply by
electrophoresis (EP). Nanoentities, such as nanomotors,
experience an electrophoretic force in the presence of an
external DC E-field due to the formation of an electric dou-
ble layer at the solid/liquid interface with opposite local
charges. The speed of nanoentities (𝒗𝐸𝑃) due to electro-
phoretic force is proportional to the dielectric constant of
suspension medium (ε), Zeta potential of nanomotor (𝜁),
and DC E-field (Eext), given by:53
𝒗𝐸𝑃 =
𝜀𝜀0𝜁
𝜂
𝑬𝑒𝑥𝑡
(5)
Given that the Zeta potentials of the catalytic nanomo-
tors and control samples are all negative, if only because of
the electrophoretic effect, the moving speed should in-
crease in the opposite direction of the DC E-field and scale
with the magnitude of the Zeta potential. However, the ex-
periments show the behaviors of nanomotors opposite to
this analysis. In-depth study shows that in addition to elec-
trophoretic (EO) forces, electroosmosis flows generated on
the surface of glass substrates are important as well, which
is given by:53
𝜀𝜀0𝜁𝑠
(6)
𝒗𝐸𝑂 = −
𝑬𝑒𝑥𝑡
𝜂
The Zeta potential of glass substrates (𝜁𝑠) is negative and
can reach -80 mV.54 As a result, the direction of electroos-
mosis flow on glass substrates is in the same direction of
DC E-field according to Eq. (6), opposite to that of the elec-
trophoretic effect on nanomotors in Eq. (5). Therefore, the
speed of a nanomotor (𝒗𝐸) in DC E-field is governed by the
combined effects of electrophoretic force on the nanomo-
tor (𝒗𝐸𝑃) in Eq. (5) and electroosmosis flows on the glass
substrate (𝒗𝐸𝑂) in Eq. (6). Considering the zeta potentials
and the moving direction of nanomotors, it can be found
that the electroosmosis flows (𝒗𝐸𝑂) dominate the motions
of nanomotors. Therefore, lower absolute values of the
negative zeta potential of the nanomotors lead to higher
speed modulations by the DC E-field. This well agrees with
the experimental results in Fig. 3(f), where the speed of cat-
alytic nanomotors can be strongly tuned by the external
DC E-field (Supporting Information).
With the demonstration and understanding of the prow-
ess of AC and DC E-fields for manipulation of catalytic na-
nomotors with high facileness, precision, and efficiency,
we exploited two applications of these catalytic nanomo-
tors: targeted cargo delivery and assembling of catalytic na-
nomotors for powering rotary NEMS devices.
Targeted Cargo Delivery
In this demonstration [Fig. 4(a) and movie S4], Au nano-
rods (250 nm in diameter and 3.6 μm in length), fabricated
by electrodeposition into nanoporous templates, serve as
cargos. They are mixed with Pt-Au catalytic nanomotors in
7.5 wt% H2O2 fuel solution. Without AC E-field, the cata-
lytic nanomotors move randomly, while the Au cargos only
exhibit weak Brownian motions. By applying a uniform AC
E-field, both nanomotors and cargo nanorods are aligned.
The nanomotors move in the alignment direction. While,
the cargo stays essentially at the original location during
the transport of the nanomotor guided by the AC E-field as
shown in Fig. 4(b)-(c). When the nanomotor are close to
the cargo, the induced E-fields can readily assemble them
tip to tip. Here, the interaction between the induced di-
poles of nanomotor and cargo ensures a simple procedure
to upload the cargo on the nanomotor and the assembly is
robust during the transport in AC E-fields. Next, guided by
the AC E-field, the nanomotor propels the cargo to a pat-
terned metallic microdock as shown in Fig. 4(c)-(d). When
the cargo is in the vicinity of the microdock, it can rapidly
anchor to the edge of the metallic microdock due to the
interaction between induced electric dipoles. Then the na-
nomotor can start its next journey to seek other targets as
shown in Fig. 4(e)-(f). Here, the on-demand release of car-
gos from the catalytic nanomotors are facilely accom-
plished by turning off the AC E-field swiftly, e.g. for a few
seconds [Fig. 4(e)]. The above process is further deter-
mined quantitatively by analyzing the instantaneous speed
of nanomotors as shown in Fig. 4(g). When the nanomotor
moves towards the cargo, the average speed is around 3.9
μm/sec. The speed swiftly increases to 5.7 μm/sec when it
gets close and attaches to the cargo due to the strong mu-
tual attraction. After capturing the cargo, the speed de-
creases to 2.2 μm/sec due to the load, which agrees with
our estimation of 2.6 μm/sec from Eq. (3). After releasing
the cargo, the nanomotor immediately restores to a speed
up to 3.3 μm/sec, which is slightly lower than the speed of
3.6 μm/sec before loading the cargo. Note that at this mo-
ment the nanomotor is not aligned by the AC E-field which
could account for the lowered speed.
Figure 4. (a) Scheme of targeted cargo delivery. (b-f) Optical
microscopy images show the dynamic process of a Pt-Au cat-
alytic nanomotor when picking up, pushing, and delivering a
nanorod cargo to the designated microdock. (g) Speed of the
nanomotor in the cargo delivery process.
The above targeted cargo delivery is the first demonstra-
tion of E-field enabled loading and unloading of cargo by
nanomotors, showing simplicity and reliability for diverse
lab-on-a-chip applications.
Powering Rotary NEMS by Precision Assembling Cata-
lytic Nanomotors
structure
Next, leveraging the precision guidance of catalytic na-
nomotors with E-fields, we designed and assembled a
unique type of chemically powered rotary NEMS by inte-
grating catalytic nanomotors with electric manipulation.
Firstly, a rotary NEMS is assembled by E-fields following
previous reports as illustrated in Fig. 5(a).47, 55 It consists of
a multi-segmented Au/Ni nanorod (250 nm in diameter
and 8 μm in length) and a patterned nanomagnet (500 nm
in diameter) serving as the rotor and bearing, respectively.
Different from previous research, the rotor is designed to
consist of three Ni segments embedded in the Au nanorod.
The
is Au(200 nm)/Ni(50 nm)/Au(3.5
µm)/Ni(500 nm)/Au(3.5 nm)/Ni(50 nm)/Au(200 nm) as
shown in Fig. 5(b). The Ni segment in the center of the ro-
tor is used to attach the rotary NEMS on the patterned
magnetic bearing. The two Ni segments next to the tips
aim to anchor the catalytic nanomotors. The patterned na-
nomagnet has a tri-layer stack of Ti (60 nm)/Ni (80 nm)/Cr
(6 nm). By using the electric tweezers, the nanorotor is
transported and attached atop of the pattern magnetic
bearing, where the magnetic interaction between the Ni
segment in the rotor and Ni layer in the magnetic bearing
fixes the position of the rotary NEMS while still allows its
rotation.47, 55 Next, we fabricated catalytic nanomotors
made of Pt-AgAu nanowires. Here, we replaced the previ-
ously used Au segments with Ag(~80 wt%)/Au(~20 wt%)
alloys in the Pt-based catalytic nanomotors to substantially
increase the output power and speed due to the enhanced
potential differences between the two segments as shown
in Fig. 5(c-d).19 The catalytic nanomotors move at a speed
up to ~30 μm/sec in 5 wt% H2O2 solution. From the speed,
we can estimate the driving force as high as 0.26
piconewtons (pN). With the guidance of the AC E-field, the
catalytic nanomotor can be efficiently maneuvered to-
wards the rotary NEMS and then be assembled at one end
of the rotor by magnetic attraction as shown in Fig. 5(a).
After assembling, the nanomotor instantly drives the
NEMS into continuous rotation, all powered by the chem-
ical energy harvested from the H2O2 fuels. The driving
torque can be readily determined as 1.04×10−18𝑁 ∙ 𝑚 ,
which is sufficient to overcome the friction and magnetic
torques between the rotor and bearing as shown in Fig. 5(e-
g) and movie S5. Here, we can observe the effects of the
angle-dependent magnetic force and torque that result in
the oscillation of rotation speeds of the rotor with a 360
periodicity as a function of the angular position, agreeing
with previous works.47, 56 The average rotation speed of the
NEMS device is determined as 0.64 rad/sec.
This work demonstrates a new approach in integrating
powering components to drive nanomechanical devices,
which could be applied to assemble various functional
components of nanorobots and complicated NEMS/MEMS
devices with far-reaching impact in electronics and bio-
medical research.
Figure 5. Assembling of catalytic a nanomotor for powering
NEMS device into continuous rotation. (a) Schematic diagram
of the rotary NEMS device with catalytic nanomotor assem-
bled as the powering component. (b) SEM images of the multi-
segment nanowire rotor. The segments from left to right are
200 nm Au, 50 nm Ni, 3.5 μm Au, 500 nm Ni, 3.5 μm Au, 50
nm Ni and 200 nm Au. (c-d) SEM and EDX images of the Pt-
Ag/Au catalytic nanomotor. The segments from left to right
are 400 nm Au, 1.7 μm Ag-Au alloy, 500 nm Ni, 400 nm Au,
and 2 μm Pt. (e) Snapshots of the rotary NEMS device taken
every 2 seconds. (f) Rotation angle versus time. (g) Rotation
speed versus time.
CONCLUSIONS
In summary, we report a new, controllable and precision
approach to guide and modulation propulsions of catalytic
nanomotors with E-fields, which allows for a motion con-
trol in 2-D and 3-D. The fundamental interactions involved
in the electric manipulation have been investigated. Lever-
aging the precision of manipulation, we experimentally
confirmed the inverse linear dependence of speed and size
of catalytic nanomotors, supporting previous modeling.
For applications, the manipulation strategy provides facile
operations of these catalytic nanomotors to realize cargo
capture, transport, and delivery to designated microdock.
The prowess of the manipulation is also demonstrated by
assembling catalytic motors as powering component of ro-
tary NEMS. This innovative approach could be applied to
construct
functional
NEMS/MEMS devices for diverse tasks in electronics and
biomedical research.
nanorobots
various
and
ASSOCIATED CONTENT
Supporting Information.
This material is available free of charge via the Internet at
http://pubs.acs.org.
Experimental details, velocity analysis, torque calculation and
additional Zeta potential and SEM characterization of cata-
lytic nanomotors. (PDF)
Movie S1. Random movements of conventional catalytic na-
nomotor versus directional motions of E-field guided catalytic
nanomotor (2× frame rate).
Movie S2. Motions of catalytic nanomotors in the vertical di-
rection with the alignment of the AC E-fields. The fuels have
concentrations of 7.5 wt% and 12.5% H2O2, respectively.
Movie S3. Speed modulation of a catalytic nanomotor (2×
frame rate).
Movie S4. Targeted cargo delivery by a catalytic nanomotor
(5× frame rate).
Movie S5. Rotary NEMS device powered a catalytic nanomotor
(5× frame rate).
AUTHOR INFORMATION
Corresponding Author
* [email protected]
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENT
This research is supported by the Welch Foundation (F-1734)
and National Science Foundation (CMMI 1150767 and EECS
1710922). We appreciate the idea from Prof. Tom Mallouk in
studying the relationship of speed and size of the catalytic na-
nomotors. We also thank the discussion with Prof. Peer
Fischer on this work.
REFERENCES
(1) Wang, H.; Pumera, M., Chem. Rev. 2015, 115, 8704.
(2) Guix, M.; Mayorga-Martinez, C. C.; Merkoçi, A., Chem. Rev.
2014, 114, 6285.
(3) Kim, K.; Guo, J.; Xu, X.; Fan, D. L., Small 2015, 11, 4037.
(4) Kim, K.; Guo, J.; Liang, Z. X.; Zhu, F. Q.; Fan, D. L.,
Nanoscale 2016, 8, 10471.
(5) Wang, J.; Gao, W., ACS Nano 2012, 6, 5745.
(6) Xu, T.; Gao, W.; Xu, L.-P.; Zhang, X.; Wang, S., Adv. Mater.
2017, 29, 1603250.
(7) Wang, W.; Li, S.; Mair, L.; Ahmed, S.; Huang, T. J.; Mallouk,
T. E., Angew. Chem. 2014, 126, 3265.
(8) Qiu, T.; Lee, T.-C.; Mark, A. G.; Morozov, K. I.; Münster, R.;
Mierka, O.; Turek, S.; Leshansky, A. M.; Fischer, P., 2014, 5, 5119.
(9) Palagi, S.; Mark, A. G.; Reigh, S. Y.; Melde, K.; Qiu, T.; Zeng,
H.; Parmeggiani, C.; Martella, D.; Sanchez-Castillo, A.;
Kapernaum, N.; Giesselmann, F.; Wiersma, D. S.; Lauga, E.;
Fischer, P., Nat Mater 2016, 15, 647.
(10) Wang, W.; Duan, W.; Ahmed, S.; Mallouk, T. E.; Sen, A.,
Nano Today 2013, 8, 531.
(11) Li, J.; Esteban-Fernández de Á vila, B.; Gao, W.; Zhang, L.;
Wang, J., Sci. Robot. 2017, 2, eaam6431.
(12) Ma, X.; Hortelão, A. C.; Patiño, T.; Sánchez, S., ACS Nano
2016, 10, 9111.
(13) Wong, F.; Sen, A., ACS Nano 2016, 10, 7172.
(14) Wang, W.; Chiang, T.-Y.; Velegol, D.; Mallouk, T. E., J. Am.
Chem. Soc. 2013, 135, 10557.
(15) Lin, X.; Wu, Z.; Wu, Y.; Xuan, M.; He, Q., Adv. Mater. 2016,
28, 1060.
(16) Mirkovic, T.; Zacharia, N. S.; Scholes, G. D.; Ozin, G. A.,
ACS Nano 2010, 4, 1782.
(17) Wang, H.; Zhao, G.; Pumera, M., J. Am. Chem. Soc. 2014,
136, 2719.
(18) Paxton, W. F.; Kistler, K. C.; Olmeda, C. C.; Sen, A.; St.
Angelo, S. K.; Cao, Y.; Mallouk, T. E.; Lammert, P. E.; Crespi, V.
H., J. Am. Chem. Soc. 2004, 126, 13424.
(19) Demirok, U. K.; Laocharoensuk, R.; Manesh, K. M.; Wang,
J., Angew. Chem. Int. Ed. 2008, 47, 9349.
(20) Laocharoensuk, R.; Burdick, J.; Wang, J., ACS Nano 2008,
2, 1069.
(21) Solovev, A. A.; Mei, Y.; Bermúdez Ureña, E.; Huang, G.;
Schmidt, O. G., Small 2009, 5, 1688.
(22) Solovev, A. A.; Xi, W.; Gracias, D. H.; Harazim, S. M.;
Deneke, C.; Sanchez, S.; Schmidt, O. G., ACS Nano 2012, 6, 1751.
(23) Li, J.; Liu, W.; Wang, J.; Rozen, I.; He, S.; Chen, C.; Kim, H.
G.; Lee, H.; Lee, H.; Kwon, S.; Li, T.; Li, L.; Wang, J.; Mei, Y., Adv.
Funct. Mater. 2017, 27, 1700598.
(24) Baraban, L.; Makarov, D.; Streubel, R.; Mönch, I.; Grimm,
D.; Sanchez, S.; Schmidt, O. G., ACS Nano 2012, 6, 3383.
(25) Manjare, M.; Yang, B.; Zhao, Y. P., Phys. Rev. Lett. 2012,
109, 128305.
(26) Dong, R.; Zhang, Q.; Gao, W.; Pei, A.; Ren, B., ACS Nano
2016, 10, 839.
(27) Yoshizumi, Y.; Honegger, T.; Berton, K.; Suzuki, H.;
Peyrade, D., Small 2015, 11, 5630.
(28) Gao, W.; Pei, A.; Dong, R.; Wang, J., J. Am. Chem. Soc.
2014, 136, 2276.
(29) Ma, X.; Wang, X.; Hahn, K.; Sánchez, S., ACS Nano 2016,
10, 3597.
(30) Dey, K. K.; Zhao, X.; Tansi, B. M.; Méndez-Ortiz, W. J.;
Córdova-Figueroa, U. M.; Golestanian, R.; Sen, A., Nano Lett. 2015,
15, 8311.
(31) Burdick, J.; Laocharoensuk, R.; Wheat, P. M.; Posner, J. D.;
Wang, J., J. Am. Chem. Soc. 2008, 130, 8164.
(32) Peng, F.; Tu, Y.; van Hest, J. C. M.; Wilson, D. A., Angew.
Chem. Int. Ed. 2015, 54, 11662.
(33) Mou, F.; Chen, C.; Zhong, Q.; Yin, Y.; Ma, H.; Guan, J., ACS
Appl. Mater. Interfaces 2014, 6, 9897.
(34) Li, J.; Shklyaev, O. E.; Li, T.; Liu, W.; Shum, H.; Rozen, I.;
Balazs, A. C.; Wang, J., Nano Lett. 2015, 15, 7077.
(35) Li, J.; Gao, W.; Dong, R.; Pei, A.; Sattayasamitsathit, S.;
Wang, J., Nat. Commun. 2014, 5, 5026.
(36) Gao, W.; Dong, R.; Thamphiwatana, S.; Li, J.; Gao, W.;
Zhang, L.; Wang, J., ACS Nano 2015, 9, 117.
(37) Xu, T.; Soto, F.; Gao, W.; Dong, R.; Garcia-Gradilla, V.;
Magana, E.; Zhang, X.; Wang, J., J. Am. Chem. Soc. 2015, 137, 2163.
(38) Balasubramanian, S.; Kagan, D.; Manesh, K. M.; Calvo-
Marzal, P.; Flechsig, G.-U.; Wang, J., Small 2009, 5, 1569.
(39) Solovev, A. A.; Smith, E. J.; Bufon, C. C. B.; Sanchez, S.;
Schmidt, O. G., Angew. Chem. Int. Ed. 2011, 50, 10875.
(40) Ibele, M.; Mallouk, T. E.; Sen, A., Angew. Chem. Int. Ed.
2009, 48, 3308.
(41) Moo, J. G. S.; Presolski, S.; Pumera, M., ACS Nano 2016, 10,
3543.
(42) Calvo-Marzal, P.; Manesh, K. M.; Kagan, D.;
Balasubramanian, S.; Cardona, M.; Flechsig, G.-U.; Posner, J.;
Wang, J., Chem. Commun. 2009, 4509.
(43) Wang, W.; Duan, W.; Zhang, Z.; Sun, M.; Sen, A.; Mallouk,
T. E., Chem. Commun. 2015, 51, 1020.
(44) Chien, C. L.; Fan, D.; Cammarata, R. C. System and
method for precision transport, positioning, and assembling of
longitudinal nano-structures. Patent: 9, 044, 808, 2015 and 9, 718,
683, 2017.
(45) Whitney, T. M.; Jiang, J. S.; Searson, P. C.; Chien, C. L.,
(52) Lee, T.-C.; Alarcón-Correa, M.; Miksch, C.; Hahn, K.;
Science 1993, 261, 1316.
Gibbs, J. G.; Fischer, P., Nano Lett. 2014, 14, 2407.
(46) Jones, T. B., Electromechanics of Particles. Cambridge
(53) Baker, D. R., Capillary Electrophoresis. John Wiley & Sons,
University Press: New York, 2005.
Inc.: 1995.
(47) Kim, K.; Xu, X.; Guo, J.; Fan, D. L., Nat. Commun. 2014, 5,
(54) Barz, D. P. J.; Vogel, M. J.; Steen, P. H., Langmuir 2009, 25,
3632.
1842.
(48) Berg, H. C., Random Walks in Biology. Pricention niversity
(55) Guo, J.; Kim, K.; Lei, K. W.; Fan, D. L., Nanoscale 2015, 7,
Press: 1993.
11363.
(49) Hydrogen Peroxide Physical Properties Data Book. Becco
Chemical Division: 1955.
(56) Kim, K.; Guo, J.; Xu, X.; Fan, D., ACS Nano 2015, 9, 548.
(50) Liu, R.; Sen, A., J. Am. Chem. Soc. 2011, 133, 20064.
(51) Wang, H.; Moo, J. G. S.; Pumera, M., ACS Nano 2016, 10,
5041.
Table of Contents (TOC)
9
|
1812.05286 | 1 | 1812 | 2018-12-13T06:58:31 | Room temperature infrared photodetectors with hybrid structure based on 2D materials | [
"physics.app-ph"
] | Two-dimensional (2D) materials, such as graphene, transition metal dichalcogenides (TMDs), black phosphorus (BP) and related derivatives, have attracted great attention due to their advantages of flexibility, strong light-matter interaction, broadband absorption and high carrier mobility, and have become a powerful contender for next-generation infrared photodetectors. However, since the thickness of two-dimensional materials is on the order of nanometers, the absorption of two-dimensional materials is very weak, which limits the detection performance of 2D materials-based infrared photodetector. In order to solve this problem, scientific researchers have tried to use optimized device structures to combine with two-dimensional materials for improving the performance of infrared photodetector. In this review, we review the progress of room temperature infrared photodetectors with hybrid structure based on 2D materials in recent years, focusing mainly on 2D-nD (n = 0, 1, 2) heterostructures, the integration between 2D materials and on-chip or plasmonic structure. Finally, we summarize the current challenges and point out the future development direction. | physics.app-ph | physics | Room temperature infrared photodetectors with
hybrid structure based on 2D materials
Tiande Liu(刘天德)1,†, Lei Tong(童磊)1,†, Xinyu Huang(黄鑫宇) 1, Lei Ye(叶镭)1,*
1 School of Optical and Electronic Information, Huazhong University of Science and
Technology, Wuhan, Hubei 430074, China
† These author contributed equally
* Corresponding authors
Lei Ye: [email protected]
1
Abstract: Two-dimensional (2D) materials, such as graphene, transition metal
dichalcogenides (TMDs), black phosphorus (BP) and related derivatives, have attracted
great attention due to their advantages of flexibility, strong light-matter interaction,
broadband absorption and high carrier mobility, and have become a powerful contender
for next-generation infrared photodetectors. However, since the thickness of two-
dimensional materials is on the order of nanometers, the absorption of two-dimensional
materials is very weak, which limits the detection performance of 2D materials-based
infrared photodetector. In order to solve this problem, scientific researchers have tried
to use optimized device structures to combine with two-dimensional materials for
improving the performance of infrared photodetector. In this review, we review the
progress of room temperature infrared photodetectors with hybrid structure based on
2D materials in recent years, focusing mainly on 2D-nD (n = 0, 1, 2) heterostructures,
the integration between 2D materials and on-chip or plasmonic structure. Finally, we
summarize the current challenges and point out the future development direction.
Keywords: two dimensional materials; heterostructure; infrared photodetectors;
PACS: 68.65.-k, 74.78.Fk, 85.60.Gz
1. Introduction
Infrared (IR) photodetectors can convert infrared light signals into easily detectable
electrical signals, which are widely used in imaging, communications, medical and
night vision applications at present.[1, 2] However, in order to obtain better detecting
performance, the traditional infrared photodetectors based on the III-V or II-VI
2
semiconductors have to work at low temperatures to reduce dark currents.[3] Because
of the necessity of large refrigeration equipment to realize low operating temperature,
it leads to the greatly increase of the manufacturing cost and limitation of the
application, making it difficult to achieve miniaturization, flexibility and portability.
In order to circumvent this disadvantage, new type semiconductors are proposed
to use in the application of IR photodetectors, which can achieve low current at room
temperature and high photocurrent. In recent years, two-dimensional materials with an
atomic layer thickness, are verified to overcome the shortcomings induced by
traditional semiconductors used in traditional IR detection due to their excellent
electrical, optical and mechanical properties.[4-13] For example, graphene is the most
studied two-dimensional materials with semi-metal, high mobility, broad spectrum
response, flexibility, and tunable Fermi level properties,[14, 15] suggesting that graphene
can be fabricated into flexible infrared photodetectors with broad spectrum absorption,
ultra-fast response.[16, 17] In addition, transition metal dichalcogenides materials MX2
(M: transition metal atoms, X=S, Se, Te), which have the properties of strong light
interactions and tunable bandgap varying with the number of layers, are also suitable
as photosensitive materials for infrared photodetectors. Black phosphorus shows a
direct bandgap property with the range of 0.3-1.3 eV, leading to its absorption spectrum
covering the near-infrared, mid-infrared and far-infrared regions, even extending to the
terahertz region.[18] Especially, wide spectral response range coupled with its unique
anisotropic crystal structure, promises the potential of black phosphorus for infrared
polarized photodetectors.[19] Although 2D materials have various excellent and unique
3
properties for the applications of photodetector, they cannot avoid such a problem that
such thin thickness cannot absorb more light. For example, the thickness of monolayer
graphene is 0.35 nm and absorbs only 2.3% of light.[20] Although the bandgap of
monolayer TMDs is a direct bandgap, the whole absorption is also less than 10%.[21]
In order to solve this problem, it is necessary to combine two-dimensional materials
with other strong light absorbing materials, or to integrate with some structures that can
enhance light absorption. For example, the band gap of quantum dots and nanoparticles
with strong light absorption can be tuned by changing its size,[22, 23] which is very
suitable as a photosensitive layer for infrared photodetectors. Carbon nanotubes have
excellent broadband absorption characteristics,[24] which is also very suitable for
forming heterostructures with two-dimensional materials. Integrating two-dimensional
materials with waveguides or cavities can enhance the photon absorption of 2D
materials to improve detection performance. In addition, this integration also reveals
the compatibility of
two-dimensional materials with existing optoelectronic
technologies. The surface plasmon resonance effect not only enhances light absorption
but also selectively absorbs light by designing different shapes and sizes,[25] therefore
the spectral response of these designed photodetectors can be extended to infrared or
terahertz region that cannot be absorbed by intrinsic two-dimensional material, due to
the surface plasmon resonance effect.
In this review, we review the development of infrared photodetectors with mixed-
dimensional structure between two-dimensional materials and other low-dimensional
materials, focusing on quantum dots, single-wall carbon nanotubes, and 2D vertical or
4
lateral heterostructures composed of different two-dimensional materials. Moreover,
we also discuss the integration of two-dimensional materials and on-chip structures,
including waveguides and cavities. We also introduce the application of surface
plasmon resonance in infrared photodetectors based on two-dimensional materials.
Finally, we summarize the faced challenges currently in infrared photodetectors with
hybrid structures based on two-dimensional materials and point out the future directions
for further development.
2. Heterostuctures
2.1 Quantum dots on 2D materials
In order to meet the demand for high-speed response of infrared photodetectors,
the materials used to fabricate the detectors is often required to have high carrier
mobility to reduce the transit time of photogenerated carriers in conducting channel.
The carrier mobility of graphene is as high as 60,000 cm2V-1s-1,[26] which can fully meet
this demand. However, the optical response of photodetectors based on intrinsic
graphene is only a few tens of mA/W,[27, 28] which is caused by weak light absorption
due to its atomic-scale thickness. Compared with graphene, monolayer molybdenum
disulfide (MoS2) has a direct band gap, so the absorption is stronger than that of
graphene, but still not enough. Moreover, the band gap of molybdenum disulfide is 1.8
eV, and the corresponding absorption wavelength range is in the visible light range,
leading to the limitation of directly fabrication for an infrared photodetector.[21, 29] The
problems of weak light absorption caused by the ultra-thin thickness and the
5
unsatisfactory photoresponse range due to broad band gap, have seriously hindered the
development of infrared photodetectors based on two-dimensional materials. To solve
this problem, scientific researchers have tried many methods. It is a good attempt to
combine two-dimensional materials with quantum dots.[16, 30-39] Because quantum dots
have stronger light absorption relative to 2D materials, and their absorption wavelength
range can be tuned by changing its size,[22, 40] and it can efficiently and selectively
absorb photons and convert them into photogenerated electron-hole pairs. Therefore,
the hybrid 2D-QDs structure is promising in the field of infrared detection technology.
Konstantatos et al. demonstrated a hybrid graphene-PbS quantum dots
phototransistor with ultrahigh gain of 108 and responsivity of 5107 AW-1.[32] The
schematic diagram of a hybrid graphene-PbS quantum dots phototransistor is shown in
Figure 1a. Monolayer or bilayer graphene was exfoliated with tape on a Si/SiO2
substrate, and then an 80-nm PbS quantum dots film was prepared by low-cost spin
casting method. PbS QDs thin film acts as light-sensitive material due to the
characteristics of strong and selective light absorption, while graphene acts as a
conduction channel owing to its extremely high carrier mobility. The energy band
diagram at the interface of graphene and PbS QDs is shown in Figure 1b. On account
of the work function mismatch between graphene and PbS QDs, the band bending
occurs at the interface. PbS QDs thin film absorbs light to create photogenerated
electron-hole pairs, which are then separated by an built-in electric field at the interface
that induced by band bending. The Dirac point (VD) will be drifted to higher backgate
voltage due to illumination. For VBG<VD, holes are transferred from PbS QDs to
6
graphene channel, and electrons are still trapped in the PbS QDs. The resistance of
graphene decreases and carrier transport in the channel is hole-dominated. For VBG>VD,
holes are transferred to graphene and recombined with electrons induced by the
backgate, so the resistance of graphene increases and carrier transport in the channel is
electron-dominated.
After the photogenerated holes are transferred to the graphene channel, they are
driven by the source-drain voltage (VSD) to reach the drain. The linear dependence of
the photocurrent on VSD is shown in Figure 1c. In order to achieve charge conservation
in graphene, holes are continuously replenished from the source. Due to the long charge
trapping time of the PbS QDs and the high mobility of the graphene,[26, 41] holes in the
channel can circulate multiple times during quantum dots trapped photogenerated
electrons, resulting in ultrahigh photoconductive gain. The relationship between the
responsivity and the applied back-gate voltage is shown in Figure 1d, which shows that
the photoconductive gain of the photodetector can be adjusted by VBG. For VBG<VD,
the measured responsivity can reach up to 5107 A/W at VBG=-20 V; when VBG=4 V,
the responsivity reduces to zero. This back-gate-tunable gain is significant so that the
required gain can be adjusted based on the photosignal intensity that needs to be
detected.
Similarly, Zheng et al. demonstrated an ambipolar broadband photodetector based
on N, S codecorated graphene-PbS quantum dot heterostructure.[42] N, S codecorated
graphene greatly reduces the p-type doping of graphene in air. The gate voltage required
to modulate the Dirac point and the carrier type is significantly reduced, thereby
7
achieving gate tuneable mutual transformation of positive and negative photoresponse.
The device exhibits excellent detection performance with responsivity of 104 A/W and
specific detectivity of 1012 Jones at 1550 nm.
Figure 1
(a) The schematic diagram of a hybrid graphene-PbS quantum dots
phototransistor. (b) The energy band diagram at the graphene/PbS QDs interface. (c)
The linear dependence of the photocurrent on the applied drain-source voltage for
different optical powers (46, 336, 516 pW). VBG=0. Inset: The total current (I=Idark +
Iphoto) as a function of the applied drain-source voltage for different optical powers. (d)
The responsivity as a function of the applied back-gate voltage VBG. VSD=5 V. (a-d) are
reproduced from Ref.[32] with permission from the Nature Publishing Group.
Another important two-dimensional material, transition metal dichalcogenides
8
(TMDs), has also been extensively studied in recent years.[43-46] However, due to the
bandgap limitation and ultrathin thickness, the response spectral range and responsivity
of TMDs-based infrared photodetectors are limited. Similarly, in order to enhance light
absorption and broaden the response spectral range, a strategy for integrating TMDs
materials with quantum dots has also been exploited.[31, 35, 36] Huo et al. demonstrated
MoS2-HgTe QDs hybrid photodetectors beyond 2m.[31] The infrared detection
capability of this detector is mainly caused by HgTe QDs thin film, which acts as a
sensitizing layer, and the response spectrum range can be further extended to the mid-
wave infrared and long-wave infrared ranges by adjusting the size of HgTe quantum
dots.[47, 48] A TiO2 buffer layer between the MoS2 transistor channel and the HgTe QDs
sensitizing layer was used as a protective layer of the MoS2 channel and as an n-type
electron acceptor medium to form an efficient p -- n junction with the HgTe QDs at the
interface facilitating the charge transfer to MoS2 channel.[31, 49] Thus, the dark current
can be adjusted by the gate voltage and the specific detectivity D* (~1012 Jones) at a
wavelength of 2 m can be obtained. The detector can operate at room temperature and
exhibit sub-milliseconds response and high responsivity of ~106 A/W, which
demonstrates that the hybrid 2D-QDs structure is an effective method to improve the
performance of photodetector based on 2D materials.
In order to obtain high gain to make up for the weak light absorption of 2D
materials, many infrared photodetectors based on hybrid 2D-QDs heterostructures are
phototransistors that have an electronic passive sensitization layer composed of strong
light absorbing material.[32, 50, 51] Thus, the external quantum efficiency (EQE), response
9
speed and linear dynamic range are limited by thin sensitizing layers thickness, long
trapping times, and high density of the sensitizing centers trap-state, respectively.[32, 41,
52] In order to solve these problems, Nikitskiy et al. proposed a photodetector that
integrates an electrically active colloidal QD photodiode with a graphene transistor.[34]
The device is composed of a bottom graphene channel, a 300-nm-thick PbS QDs layer
and a top ITO electrode. The photodiode operation and phototransistor operation of the
hybrid 2D-QDs photodetector are shown in Figure 2a and b, respectively.
Photogenerated electron-hole pairs are created at the QDs layer, and then drifted by the
built-in electric field at the interface and by the applied bias voltage to form
photocurrent directly in photodiode operation. However, after the electron-hole pairs
are separated, electrons remain in the QDs layer and holes are driven by depletion
region on the graphene/PbS QD interface and transferred to graphene in phototransistor
operation. For VTD>0, the width of the depletion region expands, increasing the
efficient charge collection area, and thus the external quantum efficiency also increases.
The graphene-QDs photodiode responsivity and EQE are shown in Figure 2c. With
increasing VTD, EQE increased from 10% at VTD=0 V to 75% at VTD=1.2 V (limited by
reflection), indicating the importance of VTD for improving EQE. Figure 2d shows the
EQE and responsivity in phototransistor operation, agreeing with the dependence of
EQETD on VTD in photodiode operation, which demonstrates that the EQE of this
integrated configuration is determined by the photodiode. Moreover, the responsivity
is 5106 V/W at VTD=1.2 V, indicating that the phototransistor still has a high gain of
105 compared with the photodiode operation. With increasing VTD, the temporal
10
response accelerates and the linear dynamic range extends. The electrical 3 dB
bandwidth of the detector reaches 1.5 kHz for the optimum VTD, and the linear dynamic
range extends from 75 dB at VTD=0 V to 110 dB at VTD=1.2 V. The above results prove
that the design and optimization of infrared photodetectors based on hybrid 2D-QDs
heterostructures can effectively enhance the detection performance.
Figure 2
(a-b) Schematic of photodiode operation and phototransistor operation of
the hybrid graphene-PbS QDs photodetector, respectively. (c) Responsivity and
external quantum efficiency EQE of photodetector in photodiode operation as a
function of VTD. Inset: energy band diagrams of device with and without VTD bias. (d)
Responsivity and EQE of photodetector in phototransistor operation as a function of
VTD. (e) Normalized photoresponse versus light modulation frequency. The specified
VTD values are shown in the legend. Inset: The linear dependence of extracted 3 dB
bandwidth values on applied VTD. (f) Responsivity of the detector as a function of
incident irradiance in the range of 10-5 to 10 Wm-2, which shows a significant extension
11
of the linear dynamic range with increasing VTD. (a-f) are reproduced from Ref.[34]
with permission from the Nature Publishing Group.
2.2 Nanowires combined with 2D materials
One-dimensional nanomaterials, such as single-wall carbon nanotubes (SWNTs),
have unique and excellent optical and electrical properties due to 1D quantum
confinement effect.[53] Because of
the
type of
tube chiralities, metallic or
semiconducting properties are exhibited, and the carrier mobility of semiconducting
SWNTs can reach up to 105 cm2V-1s-1 at room temperature.[54-56] SWNTs have
broadband and tunable absorption, so it has great potential in the field of infrared
detection. However, the photoresponsivity is limited by the small light-absorbing area,
and the responsivity of single-tube photovoltaic devices is always <10-3 A/W.[57] Paulus
et al. reported that the junction between a metallic SWNT and graphene can perform
effective charge transfer.[58] This is because that the SWNT is able to dope the graphene
in a spatially controlled way without affecting its sp2 lattice structure.[58] Liu et al.
demonstrated a high-performance photodetector based on a planar atomically thin
SWNT-graphene hybrid film across visible to near-infrared range (400-1,550 nm).[50]
Schematic of the photodetector and atomic force microscope (AFM) characterization
of SWNT-graphene hybrid film are shown in Figure 3a. SWNTs have a diameter in the
range of 1.0-1.6 nm, and the S11 and S22 bands are found to be located at ~1,800 and
~1,000 nm with higher SWNTs loadings. The photogenerated electron-hole pairs are
generated at the SWNT/ graphene interface under optical illumination, and then
12
separated by the built-in electric field. Electrons are transferred from SWNTs to
graphene, while holes are trapped in the SWNTs, forming a photogating effect. Figure
3b shows that the source-drain current (ISD) as a function of the applied back-gate
voltage (VG). The Dirac point shifts toward the lower VG as the light increases,
indicating that the graphene resistance can be effectively modulated by the photogating
effect. Unlike other photodetectors that use the photogating effect to increase the
responsivity, which generally has slow response speed, the photodetector based on
hybrid SWNTs-graphene exhibits a fast response time of ~100 s (Figure 3c). This is
mainly due to the high mobility of graphene and SWNTs and fast charge transfer at the
SWNTs-graphene interface, and the field-effect mobility of electrons and holes are
measured as 3,920 and 3,663 cm2V-1s-1, respectively. The transit time of the hybrid
SWNTs-graphene photodetector is estimated to be ~ 10-9s, so the photoconductive gain
is ~ 105 and the gain-bandwidth product is ~109 Hz. Moreover, since SWNTs act as a
sensitizing layer, the detector also exhibits the characteristics of a broadband
photoresponse. The relationship between the responsivity and the incident optical
power at different illumination wavelengths in the range of visible to near-infrared
range (405, 532, 650, 980 and 1,550 nm) is shown in Figure 3d, which shows a high
responsivity of >100 A/W. Further, Liu et al. fabricated a photodetector based on
graphene/carbon nanotube hybrid films on a PET substrate and measured the
performance under bending conditions.[59] It was found that this device has good
robustness against repetitive bending, which fully demonstrated the huge application
potential of this heterostructure in the field of flexible electronics.
13
Jariwala et al. demonstrated a gate-tunable carbon nanotube-MoS2 heterjunction p-
n diode.[60] Since both monolayer MoS2 and semiconducting SWNTs have direct
bandgaps and tightly bound excitation states in absorption spectra,[61] the photodiode
based on hybrid MoS2-SWNTs also exhibits excellent detection performance. The
external quantum efficiency (EQE) and the response time are ~25% and <15 s,
respectively. Figure 3e shows that photocurrent spectral response can be tuned by
changing the gate voltage. The photocurrent in the near-infrared region continuously
decreases as the gate voltage decreases. Furthermore, the performance of devices based
on 1D-2D heterostructures can be further optimized by strain modulation. For example,
the responsivity of the photodetector based on the graphene-ZnO nanowire
heterostructure is improved by 26% under 0.44% tensile strain on the ZnO nanowire.[62]
The above results show that it is significant to carry out the research of the hybrid 1D-
2D heterostructure in the field of infrared detection.
14
Figure 3 (a) Schematic of the photodetector and AFM characterization of SWNT --
graphene hybrid film (middle: scale bar, 200 nm; right: scale bar, 20 nm). (b) The ISD-
VG transfer curve of device with increasing 650 nm illumination powers. VSD=0.5 V.
Inset: the optical image of the SWNT-graphene device (Scale bar, 10m). (c) Temporal
response of the SWNT -- graphene hybrid photodetector under 650 nm illumination. (d)
The dependence of responsivities of photodetector on the optical power at different
illumination wavelengths from visible to near-infrared range (405, 532, 650, 980 and
1,550 nm). (e) Photocurrent spectral response can be tuned by varying the gate voltage.
VD=-10 V. (a-d) are reproduced from Ref.[50] with permission from the Nature
15
Publishing Group. (e) are reproduced from Ref.[60] with permission from the National
Academy of Sciences.
2.3 2D vertical and lateral heterostructures
2D vertical heterostructures
Two-dimensional vertical heterostructure is fabricated by stacking different two-
dimensional materials in the vertical direction. This heterostructure is generally
prepared by targeted transfer method or CVD growth technique.[63-68] 2D vertical
heterostructures have many advantages, such as atomically sharp interface, no atomic
diffusion between layers, and no strict lattice matching problems due to weak van der
Waals interaction between 2D materials. Moreover, the strong interlayer coupling and
ultrafast transfer of charge carriers make it have great potential in the field of
optoelectronics and electronic devices.[68-73]
In order to achieve infrared detecting, photodetectors must have a broad spectral
response. However, the development of TMDs for infrared detection is hindered due to
the lack of 2D materials with narrow bandgap. Monolayer -MoTe2 has a band gap of
1.1 eV and can absorb light in the near-infrared range.[74, 75] The response spectral of
the photodetectors based on the -MoTe2/MoS2 dichalcogenide heterojunction covers
in the range from the visible to near-infrared (800 nm).[76] However, this is not enough
for mid-infrared and far-infrared detection, and the response range of the photodetector
should be further extended. Graphene can absorb light indiscriminately due to the
gapless band structure and monolayer black phosphorene shows a direct optical band
16
gap of 1.3 eV,[77, 78] so they are effective absorption material in the infrared region.
Long et al. demonstrated a broadband photodetector based on MoS2-graphene-
WSe2 heterostructure covering visible to short-wavelength infrared range at room
temperature.[79] The schematic diagram and optical image of the photodetector based
on a MoS2-graphene-WSe2 heterostructure are shown in Figure 4a. The p-doped WSe2
is induced by large work-function metal Pd, whereas MoS2 flakes remain n-type due to
Fermi level pinning,[80] and thus forming a p-n junction with a strong internal electric
field of ~2108 V/m. Photogenerated electron-hole pairs can be effectively separated in
the presence of the internal electric field. Therefore, the dark current of the device is
effectively suppressed and the specific detectivity increases accordingly. Figure 4b
shows the responsivity R and specific detectivity D* of the photodetector for
wavelengths ranging from 400 to 2,400 nm with an interval of 100 nm. R and D* are
as high as 104 A/W and 1015 Jones in the visible region, while they decrease sharply to
100 mA/W and 109 Jones in the infrared wavelength of 2400 nm, respectively. Figure
4c explains the reason for the sharp changes in responsivity and specific detectivity.
The band gap of MoS2 (Eg1) and WSe2 (Eg2) are 1.88 eV and 1.65 eV, respectively, and
the corresponding wavelengths are 660 nm and 750 nm.[21, 29, 81] When the incident
photon energy h exceeds the bandgap of the material, the material can absorb photons
to generate electron-hole pairs. Therefore, for h >Eg1>Eg2, MoS2, graphene and WSe2
all can generate photoelectron electron-hole pairs, resulting in a higher responsivity.
For h <Eg1<Eg2, only graphene can generate photoelectron electron-hole pairs, and
thus the photocurrent is drastically reduced. Qiao et al. reported a broadband
17
photodetectors based on graphene-Bi2Te3 heterostructure (Figure 4d).[82] Bi2Te3 has
only 2.7% lattice mismatch with graphene, so the graphene-Bi2Te3 heterostructure can
be directly prepared on graphene by Van der Waals epitaxial growth method.[83] Figure
4e plots photocurrent of devices based on pure graphene and graphene-Bi2Te3
heterostructure as a function of source-drain voltage without applied gate bias, which
shows that the latter is ten times higher than the former. Both graphene and Bi2Te3
absorb light to generate photoelectron-hole pairs, which are subsequently separated by
a built-in electric field. Then, photogenerated electrons in graphene are transferred into
Bi2Te3, while the photogenerated holes in Bi2Te3 are transferred into graphene, which
suppresses the recombination of carriers and generates a larger photocurrent. Figure 4f
shows the dependence of photocurrent on wavelengths ranging from 400 nm to 1,550
nm. Due to the gapless nature of graphene and a small bandgap (0.15-0.3 eV) of Bi-
2Te3,[84] both graphene and Bi2Te3 are effective absorbent materials for infrared light,
and thus the detector exhibits broadband detection characteristics. The effective
detection range can be extended to the near infrared (980 nm) and c-communication
band (1,550nm). The responsivity is 35 A/W, 10 A/W and 0.22 A/W in the visible (532
nm), infrared (980 nm) and c-communication band (1,550 nm) regions, respectively.
18
Figure 4
(a) Top: optical
image of
the photodetector based on a
MoS2−graphene−WSe2 heterostructure. Scale bar, 5 m. Bottom: Schematic diagram
of the device. (b) Responsivity and specific detectivity of the photodetector for
wavelengths ranging from 400 to 2,400 nm with an interval of 100 nm. (c) Top: Spectra
from ultraviolet to infrared. The band gap of MoS2 (Eg1) and WSe2 (Eg2) as well as
corresponding wavelengths are plotted on the graph, respectively. Bottom left: For
h >Eg1>Eg2, MoS2, graphene and WSe2 all can generate photogenerated electron-hole
pairs. Bottom right: For h <Eg1<Eg2, only graphene can generate photogenerated
electron-hole pairs. (d) The schematic diagram of the device based on graphene-Bi2Te3
heterostructure. (e) The dependence of photocurrent of devices based on pure graphene
and graphene-Bi2Te3 heterostructure on source-drain voltage VSD. VG=0 V. (f)
Dependence of photocurrent on wavelengths ranging from 400 nm to 1,550 nm. (a-c)
are reproduced from Ref.[79] with permission from the American Chemical Society.
19
(d-f) are reproduced from Ref.[82] with permission from the American Chemical
Society.
So far, highly polarization sensitive infrared photodetector based on two-
dimensional materials working at room temperature are still lacking. Ye et al. first
demonstrated a broadband photodetector based on a vertical photogate heterostructure
of BP-on-WSe2 with highly polarization sensitive for infrared detection (Figure 5a).[85]
WSe2 acts as a conductive channel, while BP with a small direct bandgap and highly
anisotropic crystal structure acts as a photogate.[18, 19, 77] Both WSe2 and BP absorb light
efficiently and generate photogenerated electron-hole pairs under visible illumination,
which are then separated by the built-in electric field. The photo-generated electrons in
BP are transferred to WSe2 and the photo-generated holes in WSe2 are transferred to
BP, which effectively reduces the recombination efficiency. An increasing number of
photo-generated electrons in WSe2 are driven by source-drain bias to form photocurrent,
thereby enhancing the responsivity. However, due to the band gap limitation of WSe2,
only BP with a small band gap is an effective absorption material to generate photo-
generated free carriers under infrared illumination. So the photoresponse in the infrared
region will be smaller than in the visible region. The responsivity is ~103 A/W under
visible illumination (637 nm), while the responsivity is ~0.5 A/W under infrared
illumination (1,550 nm). Figure 5b shows the photoconductive gain G and detectivity
D* of the photodetector for wavelengths ranging from 400 to 1,600 nm. The detectivity
in the visible and infrared regions are 1014 and 1010 Jones, and the gain in the visible
20
and infrared regions are 106 and 102, respectively, demonstrating excellent broadband
detection performance. The optical image of the BP-on-WSe2 photodetector and the
relationship between the photocurrent image of the photodetector at 1,550 nm and the
polarization angle are shown in Figure 5c. The advantage of this device structure is that
it can eliminate the disturbance of two-fold polarization-dependent photocurrent
resulted from the geometric edge effect at the metal-BP edge and fully collect the photo-
generated free carriers
isotropically for highly polarization-sensitive
infrared
detection.[85] As can be seen from the photocurrent image, the polarization
photoresponse originates from the overlap of BP and WSe2, and it changes obviously
with the polarization angle changes. The polar plot in Figure 5d shows the relationship
between the photocurrent and the polarization angle of the incident light. The maximum
photocurrent, corresponding responsivity of ~40 mA/W, is taken along the horizontal
axis (defined as 0o polarization), while the minimum photocurrent, corresponding
responsivity of ~6.8 mA/W, is taken along the vertical axis (defined as 90o
polarization).[85] The ratio of the maximum responsivity to the minimum responsivity
is about 6, showing high polarization sensitive infrared detection performance.
Similarly, the mid-wave infrared photodetector based on BP and MoS2 heterostructure
designed by Bullock et al. also utilized the anisotropy of BP to achieve polarization
detection, which is bias-selectable and operates without the need for external optics.[86]
The device exhibits high EQE (35%) and D* (1.11010 cm Hz1/2 W-1) at room
temperature.
The above results show that the 2D vertical heterostructure can effectively extend
21
the response range of the photodetector to achieve infrared detection. In addition, CVD
epitaxial growth technology provides a guarantee for the scaled preparation of vertical
heterostructures. By combining 2D materials with different unique and excellent
physical properties, a multifunctional infrared detector with excellent detection
performance can be prepared.
Figure 5
(a) Schematic diagram of the photodetector with BP-on-WSe2 photogate
structure. (b) The photoconductive gain G and detectivity D* of the photodetector for
wavelengths ranging from 400 to 1,600 nm at 1 mW/cm2 incident light power density.
22
Vds=0.5 V. (c) Top: the photocurrent image of the photodetector under 1,550 nm
illumination with different light polarizations (red arrows). Bottom left: the optical
image of the photodetector with BP-on-WSe2 photogate structure. (d) The photocurrent
as a function of light polarization at 100 mW/cm2 incident light power density. Vds=0.5
V. (a-d) are reproduced from Ref.[85] with permission from Elsevier.
2D lateral heterostructures
The heterostructures mentioned above are basically vertical, and the built-in
electric field at the interface promotes the separation of photogenerated electron-hole
pairs and charge transfer, thereby reducing the recombination and accelerating the
response speed. In addition, two-dimensional lateral heterostructures, a heterostructure
composed of materials with different work functions or homojunction composed of the
same material but different doping types in the lateral direction, form a built-in electric
field at the interface and play a similar role. At present, the methods for forming lateral
heterostructures generally include chemical doping,[87, 88] electrostatic doping,[89-91] and
one-pot or two-step chemical vapor deposition (CVD) growth methods.[64, 65, 92-96]
Recently, Yu et al. demonstrated a lateral black phosphorene P -- N junction formed
via chemical doping for high performance near-infrared photodetector.[88] The n-type
doping of few-layer black phosphorene is achieved by using benzyl viologen (BV) as
an electron dopant, which can obtain a high electron concentration while still maintain
the high carrier mobility of BP as the processes do not induce defects in the crystal
lattices, and Al2O3 is used to selectively protect part of the area of a p-type BP at the
23
same time.[88] The electron doping concentration of BP can be adjusted by doping time,
and the built-in electric field at the P-N junction can effectively separate photogenerated
electron-hole pairs and reduce recombination efficiency. Photocurrent tests show that
the off-current under 1.47 m illumination is 4 orders of magnitude higher than in dark
condition, demonstrating that the photocurrent dominates over thermionic and
tunneling currents in the photodiode operation.[88] The photodetector exhibits high
responsivity of ~180 mA/W, fast response of tens of milliseconds and high detectivity
in the range of 1011-1013 Jones under infrared light illumination (=1.47 m).
However, the current methods for preparing lateral heterojunctions have their own
limitations. Chemical doping has the characteristic that it is difficult to find suitable
dopants. The density of the carrier modulated by the electrostatic gating is relatively
low, so the modulation effect is limited.[19, 97] CVD technology is an effective method
for large-scale preparation of lateral heterostructures, but the main focus is on epitaxial
growth of TMDs with large band gap at present, and thus CVD technology has
limitation in the field of infrared detection. The infrared light detection technology
based on 2D lateral heterostructure needs further research and development.
3. On-chip structure
3.1 2D materials on waveguide
24
Two-dimensional materials show strong interaction with light, but the spectral
response range and responsivity of the photodetector based on 2D materials are limited
due to the reason of the atomically ultra-thin thickness and the limitation of the band
gap. In order to obtain better infrared detection performance to achieve practical
applications, it is necessary to enhance the infrared light absorption of two-dimensional
materials. Integrating two-dimensional materials with waveguides is a good strategy to
enhance light absorption.[91, 98-104] The propagation of light in the waveguide is
accompanied by the evanescent field, which can be absorbed by the two-dimensional
materials to generate photogenerated electron-hole pairs.[105-108] Different with the
incident light perpendicularly incident on the two-dimensional materials, the in-plane
absorption of two-dimensional materials makes its nanoscale thickness to be no longer
a major limiting factor for light absorption. The length of the two-dimensional material
in contact with the waveguide is on the order of micrometers, so the length of the
interaction with the light is also on the order of micrometers and therefore can
effectively enhance light absorption.
Recently, a series of similar but independent waveguide integrated graphene
photodetectors have been reported,[98, 101, 102] and the device structures are shown in
Figures 6a, c and e, respectively. Figure 6a shows a scanning electron microscope (SEM)
image of a waveguide-integrated graphene photodetector using a GND-S-GND
configuration. It should be noted that in addition to the length L of graphene, the width
W of central electrode S in this device also affects the light absorption of the graphene.
The absorption coefficient of metal electrode M increases with increasing electrode
25
width W. For W>100 nm (W>160 nm), the light absorption of the metal electrode will
exceed the absorption of monolayer (bilayer) graphene. However, the contact resistance
between metal and graphene also increases with decreasing width of electrode, so the
metal electrode absorption coefficient and contact resistance should be considered
comprehensively to obtain the best device performance. Figure 6b shows the absorption
of graphene as a function of graphene length L at different widths, indicating that as
long as the appropriate W and L values are obtained, the light absorption of graphene
can be greatly enhanced. For example, when W = 100 nm, more than 50% light
absorption can be obtained with only 22 m length bilayer graphene. Figure 6d shows
an SEM image of a waveguide-integrated graphene photodetector with high
responsivity and photocurrent image at zero bias. A 53 m bilayer graphene was
overlaid on the waveguide, and two electrodes with waveguide distances of 100 nm and
3.5 m, respectively, were used to collect the photogenerated carriers. The lateral metal-
doped junction overlapping the waveguide is formed by doping of metal electrode 100
nm away from the waveguide, and effectively promotes the separation of
photogenerated electron-hole pairs. The photocurrent image shows that the built-in
electric field between metal-doped graphene and graphene appears at the metal and
graphene interface and leads to two narrow photocurrent regions. Moreover, the region
of maximum photocurrent coincides with the position of waveguide. Wavelength
transmission measurements show that the transmission loss are 0.1 dB and 4.8 dB
before and after the graphene transfer, respectively, indicating that the optical
absorption of
the waveguide-integrated graphene
is greatly
increased. The
26
photodetectors demonstrated an ultrafast response rates exceeding 20 GHz at zero bias
and a high response rate of 0.108 A/W at 1 V bias. Thanks to the gapless graphene, the
spectral response of photodetector ranged from 1,450 nm to 1,590 nm, which can
effectively detect the c-communication band of the near-infrared region. The response
spectrum of the waveguide-integrated graphene photodetector can be further extended
to the mid-infrared region. Xu and coworkers demonstrated a graphene/silicon
heterostructure waveguide photodetector with a responsivity of up to 0.13 A/W at room
temperature in the mid-IR region (2.75 m).[102] In addition, this photodetector exhibits
a low dark current due to the presence of a graphene/silicon heterojunction potential
barrier, and the on/off photocurrent ratio is 30 at a bias of -1.5 V for 2.75 m
illumination at room temperature (Figure 6f). This work further explains the
photocurrent generation mechanism of waveguide-integrated graphene in the visible
light region, near infrared region and mid-infrared region. For the visible light region,
since silicon has an indirect bandgap of 1.1 eV, photoresponse is mainly due to electron-
hole pairs generated after the silicon absorbs light and graphene acts as a
photogenerated free carriers conduction channel in this process. For the near-infrared
region (1.55 m), the photon energy is more than twice that of the graphene Fermi
energy level and the infrared light cannot be absorbed by silicon. The photoexcitation
of graphene takes place in a direct transition way that electrons are excited from the
valence band to the conduction band. For the mid-infrared region (2.75 m), the photon
energy is less than twice the energy difference between the Fermi level and the Dirac
point, thus preventing direct transitions of graphene.[102] The excitation of graphene
27
follows the indirect transition processes, and electrons are excited from Fermi level in
the valence band to the conduction band with a large momentum mismatch under mid-
infrared light illumination.
Figure 6
(a) SEM image of a waveguide-integrated graphene photodetector using a
GND-S-GND configuration (false colour). Inset: An enlarged view of the inside of the
black dashed line. (b) The absorption of graphene as a function of graphene length L at
different widths W. Solid line: bilayer graphene. Dashed line: monolayer graphene. (c)
28
Schematic of the chip-integrated graphene photodetector. (d) The spatially resolved
photocurrent image of the waveguide-integrated graphene photodetector at zero bias.
The black solid line shows the relative potential distribution across the graphene
channel along the white dashed line. Inset: corresponding SEM image. (e) Schematic
of a graphene/silicon heterostructure waveguide photodetector. (f) Photocurrent as a
function of applied bias voltage with 2.75 m laser illumination at different light power.
(a-b) are reproduced from Ref.[101] with permission from the Nature Publishing Group.
(c-d) are reproduced from Ref.[98] with permission from the Nature Publishing Group.
(e-f) are reproduced from Ref.[102] with permission from the Nature Publishing Group.
Similarly, Youngblood et al. demonstrated a waveguide-integrated few-layer black
phosphorus photodetector with high responsivity and low dark current.[103] Figure 7a
shows the structure of the photodetector, having a graphene top-gate, a BP channel and
the waveguide patterned on the silicon-on-insulator substrate. Since the incident light
propagates along the waveguide, the optical interaction length is equal to the width of
the black phosphorus covering on the waveguide (6.5 m). The device can absorb 78.7%
of the optical power in the waveguide, of which only 17.5% is absorbed by the black
phosphorus and the rest is absorbed by the graphene top-gate.[103] Figure 7c shows that
the photocurrent versus the gate voltage VG and the bias voltage VDS. VG can
electrostatically tune the hole doping concentration of black phosphorus and the BP is
nearly intrinsic at VG=-8 V. When black phosphorus is in a low doping state (-10 V <
VG < -1 V), the photocurrent is larger and its sign is positive relative to VDS, which
suggests photocurrent generation mechanism is photovoltaic effect at this time.[109-111]
29
For VG > 0 V, BP is tuned to be more heavily n-doped, the photocurrent is getting
smaller and its sign changes to negative relative to VDS, and the photocurrent generation
mechanism is bolometric effect at this time.[109, 112] The frequency response of the
photocurrent shown in Figure 7b can also demonstrate the photocurrent generation
mechanism at different gate voltages. The roll-off frequency is measured to be 2.8 GHz
when BP is in a low doping state (VG=-8 V), showing a fast response due to
photovoltaic effect. But the roll-off frequency is measured to be 0.2 MHz when BP is
in a high doping state (VG=8 V), showing a slower response due to bolometric effect.
Figure 7d shows the intrinsic responsivity R and internal quantum efficiency IQE as a
function of bias for different thicknesses. The intrinsic responsivity and internal
quantum efficiency of 11.5-nm-thick BP device with top-gate are 135 mA/W and 10%
at Vbias=-0.4 V, respectively. For 100-nm-thick BP device without top-gate, a larger bias
voltage can be applied because there is no need to consider the breakdown of the gate
insulation layer. Therefore, this photodetector shows better detection performance with
high intrinsic responsivity (657 mA/W) and high internal quantum efficiency (50%) at
Vbias=-0.4 V. Unlike gapless graphene, BP has a finite direct band gap,[113] which make
the dark current of waveguide-integrated BP photodetector extreme low (220 nA) at
VG=-8 V and Vbias=-0.4 V.
30
Figure 7
(a) Schematic of the waveguide-integrated few-layer BP photodetector. (b)
The frequency response of the waveguide-integrated BP photodetector at low doping
state (VG=-8 V) and high doping state (VG=8 V). (c) Two-dimensional contour plot of
photocurrent versus the gate voltage VG and the bias voltage VDS. (d) The intrinsic
responsivity R and internal quantum efficiency IQE as a function of applied bias for
different thicknesses. tBP: the thickness of BP. (a-d) are reproduced from Ref.[103] with
permission from the Nature Publishing Group.
Recently, Bie et al. reported a silicon photonic-crystal (PhC) waveguide-integrated
light source and photodetector based on a p -- n junction of bilayer MoTe2.[91] Figure 8a
shows the device schematic, and Figure 8b shows LED mode and photodetector mode
of device, respectively. The MoTe2 p-n junction is composed of p-type and n-type
31
doping that are electrostatically induced by left and right graphite top gates above the
hBN dielectric layer. The bilayer MoTe2 not only has stronger excitation and longer
emission wavelength than the monolayer MoTe2, but also has excitation caused by
direct bandgap optical transitions similar to that of the monolayer.[74, 114, 115] There is a
grating coupler for excitation and collection at the far end of the waveguide. Excitation
light is coupled into the waveguide and transmitted along the waveguide to the grating
coupler in LED mode, while incident light is coupled into the waveguide at the grating
coupler and detected by the MoTe2 p-n junction in the photodetector mode (Figure 8b).
Figure 8c shows an EL emission intensity image at room temperature cover on top of a
false-colour optical image of the device. Two extra emission spots are observed on the
grating coupler on both sides, which coincides with the aforementioned LED mode
(Figure 8b). The orange and green lines in Figure 8d represent the transmission spectra
of the PhC waveguide before and after the bilayer MoTe2 transfer, respectively. The
orange curve in Figure 8d shows that the PhC transmission peak is at 1,160 nm, but this
peak disappears after MoTe2 transfer, which is mainly due to the interference of MoTe2
absorption. The normalized EL emission spectrum collected at the grating coupler
represented by the blue line in Figure 8d, showing a narrow peak centred at 1,160 nm
on top of a broader peak centred at 1,175 nm.[91] The former coincides with the PhC
transmission peak, and the latter matches the free-space emission of the bilayer MoTe2.
The reason that the PhC transmission peak can be observed at the blue line is that the
excited states of MoTe2 have been filled by the electrically excited carriers in the LED
mode, so the absorption of MoTe2 will not cause interference. When the device is
32
operating in photodetector mode, the responsivity and external quantum efficiency of
the photodetector at the wavelength of 1,160 nm are 4.8 mA/W and 0.5%, respectively.
Figure 8e shows the temporal response in photodetector mode. The photocurrent
response bandwidth is 200 MHz (limited by experimental set-up), which is much higher
than other MoTe2 photodetectors based on photogating and photoconductive effect,[116,
117] indicating the fast response of waveguide-integrated photodetector with bilayer
MoTe2 p-n junction. As can be seen from Figure 8f, the response spectrum of
photodetector in the range from 1040 nm to 1240 nm and the largest responsivity is at
1,160 nm (the location of the PhC transmission peak), which shows the dependence of
the responsivity on the wavelength. This dependence indicates that the photocurrent
originates mainly from the light transmitted by the PhC waveguide, and thus the light
beyond this spectral response range cannot be detected due to the bandgap of the bilayer
MoTe2 and the transmission cut-off of the waveguide. This work shows that all active
optical components for point-to-point interconnects are possible with 2D TMD devices
transferred onto otherwise passive photonic integrated circuits.[91]
33
Figure 8
(a) Schematic of the silicon photonic-crystal (PhC) waveguide-integrated
light-emitting diode and photodetector based on bilayer MoTe2 p -- n junction. (b) The
schematic diagram of the device operating in LED mode and photodetector mode. The
arrow indicates the direction of light propagation in the PhC waveguide. (c) EL
emission intensity image at room temperature cover on top of a false-colour optical
image of the device. (d) The orange and green lines represent the transmission spectra
of the waveguide before and after the bilayer MoTe2 transfer, respectively. The
normalized EL emission spectrum collected at the grating coupler represented by the
blue line. (e) The temporal response of device in photodetector mode at zero bias. The
34
response bandwidth of the device is 200 MHz. (f) The response spectrum of
photodetector in the range from 1,040 nm to 1,240 nm, which shows the dependence
of the responsivity on the wavelength due to photonic-crystal waveguide. (a-f) are
reproduced from Ref.[91] with permission from the Nature Publishing Group.
3.2 2D materials on cavity
The integration of two-dimensional materials and cavities, including photonic
crystal cavities,[118, 119] optical microcavities[120, 121] and optical ring resonators,[122] is
also an effective method for enhancing the light absorption of 2D materials. The
incident light at the resonant frequency of the cavity is absorbed many times by the
atomically thin two-dimensional material in the cavity, thereby enhancing absorption.
Furchi et al. achieved a 26-fold absorption enhancement of graphene (>60%) by
monolithically integrating graphene with a Fabry-Perot microcavity.[123] The schematic
of the microcavity-integrated graphene photodetector and the corresponding electric
field distribution are shown in Figure 9a. Two Bragg mirrors are alternately composed
of quarter-wavelength thick materials with different refractive index, and graphene
layer is sandwiched between these mirrors. The presence of the barrier layer ensures
the position of resonance occur at the graphene. The incident light are reflected back
and forth multiple times between two Bragg mirrors and absorbed by the graphene,
getting a higher absorption, which agrees with the electric field distribution. The
spectral response of the device is shown in Figure 9b, and it can be clearly seen that a
significant photocurrent enhancement at the cavity resonance (855 nm). Atomic thin
monolayer graphene (only 0.335 nm) can absorb more than 60% of light, indicating
35
that the integration of 2D materials with cavity has great potential in enhancing the light
absorption of 2D materials.
However, the enhancement of light absorption by the cavity occurs only at the
resonant wavelength of the cavity. In other words, the enhancement of the cavity is at
the expense of the broadband photodetection. The strategy of integrating cavity and
two-dimensional materials cannot be used for broadband detectors, and thus it can only
be used to enhance light absorption at specific wavelengths. For example, Casalino et
al. demonstrated a vertically illuminated, resonant cavity enhanced, graphene−Si
Schottky photodetector operating at communication band (1,550 nm).[124] Figure 9c
shows the resonant cavity enhanced graphene/Si Schottky photodetector and the
resonant structure consists of a λ/2 Si slab layer confined between SLG/Si top and Au
bottom mirrors.[124] The response spectrum of the device with and without Au bottom
mirror is shown in Figure 9d. A 3-fold external responsivity Rext enhancement as
compared with no Au bottom mirror. The above result shows the integration of cavity
with two-dimensional materials can achieve stronger and selective light absorption of
2D materials, so this method is possible to be applied to specific wavelength detection.
36
Figure 9
(a) The schematic of the resonant cavity enhanced graphene photodetector
and the corresponding electric field distribution. Red arrow: the position of graphene.
(b) The dashed lines indicates the calculation result: reflection R (red), transmission T
(green) and absorption A (blue), while the solid lines indicates the measurement result:
reflection (red) and photocurrent (blue). A 26-fold absorption enhancement at the cavity
resonance wavelength (855 nm). Inset: calculated theoretical results. (c) Schematic
diagram of the resonant cavity enhanced graphene/Si Schottky photodetector. (d) The
response spectrum of the device with and without Au mirror. A 3-fold absorption
enhancement can be observed in the presence of Au mirror. The blue shift of the
resonance peak is caused by the difference between different devices. (a-b) are
reproduced from Ref.[123] with permission from the American Chemical Society. (c-
d) are reproduced from Ref.[124] with permission from the American Chemical Society.
37
4. Plasmonic structure
In order to further enhance the photoresponse of the photodetector and obtain a
tunable response spectrum, plasmonic nanostructures can also be applied to the infrared
photodetector based on
two-dimensional materials.[125-133] Metal plasmonic
nanostructures mainly enhance the response of the detector in two ways. First, the near-
field enhancement capability of the plasmonic nanostructures can focus the light field
in a small area, effectively enhancing the light absorption of the two-dimensional
material. Moreover, the plasmonic structure can generate plasma-induced hot electrons
due to plasmon decay under light illumination, and then hot electrons are injected into
the device across the Schottky barrier between the metal and the two-dimensional
material to form photocurrent.
Fang et al. demonstrated a graphene-antenna sandwich photodetector with
response spectrum in the range of visible to near-infrared and explained the reasons for
photocurrent enhancement.[131] The schematic of photodetector and optical image are
shown in Figure 10a and b. The Fano resonance plasma clusters (dimers, heptamers,
and nonamers) sandwiched between two layers of graphene are respectively located in
the 1, 3 and 5 regions, while the 2 and 4 regions have no plasmonic structure (Figure
10b). The Raman image in Figure 10b shows that the plasmonic antenna clusters of
heptamers and nonamers significantly enhance Raman intensity due to resonance under
785 nm laser illumination. The photocurrent response in different regions in Figure 10c
shows that the metal plasmonic antenna clusters can effectively enhance the
photocurrent. As heptamers can provide larger near-field enhancement and higher
38
thermal electron yield, the enhancement effect of heptamers is significantly better than
that of the dimer. In addition, Figure 10d shows that the photocurrent enhancement of
the dimer plasmonic clusters is related to the polarization of the incident light,
indicating that this structure has potential in the field of infrared polarization detection.
Figure 10e and f show the measured photocurrent and the corresponding band diagram
for different applied gate voltages, respectively. The gate voltage VG can effectively
adjust the Fermi level of the graphene channel, and then obtain different energy band
bending between the Fermi level of graphene channel and the Ti electrode-doped
graphene. The steeper band bending results in a correspondingly larger built-in electric
field, which can obtain higher the separation efficiency of the photogenerated electron-
hole pairs and thus produces a greater photocurrent. The resonance wavelength of the
plasmonic clusters can be controlled by varying the diameter of disks. For example,
when the diameter of heptamer plasmonic clusters changes from 80 nm to 180 nm, the
corresponding resonance wavelength changes from 650 nm to 950 nm. Figure 10g
shows the experimentally measured photocurrent as a function of different resonant
wavelength (i.e. different heptamer diameters). It can be clearly seen that the
photocurrent consists of plasmon-induced hot electrons and photogenerated carriers of
plasmon-enhanced direct carrier excitation. The mechanism of photocurrent
enhancement by plasmonic nanostructures is as follows. When the incident light
resonates with the transmission window of the plasmonic clusters, a near-field
enhancement effect occurs.[134] The light absorption of graphene is effectively enhanced
and more electron-hole pairs are generated by direct excitation (DE). Moreover, hot
39
electrons (HE) generated in metal nanostructures can transfer into the conduction band
of graphene over the Schottky barrier between metal and graphene.[135] Both plasmon-
induced hot electrons and directly excited carriers can be driven by the source-drain
bias into the circuit to from photocurrent. Because the contribution of absorption
relative to scattering decreases as the size of the nanoparticles increases, the
contribution of HE to photocurrent tends to saturate, while the contribution of DE plays
a greater role with increasing the diameter of the heptamer.[131]
Since hot electrons are generated by plasmon decay, the response spectra of
photodetectors based on
two-dimensional materials and metal plasmonic
nanostructures are not limited by the bandgap of two-dimensional materials.[129, 130] For
example, bilayer MoS2 has an indirect bandgap of 1.65 eV, corresponding to a
wavelength of 750 nm.[136] The hot electron-based bilayer MoS2 photodetector has a
high responsivity of 5.2 A/W at 1,070 nm (Figure 10h).[129] MoS2 field-effect
phototransistors were also found to have negative infrared photoresponse, which is
caused by bolometric effect, and exhibited responsivity of 2.3 A/W at 980 nm.[137] The
above results show that plasma-induced hot electrons can achieve effective light
detection with below bandgap photon illumination.
Ni et al. reported a photodetector based on plasmonic Si QD/graphene
heterostructure that enables effective detection from UV to mid-infrared.[138] B-doped
Si QDs not only enhance the light absorption of graphene in the infrared region due to
surface plasmon resonance but also form photogating effect, which greatly improves
the performance of the detector. Thus, the device exhibits ultrahigh gain of ~1012 and
40
responsivity of ~109 A/W.
Figure 10
(a) The schematic of graphene-antenna sandwich photodetector. (b)
41
The optical image of device before (left) and after (right) transfer of top layer graphene.
Inset: Raman mapping of the device under 785 nm laser illumination. (c) Photocurrents
measured in the photodetector with dimer array and heptamer array and without pattern
exhibit antisymmetric properties. The results show that the metal plasmonic antenna
clusters can effectively enhance the photocurrent. The line scan position is shown by
the white arrows in the inset. (d) The polarization dependent properties of the
photocurrent measured in dimer and heptamer devices. (e) Photocurrents measured at
different gate voltages in a heptamer device. (f) Energy band diagram of devices with
different gate voltages in a heptamer device. The purple dashed line indicates the Dirac
point of graphene. (g) Photocurrent as a function of different resonant wavelength.
Black triangles: the measured photocurrent (IEXP(λ)). Blue triangles: the calculated DE
photocurrent (IDE(λ)). Solid red triangles: the estimated HE photocurrent contribution
EDE(λ) = IEXP(λ) −IDE(λ). Hollow red triangles: the calculated IHE(λ). (h) Schematic of
hot electron-based bilayer MoS2 with near-infrared laser illumination. (a-g) are
reproduced from Ref.[131] with permission from the American Chemical Society. (h)
is reproduced from Ref.[129] with permission from the American Chemical Society.
Table 1. Performance of room temperature infrared photodetectors with hybrid
structure based on 2D materials. (UV: Ultraviolet. NIR: Near infrared. MIR: Mid-
infrared. BP: Black phosphorus. Gr: Graphene. QD: Quantum dot. FET: Field effect
transistor.)
Device type
Device
Spect Responsi Detectivi Respon
EQE/I
gain
Ref.
42
structure
ral
vity
ty
se time
QE
range
(A/W)
(Jones)
(s)/f3dB
bandwi
dth
Gr-PbS QD NIR
107
-
0.26
-
-
[16][16]
Gr-PbS QD
hybrid
MoS2-HgTe
QD
NIR
1.6104
8.61010
810-3
2.410
6%
-
[30]
Visib
le-
106
1012
10-3
-
10
6
[31]
NIR
Visib
Gr/PbS QD
le-
107
71013
0.01
25%
108
[32]
NIR
Visib
Gr/PbS
QDs/Gr
ITO/PbS
QD/Gr
MoS2/PbS
QD
MoS2/PbSe
QD
le-
58
21011
0.03
-
-
[33]
NIR
Visib
le-
2106
1013
1.5
70%-
KHz
80%
105
[34]
NIR
Visib
le-
105-106
51011
0.3-0.4
NIR
NIR
1.910-6
-
0.25
-
-
-
[35]
-
[36]
SWNT-Gr
400-
>100 (650
nm)
110-4
34%
105
hybrid
1,550
-
(650
(650
(650
[50]
films
nm
SWNT-
MoS2
Visib
le-
NIR
40 (1550
nm)
>0.1 (650
nm)
nm)
nm)
nm)
<1.51
25%
-
0-5 (650
(650
-
[60]
nm)
nm)
Hybrid 2D-
0D
structures
Hybrid 2D-
1D
structures
43
Visib
WSe2-In2O3
le-
NIR
7.5105
4.171017
(637 nm)
(637 nm)
3.5104
1.951016
(940 nm)
(940 nm)
0.1210-
210-2
(940
-
-
[139]
nm)
Gr-WSe2-
Gr
NIR
3 (1500
-
-
2%
-
[140]
nm)
MoS2 PN
Visib
homojuncti
le-
7104
3.51014
10-2
>10%
>105
[141]
on
NIR
Visib
Gr-Bi2Se3
le-
MIR
1.97
(3500
nm)
8.18
(1300
nm)
1.7109
410-6
2.3%
30
[142]
Hybrid 2D-
2D
Few-Layer
Visib
structures
-MoTe2
le-
/MoS2
NIR
0.322
(470 nm)
0.037
(800 nm)
2.510-
-
2 (620
nm)
85%
(470
nm)
6%
(800
nm)
-
[76]
MoS2-Gr-
WSe2
Visib
le-
MIR
0.306
1011 (940
(940 nm)
nm)
3.0310
106%
-5 (637
(532
-
[79]
nm)
nm)
Gr-Bi2Te3
UV-
10 (980
NIR
nm)
-
Visib
103 (637
1014 (637
9.310-
3
(1550n
m)
83
-
(532
[82]
nm)
106
BP/WSe2
le-
nm)
NIR
0.5
nm)
1010
810-4
-
(637
[85]
nm)
44
(1550
(1550
nm)
nm)
102
(1550
nm)
BP p-n
junction
1011-
1013
NIR
0.18
NIR
4.810-3
MoTe2 p-n
(1110
junction
-1200
(1160
-
200
MHz
nm)
nm)
metal-
0.015
0.75%
-
[88]
0.5%
(1160
-
[91]
nm)
doped Gr
NIR
0.1
-
20 GHz
3.8%
-
[98]
junction
Metal-Gr-Si
1550
nm
0.37
-
-
7%
2
[99]
0.023
(19.5
Gr-MoTe2-
Waveguide-
Au
integrated
structures
NIR
nm)
-
1 GHz
0.4 (60
nm)
1.5%
(19.5
nm)
35%
(60
nm)
-
[100]
Gr
NIR
0.05
Visib
0.13
Gr/Si
le-
(2750
MIR
nm)
-
-
18
GHz
10%
-
[101]
410-8
71.5%
-
[102]
0.135
(11.5-
nm-
10%
(11.5-
nm-
BP FET
NIR
thick)
-
>3 GHz
thick)
-
[103]
0.657
(100-nm-
thick)
50%
(100-
nm-
45
Microcavity
Gr
-integrated
structures
Gr-Si
864.5
nm
1550
nm
2.110-2
-
-
210-2
5.1107
1.3510
-9
thick)
-
-
-
-
[123]
[124]
Pt
14 (325
nm)
nanostrips
UV-
312.5
-
NIR
(532 nm)
MoS2
69.2
Gr-Au
UV-
nanorods
NIR
(980 nm)
4104
(1310
nm)
1.1105
Plasmonic-
integrated
Bilayer
structures
MoS2-Au
Visib
(532 nm)
le-
5.2
-
NIR
(1070
nm)
410-3
Few-layer
Visib
(650 nm)
1.11310
0.707
10 (532
(532
nm)
nm)
7.283104
(532 nm)
-
[126]
0.7 s
-
(1310
-
1.8109
(1310 nm)
[128]
nm)
28.5
(532
nm)
44.5
(1070
nm)
1.05105
-
(532 nm)
[129]
MoS2-metal
le-
0.510-
-
-
-
-
[130]
junctions
NIR
4 (850
Visib
Gr-antenna
le-
NIR
nm)
1.251
0-2
5. Conclusion and prospects
-
-
22%
-
[131]
In this review, we have reviewed various infrared photodetectors operating at room
46
temperature with hybrid structure based on two-dimensional materials. Hybrid 2D-QDs
heterostructures possess two prominent advantages of strong light absorption
contributed by quantum dots and high carrier mobility contributed by two-dimensional
materials, and thus exhibiting extremely high gain and responsivity. The detection
performance of photodetectors based on hybrid graphene-SWNTs structure is superior
to that of photodetectors based on only graphene or SWNTs, showing high responsivity
and fast response, in which the bandgap of SWNTs can be adjusted by the diameter of
SWNTs to extend the spectral response range to infrared. The 2D-2D van der Waals
heterostructure is also suitable for light detection in the infrared region because of
ultrafast charge transfer at the interface, and the combination of two or more two-
dimensional materials with different optical and electrical properties not only
compensates for the lack of both (eg. the combination of 2D materials with different
bandgap structures can achieve broadband detection), but also exploits the
characteristics of 2D materials to prepare a photodetector with specific function (such
as high polarization sensitivity of a BP-on-WSe2 photogate structure). The integration
of two-dimensional material on waveguide makes the light-matter interaction no longer
limited by the ultrathin thickness of two-dimensional material, greatly increasing the
light absorption of photodetectors to obtain a higher responsivity. In addition, this
integration also shows that two-dimensional materials are compatible with the existing
mature optical waveguide technology, providing a pathway for the commercialized
practical application of two-dimensional materials. The integration of two-dimensional
material with the cavity enhances the light absorption of the material at the resonant
47
wavelength of the cavity. This selective absorption method is very suitable for precise
detection of specific wavelengths. Metal plasmonic nanostructures can also enhance
light absorption and photocurrent through near-field enhancement effects and hot
electrons. In addition, plasma-induced hot electrons can extend the spectral response
range, which is useful for infrared detection based on some two-dimensional materials
with wide bandgap.
Infrared photodetectors with hybrid structure based on two-dimensional materials
still face many challenges and need us to explore further.
1) Low-cost, large-scale preparation
Many of the currently reported hybrid structures based on two-dimensional
materials are prepared just only through mechanical exfoliation and targeted transfer
techniques, leading to poor reproducibility and relatively low yield of photodetectors.
This method can only meet the needs of scientific research and cannot be used for
commercial applications. Although the 2D vertical or lateral heterostructure can be
prepared by the CVD method and 2D-QDs structure can be prepared by the solution-
processed method, which shows the potential of large-scale production, but they still
have many limitations and need further improve.
2) Device performance optimization
The reported overall performance parameters of infrared photodetectors with
hybrid structure based on two-dimensional materials are excellent, but they are still far
from the theoretical value. The excellent optical and electrical properties of two-
dimensional materials have not been fully exploited. In order to obtain better detection
48
performance, it is necessary to further study the interfacial charge transfer and
photocurrent generation mechanism in hybrid structure based on two-dimensional
materials for designing or optimizing various hybrid structures.
3) Research on novel and suitable two-dimensional materials
There are various currently limitations for two-dimensional materials used in the
field of infrared detection. For example, the band gap of graphene and many TMDs
(such as MoS2, WSe2) are not conducive to the absorption of infrared light. The
instability of MoTe2 and BP with an appropriately bandgap in the air also limits their
use. In order to obtain better infrared detection performance, novel two-dimensional
materials with suitable band gap and good stability have to be developed.
4) Design of novel heterostructure with 2D materials
Many existing heterostructures based on two-dimensional materials simply
combine two materials that appear to be suitable. However, this combination may not
be optimal, which may leads to the inability to maximize the advantages of
heterostructure. It is a worthwhile research direction to further understand the properties
of various two-dimensional materials through theoretical research and computational
simulation methods, and to design a more rational and ingenious novel heterostructure.
It has the potential to dramatically improve the performance of the photodetector and
even discover new phenomena.
5) Interface research of heterostructures
Two-dimensional materials are particularly sensitive to interface problem due to
their large specific surface area and atomic thickness. On the one hand, interface
49
pollution and the introduction of defects have a huge impact on the physical properties
of two-dimensional materials. Therefore, how to obtain a clean and defect-free interface
in the preparation process of heterostructure based on two-dimensional materials has
always been a research hotspot. On the other hand, the study of charge transfer and
interaction at the interface of two materials also helps to clarify the working mechanism
of the photodetector and to improve performance of photodetector after the formation
of the heterostructure.
Declaration of conflicts of Interest
The authors declare no conflict of interest.
Acknowledgements
The authors acknowledge funding by National Natural Science Foundation of China
(No. 61704061)
References
[1]
Norton P R, Campbell J B, Horn S B and Reago D A International Symposium
on Optical Science and Technology, p. 11
Rogalski A 2012 Opto-Electronics Review 20 279
Rogalski A, Antoszewski J and Faraone L 2009 Journal of Applied Physics 105
091101
Xu M, Liang T, Shi M and Chen H 2013 Chemical Reviews 113 3766
Butler S Z, Hollen S M, Cao L, Cui Y, Gupta J A, Gutiérrez H R, Heinz T F,
Hong S S, Huang J, Ismach A F, Johnston-Halperin E, Kuno M, Plashnitsa V V,
Robinson R D, Ruoff R S, Salahuddin S, Shan J, Shi L, Spencer M G, Terrones
[2]
[3]
[4]
[5]
50
M, Windl W and Goldberger J E 2013 ACS Nano 7 2898
[6] Wu Pei H X, Zhang Jian, Sun Lian-Feng 2017 Acta Physica Sinica 66 218102
[7]
Tan Wee C, Huang L, Ng Rui J, Wang L, Hasan Dihan Md N, Duffin Thorin J,
Kumar Karuppannan S, Nijhuis Christian A, Lee C and Ang K W 2017
Advanced Materials 30 1705039
[8]
Lai J, Liu X, Ma J, Wang Q, Zhang K, Ren X, Liu Y, Gu Q, Zhuo X, Lu W, Wu
Y, Li Y, Feng J, Zhou S, Chen J-H and Sun D 2018 Advanced Materials 30
1707152
[9]
Xuefei Li X X, Yanqing Wu 2017 Chin. Phys. B 26 37307
[10]
Jianlu Wang W H 2017 Chin. Phys. B 26 37106
[11] Yuanfang Yu F M, Jun He, Zhenhua Ni 2017 Chin. Phys. B 26 36801
[12] Zi-Wei Li Y-H H, Yu Li, Zhe-Yu Fang 2017 Chin. Phys. B 26 36802
[13] Xiaomu Wang X G 2017 Chin. Phys. B 26 34203
[14] Novoselov K S, Fal′ko V I, Colombo L, Gellert P R, Schwab M G and Kim
K 2012 Nature 490 192
[15] Allen M J, Tung V C and Kaner R B 2010 Chemical Reviews 110 132
[16] Sun Z, Liu Z, Li J, Tai G a, Lau S P and Yan F 2012 Advanced Materials 24
5878
[17] Zhang B Y, Liu T, Meng B, Li X, Liang G, Hu X and Wang Q J 2013 Nature
Communications 4 1811
[18] Castellanos-Gomez A 2015 The Journal of Physical Chemistry Letters 6 4280
[19] Xia F, Wang H and Jia Y 2014 Nature Communications 5 4458
[20] Nair R R, Blake P, Grigorenko A N, Novoselov K S, Booth T J, Stauber T, Peres
N M R and Geim A K 2008 Science 320 1308
[21] Mak K F, Lee C, Hone J, Shan J and Heinz T F 2010 Physical Review Letters
105 136805
[22] Moreels I, Justo Y, De Geyter B, Haustraete K, Martins J C and Hens Z 2011
ACS Nano 5 2004
[23] Yu X, Li Y, Hu X, Zhang D, Tao Y, Liu Z, He Y, Haque M A, Liu Z, Wu T and
Wang Q J 2018 Nature Communications 9 4299
[24] Eatemadi A, Daraee H, Karimkhanloo H, Kouhi M, Zarghami N, Akbarzadeh
A, Abasi M, Hanifehpour Y and Joo S W 2014 Nanoscale Research Letters 9
393
51
[25] Brongersma M L, Halas N J and Nordlander P 2015 Nature Nanotechnology 10
25
[26] Dean C R, Young A F, Meric I, Lee C, Wang L, Sorgenfrei S, Watanabe K,
Taniguchi T, Kim P, Shepard K L and Hone J 2010 Nature Nanotechnology 5
722
[27] Mueller T, Xia F and Avouris P 2010 Nature Photonics 4 297
[28] Xia F, Mueller T, Lin Y-m, Valdes-Garcia A and Avouris P 2009 Nature
Nanotechnology 4 839
[29] Splendiani A, Sun L, Zhang Y, Li T, Kim J, Chim C-Y, Galli G and Wang F
2010 Nano Letters 10 1271
[30] Che Y, Zhang Y, Cao X, Zhang H, Song X, Cao M, Yu Y, Dai H, Yang J, Zhang
G and Yao J 2017 ACS Applied Materials & Interfaces 9 32001
[31] Huo N, Gupta S and Konstantatos G 2017 Advanced Materials 1606576
[32] Konstantatos G, Badioli M, Gaudreau L, Osmond J, Bernechea M, Arquer F P
G D, Gatti F and Koppens F H L 2012 Nature Nanotechnology 7 363
[33] Nian Q, Gao L, Hu Y, Deng B, Tang J and Cheng G J 2017 ACS Applied
Materials & Interfaces 9 44715
[34] Nikitskiy I, Goossens S, Kufer D, Lasanta T, Navickaite G, Koppens F H and
Konstantatos G 2016 Nature Communications 7 11954
[35] Kufer D, Nikitskiy I, Lasanta T, Navickaite G, Koppens Frank H L and
Konstantatos G 2015 Advanced Materials 27 176
[36] Schornbaum J, Winter B, Schiessl Stefan P, Gannott F, Katsukis G, Guldi Dirk
M, Spiecker E and Zaumseil J 2014 Advanced Functional Materials 24 5798
[37] Zhang Y, Cao M, Song X, Wang J, Che Y, Dai H, Ding X, Zhang G and Yao J
2015 The Journal of Physical Chemistry C 119 21739
[38] Zhang D, Gan L, Cao Y, Wang Q, Qi L and Guo X 2012 Advanced Materials
24 2715
[39] Sun M, Fang Q, Xie D, Sun Y, Qian L, Xu J, Xiao P, Teng C, Li W, Ren T and
Zhang Y 2018 Nano Research 11 3233
[40] Giansante C, Infante I, Fabiano E, Grisorio R, Suranna G P and Gigli G 2015
Journal of the American Chemical Society 137 1875
[41] Konstantatos G, Clifford J, Levina L and Sargent E H 2007 Nature Photonics 1
531
52
[42] Zheng L, Zhou W, Ning Z, Wang G, Cheng X, Hu W, Zhou W, Liu Z, Yang S,
Xu K, Luo M and Yu Y 2018 Advanced Optical Materials 0 1800985
[43] Radisavljevic B, Radenovic A, Brivio J, Giacometti V and Kis A 2011 Nature
Nanotechnology 6 147
[44]
Jin W, Yeh P-C, Zaki N, Zhang D, Sadowski J T, Al-Mahboob A, van der Zande
A M, Chenet D A, Dadap J I, Herman I P, Sutter P, Hone J and Osgood R M
2013 Physical Review Letters 111 106801
[45] Eda G, Yamaguchi H, Voiry D, Fujita T, Chen M and Chhowalla M 2011 Nano
Letters 11 5111
[46] Wu D, Ma Y, Niu Y, Liu Q, Dong T, Zhang S, Niu J, Zhou H, Wei J, Wang Y,
Zhao Z and Wang N 2018 Science Advances 4
[47] Keuleyan S E, Guyot-Sionnest P, Delerue C and Allan G 2014 ACS Nano 8
8676
[48] Lhuillier E, Scarafagio M, Hease P, Nadal B, Aubin H, Xu X Z, Lequeux N,
Patriarche G, Ithurria S and Dubertret B 2016 Nano Letters 16 1282
[49] Kufer D, Lasanta T, Bernechea M, Koppens F H L and Konstantatos G 2016
ACS Photonics 3 1324
[50] Liu Y, Wang F, Wang X, Wang X, Flahaut E, Liu X, Li Y, Wang X, Xu Y, Shi Y
and Zhang R 2015 Nature Communications 6 8589
[51] Roy K, Padmanabhan M, Goswami S, Sai T P, Ramalingam G, Raghavan S and
Ghosh A 2013 Nature Nanotechnology 8 826
[52] Konstantatos G, Levina L, Fischer A and Sargent E H 2008 Nano Letters 8 1446
[53] Avouris P, Freitag M and Perebeinos V 2008 Nature Photonics 2 341
[54] Saito R, Fujita M, Dresselhaus G and Dresselhaus M S 1992 Applied Physics
Letters 60 2204
[55] Kataura H, Kumazawa Y, Maniwa Y, Umezu I, Suzuki S, Ohtsuka Y and Achiba
Y 1999 Synthetic Metals 103 2555
[56] Dürkop T, Getty S A, Cobas E and Fuhrer M S 2004 Nano Letters 4 35
[57] Gabor N M, Zhong Z, Bosnick K, Park J and McEuen P L 2009 Science 325
1367
[58] Paulus Geraldine L C, Wang Qing H, Ulissi Zachary W, McNicholas Thomas P,
Vijayaraghavan A, Shih C J, Jin Z and Strano Michael S 2012 Small 9 1954
[59] Liu Y, Liu Y, Qin S, Xu Y, Zhang R and Wang F 2017 Nano Research 10 1880
53
[60]
Jariwala D, Sangwan V K, Wu C-C, Prabhumirashi P L, Geier M L, Marks T J,
Lauhon L J and Hersam M C 2013 Proceedings of the National Academy of
Sciences 110 18076
[61] Mak K F, He K, Lee C, Lee G H, Hone J, Heinz T F and Shan J 2012 Nature
Materials 12 207
[62] Liu S, Liao Q, Zhang Z, Zhang X, Lu S, Zhou L, Hong M, Kang Z and Zhang
Y 2017 Nano Research 10 3476
[63] Pizzocchero F, Gammelgaard L, Jessen B S, Caridad J M, Wang L, Hone J,
Bøggild P and Booth T J 2016 Nature Communications 7 11894
[64] Gong Y, Lin J, Wang X, Shi G, Lei S, Lin Z, Zou X, Ye G, Vajtai R, Yakobson
B I, Terrones H, Terrones M, Tay Beng K, Lou J, Pantelides S T, Liu Z, Zhou
W and Ajayan P M 2014 Nature Materials 13 1135
[65] Gong Y, Lei S, Ye G, Li B, He Y, Keyshar K, Zhang X, Wang Q, Lou J, Liu Z,
Vajtai R, Zhou W and Ajayan P M 2015 Nano Letters 15 6135
[66] He Y, Yang Y, Zhang Z, Gong Y, Zhou W, Hu Z, Ye G, Zhang X, Bianco E, Lei
S, Jin Z, Zou X, Yang Y, Zhang Y, Xie E, Lou J, Yakobson B, Vajtai R, Li B and
Ajayan P 2016 Nano Letters 16 3314
[67] Yu Y, Hu S, Su L, Huang L, Liu Y, Jin Z, Purezky A A, Geohegan D B, Kim K
W, Zhang Y and Cao L 2015 Nano Letters 15 486
[68] Wang Q, Wen Y, Cai K, Cheng R, Yin L, Zhang Y, Li J, Wang Z, Wang F, Wang
F, Shifa T A, Jiang C, Yang H and He J 2018 Science Advances 4
[69] Fang H, Battaglia C, Carraro C, Nemsak S, Ozdol B, Kang J S, Bechtel H A,
Desai S B, Kronast F, Unal A A, Conti G, Conlon C, Palsson G K, Martin M C,
Minor A M, Fadley C S, Yablonovitch E, Maboudian R and Javey A 2014
Proceedings of the National Academy of Sciences 111 6198
[70] Hong X, Kim J, Shi S-F, Zhang Y, Jin C, Sun Y, Tongay S, Wu J, Zhang Y and
Wang F 2014 Nature Nanotechnology 9 682
[71] Ceballos F, Bellus M Z, Chiu H-Y and Zhao H 2014 ACS Nano 8 12717
[72] Long R and Prezhdo O V 2016 Nano Letters 16 1996
[73] Xia Wei F-G Y, Chao Shen, Quan-Shan Lv, Kai-You Wang 2017 Chin. Phys. B
26 38504
[74] Lezama I G, Arora A, Ubaldini A, Barreteau C, Giannini E, Potemski M and
Morpurgo A F 2015 Nano Letters 15 2336
54
[75] Ruppert C, Aslan O B and Heinz T F 2014 Nano Letters 14 6231
[76] Pezeshki A, Shokouh Seyed Hossein H, Nazari T, Oh K and Im S 2016
Advanced Materials 28 3216
[77] Liu H, Neal A T, Zhu Z, Luo Z, Xu X, Tománek D and Ye P D 2014 ACS Nano
8 4033
[78] Tran V, Soklaski R, Liang Y and Yang L 2014 Physical Review B 89 235319
[79] Long M, Liu E, Wang P, Gao A, Xia H, Luo W, Wang B, Zeng J, Fu Y, Xu K,
Zhou W, Lv Y, Yao S, Lu M, Chen Y, Ni Z, You Y, Zhang X, Qin S, Shi Y, Hu
W, Xing D and Miao F 2016 Nano Letters 16 2254
[80] Das S, Chen H-Y, Penumatcha A V and Appenzeller J 2013 Nano Letters 13
100
[81] Ross J S, Wu S, Yu H, Ghimire N J, Jones A M, Aivazian G, Yan J, Mandrus D
G, Xiao D, Yao W and Xu X 2013 Nature Communications 4 1474
[82] Qiao H, Yuan J, Xu Z, Chen C, Lin S, Wang Y, Song J, Liu Y, Khan Q, Hoh H
Y, Pan C-X, Li S and Bao Q 2015 ACS Nano 9 1886
[83] Dang W, Peng H, Li H, Wang P and Liu Z 2010 Nano Letters 10 2870
[84] Hasan M Z and Kane C L 2010 Reviews of Modern Physics 82 3045
[85] Ye L, Wang P, Luo W, Gong F, Liao L, Liu T, Tong L, Zang J, Xu J and Hu W
2017 Nano Energy 37 53
[86] Bullock J, Amani M, Cho J, Chen Y-Z, Ahn G H, Adinolfi V, Shrestha V R, Gao
Y, Crozier K B, Chueh Y-L and Javey A 2018 Nature Photonics 12 601
[87] Choi M S, Qu D, Lee D, Liu X, Watanabe K, Taniguchi T and Yoo W J 2014
ACS Nano 8 9332
[88] Yu X, Zhang S, Zeng H and Wang Q J 2016 Nano Energy 25 34
[89] Li L, Yu Y, Ye G J, Ge Q, Ou X, Wu H, Feng D, Chen X H and Zhang Y 2014
Nature Nanotechnology 9 372
[90] Pradhan N R, Rhodes D, Xin Y, Memaran S, Bhaskaran L, Siddiq M, Hill S,
Ajayan P M and Balicas L 2014 ACS Nano 8 7923
[91] Bie Y-Q, Grosso G, Heuck M, Furchi M M, Cao Y, Zheng J, Bunandar D,
Navarro-Moratalla E, Zhou L, Efetov D K, Taniguchi T, Watanabe K, Kong J,
Englund D and Jarillo-Herrero P 2017 Nature Nanotechnology 12 1124
[92] Duan X, Wang C, Shaw J C, Cheng R, Chen Y, Li H, Wu X, Tang Y, Zhang Q,
Pan A, Jiang J, Yu R, Huang Y and Duan X 2014 Nature Nanotechnology 9
55
1024
[93] Huang C, Wu S, Sanchez A M, Peters J J P, Beanland R, Ross J S, Rivera P, Yao
W, Cobden D H and Xu X 2014 Nature Materials 13 1096
[94] Li M-Y, Shi Y, Cheng C-C, Lu L-S, Lin Y-C, Tang H-L, Tsai M-L, Chu C-W,
Wei K-H, He J-H, Chang W-H, Suenaga K and Li L-J 2015 Science 349 524
[95] Sahoo P K, Memaran S, Xin Y, Balicas L and Gutiérrez H R 2018 Nature 553
63
[96] Zhang Z, Chen P, Duan X, Zang K, Luo J and Duan X 2017 Science 357 788
[97] Du Y, Liu H, Deng Y and Ye P D 2014 ACS Nano 8 10035
[98] Gan X, Shiue R-J, Gao Y, Meric I, Heinz T F, Shepard K, Hone J, Assefa S and
Englund D 2013 Nature Photonics 7 883
[99] Goykhman I, Sassi U, Desiatov B, Mazurski N, Milana S, de Fazio D, Eiden A,
Khurgin J, Shappir J, Levy U and Ferrari A C 2016 Nano Letters 16 3005
[100] Ma P, Flöry N, Salamin Y, Baeuerle B, Emboras A, Josten A, Taniguchi T,
Watanabe K, Novotny L and Leuthold J 2018 ACS Photonics
[101] Pospischil A, Humer M, Furchi M M, Bachmann D, Guider R, Fromherz T and
Mueller T 2013 Nature Photonics 7 892
[102] Wang X, Cheng Z, Xu K, Tsang H K and Xu J-B 2013 Nature Photonics 7 888
[103] Youngblood N, Chen C, Koester S J and Li M 2015 Nature Photonics 9 247
[104] Yubing Wang W Y, Qin Han, Xiaohong Yang, Han Ye, Qianqian Lv, Dongdong
Yin 2016 Chin. Phys. B 25 118103
[105] Liu M, Yin X, Ulin-Avila E, Geng B, Zentgraf T, Ju L, Wang F and Zhang X
2011 Nature 474 64
[106] Echtermeyer T J, Britnell L, Jasnos P K, Lombardo A, Gorbachev R V,
Grigorenko A N, Geim A K, Ferrari A C and Novoselov K S 2011 Nature
Communications 2 458
[107] Gu T, Petrone N, McMillan J F, van der Zande A, Yu M, Lo G Q, Kwong D L,
Hone J and Wong C W 2012 Nature Photonics 6 554
[108] Bao Q, Zhang H, Wang B, Ni Z, Lim C H Y X, Wang Y, Tang D Y and Loh K
P 2011 Nature Photonics 5 411
[109] Freitag M, Low T, Xia F and Avouris P 2012 Nature Photonics 7 53
[110] Low T, Rodin A S, Carvalho A, Jiang Y, Wang H, Xia F and Castro Neto A H
2014 Physical Review B 90 075434
56
[111] Hong T, Chamlagain B, Lin W Z, Chuang H J, Pan M H, Zhou Z X and Xu Y
Q 2014 NANOSCALE 6 8978
[112] Low T, Engel M, Steiner M and Avouris P 2014 Physical Review B 90 081408
[113] Das S, Zhang W, Demarteau M, Hoffmann A, Dubey M and Roelofs A 2014
Nano Letters 14 5733
[114] Robert C, Picard R, Lagarde D, Wang G, Echeverry J P, Cadiz F, Renucci P,
Högele A, Amand T, Marie X, Gerber I C and Urbaszek B 2016 Physical Review
B 94 155425
[115] Froehlicher G, Lorchat E and Berciaud S 2016 Physical Review B 94 085429
[116] Yin L, Zhan X, Xu K, Wang F, Wang Z, Huang Y, Wang Q, Jiang C and He J
2016 Applied Physics Letters 108 043503
[117] Octon Tobias J, Nagareddy V K, Russo S, Craciun Monica F and Wright C D
2016 Advanced Optical Materials 4 1750
[118] Gan X, Gao Y, Fai Mak K, Yao X, Shiue R-J, van der Zande A, Trusheim M E,
Hatami F, Heinz T F, Hone J and Englund D 2013 Applied Physics Letters 103
181119
[119] Wu S, Buckley S, Schaibley J R, Feng L, Yan J, Mandrus D G, Hatami F, Yao
W, Vučković J, Majumdar A and Xu X 2015 Nature 520 69
[120] Liu X, Galfsky T, Sun Z, Xia F, Lin E-c, Lee Y-H, Kéna-Cohen S and Menon V
M 2014 Nature Photonics 9 30
[121] Dufferwiel S, Schwarz S, Withers F, Trichet A A P, Li F, Sich M, Del Pozo-
Zamudio O, Clark C, Nalitov A, Solnyshkov D D, Malpuech G, Novoselov K
S, Smith J M, Skolnick M S, Krizhanovskii D N and Tartakovskii A I 2015
Nature Communications 6 8579
[122] Ye Y, Wong Z J, Lu X, Ni X, Zhu H, Chen X, Wang Y and Zhang X 2015 Nature
Photonics 9 733
[123] Furchi M, Urich A, Pospischil A, Lilley G, Unterrainer K, Detz H, Klang P,
Andrews A M, Schrenk W, Strasser G and Mueller T 2012 Nano Letters 12 2773
[124] Casalino M, Sassi U, Goykhman I, Eiden A, Lidorikis E, Milana S, De Fazio D,
Tomarchio F, Iodice M, Coppola G and Ferrari A C 2017 ACS Nano 11 10955
[125] Guo Q, Li C, Deng B, Yuan S, Guinea F and Xia F 2017 ACS Photonics 4 2989
[126] Kumar R, Sharma A, Kaur M and Husale S 2017 Advanced Optical Materials
5 1700009
57
[127] Luxmoore I J, Liu P Q, Li P, Faist J and Nash G R 2016 ACS Photonics 3 936
[128] Xia Z, Li P, Wang Y, Song T, Zhang Q and Sun B 2015 ACS Applied Materials
& Interfaces 7 24136
[129] Wang W, Klots A, Prasai D, Yang Y, Bolotin K I and Valentine J 2015 Nano
Letters 15 7440
[130] Hong T, Chamlagain B, Hu S, Weiss S M, Zhou Z and Xu Y-Q 2015 ACS Nano
9 5357
[131] Fang Z, Liu Z, Wang Y, Ajayan P M, Nordlander P and Halas N J 2012 Nano
Letters 12 3808
[132] Somayyeh Asgari H R, Nosrat Granpayeh, Homayoon Oraizi 2018 Chin. Phys.
B 27 84212
[133] Chen Z, Li X, Wang J, Tao L, Long M, Liang S-J, Ang L K, Shu C, Tsang H K
and Xu J-B 2017 ACS Nano 11 430
[134] Stockman M I 2010 Nature 467 541
[135] Knight M W, Sobhani H, Nordlander P and Halas N J 2011 Science 332 702
[136] Lee H S, Min S-W, Chang Y-G, Park M K, Nam T, Kim H, Kim J H, Ryu S and
Im S 2012 Nano Letters 12 3695
[137] Wu J-Y, Chun Y T, Li S, Zhang T, Wang J, Shrestha P K and Chu D 2018
Advanced Materials 30 1705880
[138] Ni Z, Ma L, Du S, Xu Y, Yuan M, Fang H, Wang Z, Xu M, Li D, Yang J, Hu W,
Pi X and Yang D 2017 ACS Nano 11 9854
[139] Guo N, Gong F, Liu J, Jia Y, Zhao S, Liao L, Su M, Fan Z, Chen X, Lu W, Xiao
L and Hu W 2017 ACS Applied Materials & Interfaces 9 34489
[140] Massicotte M, Schmidt P, Vialla F, Watanabe K, Taniguchi T, Tielrooij K J and
Koppens F H L 2016 Nature Communications 7 12174
[141] Huo N and Konstantatos G 2017 Nature Communications 8 572
[142] Kim J, Park S, Jang H, Koirala N, Lee J-B, Kim U J, Lee H-S, Roh Y-G, Lee H,
Sim S, Cha S, In C, Park J, Lee J, Noh M, Moon J, Salehi M, Sung J, Chee S-
S, Ham M-H, Jo M-H, Oh S, Ahn J-H, Hwang S W, Kim D and Choi H 2017
ACS Photonics 4 482
58
|
1907.11135 | 1 | 1907 | 2019-06-05T03:00:49 | Fast electro-optic switching for coherent laser ranging and velocimetry | [
"physics.app-ph",
"physics.optics"
] | The growth of 3D imaging across a range of sectors has driven a demand for high performance beam steering techniques. Fields as diverse as autonomous vehicles and medical imaging can benefit from a high speed, adaptable method of beam steering. We present a monolithic, sub-microsecond electro-optic switch as a solution satisfying the need for reliability, speed, dynamic addressability and compactness. Here we demonstrate a laboratory-scale, solid-state lidar pointing system, using the electro-optic switch to launch modulated coherent light into free space, and then to collect the reflected signal. We use coherent detection of the reflected light to simultaneously extract the range and axial velocity of targets at each of several electronically addressable output ports. | physics.app-ph | physics | Fast electro-optic switching for coherent laser ranging and velocimetry
B. Haylock,1 M.A. Baker,2 T.M. Stace,2,* and M. Lobino1,3,*
1Centre for Quantum Computation & Communication Technology and Centre for Quantum Dynamics, Griffith
University, Brisbane QLD 4111, Australia
2ARC Centre of Excellence for Engineered Quantum Systems (EQUS), School of Mathematics and Physics, University
of Queensland, Brisbane, 4072, Australia
3Queensland Micro- and Nanotechnology Centre, Griffith University, Brisbane, 4111, Australia
The growth of 3D imaging across a range of sectors has driven a demand for high performance beam steering
techniques. Fields as diverse as autonomous vehicles and medical imaging can benefit from a high speed, adaptable
method of beam steering. We present a monolithic, sub-microsecond electro-optic switch as a solution satisfying the
need for reliability, speed, dynamic addressability and compactness. Here we demonstrate a laboratory-scale, solid-
state lidar pointing system, using the electro-optic switch to launch modulated coherent light into free space, and then
to collect the reflected signal. We use coherent detection of the reflected light to simultaneously extract the range and
axial velocity of targets at each of several electronically addressable output ports.
Optical scanners, capable of high-speed optical beam pointing are essential for many imaging techniques, including lidar
and medical imaging applications. The first commercial demonstrations of multi-pixel lidar sensors relied on mechanical
spinning mirrors, which are cumbersome and lack dynamic addressability. Similarly full-field optical coherence tomography
relied on sample stage movement, or mechanical mirror steering to scan a sample. Recent advances have moved to simple
integrated beam scanning techniques, including MEMS mirrors [1,2], optical phase arrays [3 -- 5], and VCSELs [6,7]. Other
major approaches include liquid crystal electro-optic scanners [8 -- 10], electro-optic beam deflectors [11 -- 13], and spectral
scanning [14 -- 16]. These new beam scanning techniques have allowed improved sensing performance by increasing the size
and refresh rate of the generated point cloud. Most of these beam scanning technologies still limit the point cloud size and
refresh rate due to speed limitations, with the notable exception of indium phosphide optical phase arrays, who have angle
sweep rates of >10∘/𝜇𝜇𝜇𝜇, [17].
An alternative approach to spatial beam manipulation is to use a device with distinct separate spatial output modes to
perform a discrete 'point-by-point' scan rather than a continuous sweep. A reconfigurable waveguide network can perform
such a discrete scan. This approach ensures high speed, side-lobe-free, single mode, and single wavelength beam steering with
the field of view and resolution set instead by the output optics. Such discrete scanning has previously been demonstrated with
a silicon photonic integrated circuit, where the output channel is controlled thermally [18]. Here, we demonstrate a fibre-to-
* Authors to whom correspondence should be addressed. Electronic mail: [email protected]. [email protected]
1
free-space multiplexing switch based on an integrated electro-optic device, which enables high speed, solid-state, single-mode
output optical beam steering and light collection. We use this capability to demonstrate integrated laser ranging and velocimetry
using coherent, modulated light and detection.
The multiplexing switch is constructed from a network of directional couplers, and is fabricated by the reverse proton
exchange technique in congruent lithium niobate [19,20]. The splitting ratio of each directional coupler is tunable between 0
and 100% by applying a voltage to electrodes patterned around two evanescently coupled waveguides. For this demonstration
we create a switch with three output channels and total device loss of ~4dB. We characterize the frequency response of the
one such electro-optic directional coupler by using it to amplitude modulate a laser beam. The modulated light is detected on
a fast photodiode (PD), and the frequency response of the PD signal is shown in figure 1b, and demonstrates switching rates
up to 300 MHz, limited by electrode design. With improved designs the response of such a system should reach the GHz range.
Figure 1 a) Sample frequency response of one electrode measured using a photodiode (DET08CFC, Thorlabs, 5GHz), with a
sine wave injected from a waveform generator(E4432-B, Agilent, 250kHz-3GHz). The frequency response of the photodiode
is not removed from the displayed signal. b) AMCW Lidar setup, full details in text. Here the first (signal) channel is shown
at the top and the second (local oscillator) channel shown at the bottom. 3dB DC -- 3dB directional coupler, EOM -- electro-
optic modulator, HD -- homodyne detector, 𝑽𝑽𝑻𝑻 -- tangential velocity of target. Light from the Switching Network is directed
by a lens that maps the spatial location of the switch to a direction in free space. c) Timing diagram for signal modulation
using EOM's, where the high level represents when the EOM transmits light
2
In what follows, we deploy this network switch for beam steering in an amplitude modulated continuous wave (AMCW)
lidar system. Lidar systems are of great interest for automotive hazard detection and navigation, and a number of techniques
have been implemented in practice. These include pulsed time of flight, frequency modulated continuous wave (FMCW), flash,
and coherent flash protocols. Any system used for automotive purposes must support an encoding that can ignore interfering
signals from other road users, such as the output of a chaotic laser for a 3D lidar system [24]. Pulsed systems can achieve this
easily using random amplitude pulse encoding, however flash and FMCW systems are reliant on spectral separation, a scheme
not suitable to produce millions of unique units. AMCW facilitates coherent time-of-flight plus Doppler analysis of the return
signal, using eye-safe average and peak output powers. Our AMCW implementation allows for simultaneous measurement of
range and velocity across discrete pixels in one dimension using only a single coherent detector. Additionally the switch
architecture allows for temporal multiplexing of the detection electronics since all pixels are measured by a single detector.
The full scheme is shown in Figure 1b.
A low noise CW laser at 1550nm (Koheras Boostik, NKT Photonics) is split by a 3dB coupler between two electro-optic
modulators (EOMs). The EOMs are used to modulate the output light to provide time-of-flight sensitivity (see below), which
is then passed through an optical circulator and into the custom-made monolithic 1-to-Nout LiNbO3 electro-optic switching
network. For demonstration purposes, we choose Nout=3. The average power of each switch output is approximately 200 μW,
limited by the maximum input power of the EOMs (100mW) and their transmission. To obtain higher output powers, and
therefore better signal to noise ratios, a low noise erbium doped fiber amplifier should be used after the EOM in the signal arm,
or a lower loss EOM. The output channels of the electro-optic switch are imaged through a lens onto a remote target. In this
case, the test target was a spinning drum of radius 64±1𝑚𝑚𝑚𝑚 with a diffuse surface placed approximately 4.7m away from the
lens.
Depending on the transverse position of the switch output channel, the output of the switch is directed by the lens into rays
in free space (see Fig. 1b), which intercept one or more target objects in the far-field. In our case, the target range is
approximately 10 meters. In the current setup, the transverse separation between light from channels 1 and 2 is 9.4mm, and
between channels 2 and 3 is 4.7mm at the target. These parameters are easily varied using different optics. After reflection
from the object, the reflected light is collected at the same switch port from which it was launched. The collected light is
recombined with the other arm of the 3 dB splitter, which acts as local oscillator, and is detected using a homodyne detector
with bandwidth 100MHz [21]. Distance and velocity can be retrieved from the Fourier spectrum of the homodyne detector
signal, which is captured at a sampling rate of 200MSPS using an oscilloscope.
3
The signal channel EOM is modulated to produce a square-wave, coherent pulse train on the output light, with repetition
rate of 10MHz and pulse width of 10ns. Including intrinsic losses, the EOM transmission is -18.8dB. The second channel's
EOM is modulated to produce the inverse pulse sequence of the first, with a delay suited so that back-reflection off the input
of the electro-optic switch is not mixed with light from the other arm before being incident on the detectors as demonstrated in
Figure 1c. The total loss of the second EOM is -6.1dB. This modulation of the reference signal reduces the detection of the
back reflection from the electro-optic switch and allows an increase in the dynamic range of the lidar measurement. In future
designs, the back reflection could be minimised by using a standard anti-reflection coating on the facets of the chip.
The monolithic electro-optic switching network is programmed to launch and collect light from each output channel in
cyclic order, i.e. a measurement is taken from channel 1, then channel 2, and so on. A timing signal is sent from the switch
control electronics to the oscilloscope to indicate which channel the received signal originates from.
Both range and velocity information can be extracted from the homodyne measurement of the collected light: time-of-
flight ranging information is encoded in the time-delay of the pulse train relative to the launch time. From the relative phase
of the signal reflected off the input of the electro-optic switch and the return signal, the time delay, and therefore the distance
to target can be calculated. Axial velocity of the target is encoded in the Doppler frequency shift from the relative motion of
the object, given a Doppler frequency shift ∆𝑓𝑓, the velocity of the target is given by 𝑣𝑣=𝜆𝜆∆𝑓𝑓=1.55×10−6∆𝑓𝑓 𝑚𝑚𝜇𝜇−1. These
quantities are extracted from the Fourier transform of the homodyne signal, examples of which are shown in Figure 2a).
Figure 2 Sample return waveform from AMCW Lidar System. Each colour represents a different channel. Large peaks at
multiples of 10MHz are caused by the reflected signal from the input of the optical switch. These are used as a timing
reference. Other smaller peaks are the signal returned from the target that has been Doppler shifted.
4
The uncertainty in range and velocity is determined by the number of samples per channel and the signal to noise ratio.
For these measurements the switching rate between different outputs is set to 10 kHz, with a duty cycle of 10% per channel,
meaning 10000 samples are taken per channel per measurement. The frequency (velocity) resolution is calculated from the full
width at half maximum of the lowest frequency Doppler peak, which was ±33𝑘𝑘𝑘𝑘𝑘𝑘(±25𝑚𝑚𝑚𝑚/𝜇𝜇). The position resolution is set
by the sampling rate (200MSPS, ±0.75𝑚𝑚), however the random phase of the noise in the signal increases this uncertainty
proportional to the signal to noise ratio (SNR). The accuracy can be improved by using a higher sampling rate, longer
acquisition times or averaging over many acquisitions.
From the Doppler lidar measurements, using a known radius, the tangential velocity of the spinning target can be
calculated as well as the angle-of-incidence for the light. The surface velocity of the target was set to 𝑣𝑣𝑠𝑠=15.3±0.1 𝑚𝑚/𝜇𝜇,
and based on the angle of incidence, we calculate the axial target velocity 𝑣𝑣𝑡𝑡=𝑣𝑣𝑠𝑠sin(𝜃𝜃𝑖𝑖𝑖𝑖𝑖𝑖) to compare with the Doppler
inferred velocity. These measured and computed quantities are summarized in Table 1, showing very good agreement
between the measured and inferred velocities. For ranging, in our demonstration, the return signal was sufficiently weak that
ranging data based on the phase shift was subject to large uncertainty. However, time-domain analysis of the return signal is
capable of giving much improved ranging accuracy.
Channel
1
2
3
Incidence
Angle, 𝜃𝜃𝑖𝑖𝑖𝑖𝑖𝑖
23∘
19∘
10∘
SNR (dB)
7.1
4.3
4.9
Doppler-measured,
Velocity (𝒎𝒎/𝒔𝒔)
[Calculated, 𝑣𝑣𝑡𝑡=𝑣𝑣𝑠𝑠sin(𝜃𝜃𝑖𝑖𝑖𝑖𝑖𝑖)]
5.97±0.03, [5.98]
4.91±0.03, [4.98]
2.64±0.03, [2.66]
Range (m)
4.5±0.6m
5.8±1.3m
4.5±1.1m
Table 1 Summary of measurement results for three channel AMCW lidar. Range uncertainty limited by both sampling
rate and SNR. The target was located a distance D=4.7m from the output.
Because of the low pulse repetition rate the unambiguous velocity measurement range is only 7.75𝑚𝑚𝜇𝜇−1, and
unambiguous range measurement interval is only 15m. This can be improved with higher pulse repetition rates and improved
modulation schemes respectively. This beam steering method is also suitable for simultaneous range and velocity
measurement using alternative lidar schemes.
An advantage of a discrete beam steering scheme such as the one presented here is the possibility for distributed sensing
heads for lidar. As the outputs of the electro-optic switch may be easily recollected into fiber, the outputs can be easily routed
to different places around a vehicle, with each fiber terminated by a taper or microlens. This means the 'sensor head' can be
extremely small, unobtrusive and easily replaced. The TX/RX module, and the electro-optic switch may be integrated onto a
daughter board connected to the central processing unit. This is ideally suited to harsh environment sensing, where any
5
external sensor head may be easily damaged, and must be cheaply and easily replaceable. Furthermore the requirement of
only one set of detection electronics for multiple sensors, can reduce the total sensor cost for automotive lidar.
We have demonstrated the utility of high speed electro-optic switches for free-space beam scanning applications by creating a
CONCLUSION
multi-pixel lidar system with a three channel switch. This scheme has several advantages including its speed, modularity,
dynamic addressability, single-mode output, and compatibility with various coherent light modulation techniques. The
approach we describe is scalable up to ~1000 output channels per chip using smaller footprint thin-film ridge waveguides in
lithium niobate [22]. Additionally it could be considered to complement such a switching chip with acousto-optic deflectors
on the waveguide outputs [23], to create a chip with 2D scanning, one direction discrete, and the other continuous.
ACKNOWLEDGEMENTS
The authors thank Stefan Morley for electronics support. BH was supported by the Australian Government Research Training
Program Scholarship. This research is financially supported by the Griffith University Research Infrastructure Programme, the
Australian Research Council Centre of Excellence for Quantum Computation and Communication Technology (CQC2T,
CE170100012), TMS and MAB were supported the Australian Research Council Centre of Excellence for Engineered
Quantum Systems (EQUS, CE170100009), and ML was supported by the Australian Research Council Future Fellowship
(FT180100055). This work was performed in part at the Queensland node of the Australian National Fabrication Facility, a
company established under the National Collaborative Research Infrastructure Strategy to provide nano- and microfabrication
facilities for Australia's researchers.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
REFERENCES
S. T. S. Holmström, U. Baran, and H. Urey, "MEMS laser scanners: A review," J. Microelectromechanical Syst. 23,
259 -- 275 (2014).
Blickfield GmbH, "Blickfield GmbH - Technology," https://www.blickfeld.com/tech-sol.
F. Xiao, W. Hu, and A. Xu, "Optical phased-array beam steering controlled by wavelength," Appl. Opt. 44, 5429 --
5433 (2005).
K. Van Acoleyen, H. Rogier, and R. Baets, "Two-dimensional optical phased array antenna on silicon-on-insulator,"
Opt. Lett. 18, 265 -- 298 (2010).
J. Sun, E. Timurdogan, A. Yaacobi, E. S. Hosseini, and M. R. Watts, "Large-scale nanophotonic phased array,"
Nature 493, 195 -- 199 (2013).
N. W. Carlson, G. A. Evans, R. Amantea, S. L. Palfrey, J. M. Hammer, M. Lurie, L. A. Carr, F. Z. Hawrylo, E. A.
James, C. J. Kaiser, J. B. Kirk, and W. F. Reichert, "Electronic beam steering in monolithic grating-surface-emitting
diode laser arrays," Appl. Phys. Lett. 53, 2275 -- 2277 (1988).
K. Shimura, Z. Ho, M. Nakahama, X. Gu, A. Matsutani, and F. Koyama, "Non-mechanical beam scanner integrated
VCSEL for solid state LiDAR," 2017 Conf. Lasers Electro-Optics Pacific Rim, CLEO-PR 2017 2017-Janua, 1 -- 2
(2017).
C. Hu, J. R. Whinnery, and M. Nabil, "Optical Deflection in Thin-Film Nematic-Liquid-Crystal Waveguides," IEEE
J. Quantum Electron. QE-10, 218 -- 222 (1974).
S. Khan and N. Riza, "Demonstration of 3-dimensional wide angle laser beam scanner using liquid crystals," Opt.
Express 12, 868 -- 882 (2004).
B. Luey, S. R. Davis, S. D. Rommel, D. Gann, J. Gamble, M. Ziemkiewicz, M. Anderson, and R. Paine, "A
Lightweight, Cost-Efficient, Solid-State LiDAR System Utilizing Liquid Crystal Technology for Laser Beam
Steering for Advanced Driver Assistance," 25th Int. Tech. Conf. Enhanc. Saf. Veh. Natl. Highw. Traffic Saf. Adm.
1 -- 9 (2017).
R. A. Meyer, "Optical Beam Steering Using a Multichannel Lithium Tantalate Crystal," Appl. Opt. 11, 613 -- 616
(1972).
12. K. Nakamura, J. Miyazu, M. Sasaura, and K. Fujiura, "Wide-angle, low-voltage electro-optic beam deflection based
on space-charge-controlled mode of electrical conduction in KTa 1-xNbxO3," Appl. Phys. Lett. 89, (2006).
13. D. B. Coyle, D. L. Rabine, D. Poulios, J. B. Blair, P. R. Stysley, R. Kay, G. Clarke, J. Bufton, and N. Goddard,
6
"Fiber Scanning Array for 3 Dimensional Topographic Imaging," 1, 3 -- 4 (2013).
14. X. Gu, T. Shimada, A. Matsutani, and F. Koyama, "Miniature nonmechanical beam deflector based on bragg
reflector waveguide with a number of resolution points larger than 1000," IEEE Photonics J. 4, 1712 -- 1719 (2012).
15. A. Töws and A. Kurtz, "Investigations on the performance of lidar measurements with different pulse shapes using a
multi-channel Doppler lidar system," in 19th Coherent Laser Radar Conference (2018), pp. 1 -- 5.
Baraja Pty Ltd, "Technology - Baraja," https://www.baraja.com/technology/.
16.
17. Weihua Guo, P. R. A. Binetti, C. Althouse, M. L. Masanovic, H. P. M. M. Ambrosius, L. A. Johansson, and L. A.
Coldren, "Two-Dimensional Optical Beam Steering With InP-Based Photonic Integrated Circuits," IEEE J. Sel. Top.
Quantum Electron. 19, 6100212 -- 6100212 (2013).
18. A. Martin, J. Bourderionnet, L. Leviander, J. F. Parsons, M. Silver, and P. Feneyrou, "Coherent Lidar for 3D-imaging
19.
20.
21.
through obscurants," in 19th Coherent Laser Radar Conference (2018), pp. 19 -- 23.
J. L. Jackel and J. J. Johnson, "Reverse exchange method for burying proton exchanged waveguides," Electron. Lett.
27, 1360 (1991).
F. Lenzini, S. Kasture, B. Haylock, and M. Lobino, "Anisotropic model for the fabrication of annealed and reverse
proton exchanged waveguides in congruent lithium niobate," Opt. Express 23, 1748 (2015).
R. Kumar, E. Barrios, A. MacRae, E. Cairns, E. H. Huntington, and A. I. Lvovsky, "Versatile wideband balanced
detector for quantum optical homodyne tomography," Opt. Commun. 285, 5259 -- 5267 (2012).
22. A. Boes, B. Corcoran, L. Chang, J. Bowers, and A. Mitchell, "Status and Potential of Lithium Niobate on Insulator
(LNOI) for Photonic Integrated Circuits," Laser Photon. Rev. 1700256 (2018).
23. D. E. Smalley, Q. Y. J. Smithwick, V. M. Bove, J. Barabas, and S. Jolly, "Anisotropic leaky-mode modulator for
holographic video displays," Nature 498, 313 -- 317 (2013).
7
|
1812.05260 | 2 | 1812 | 2019-06-24T14:49:56 | On the Ferroelectric Polarization Switching of Hafnium Zirconium Oxide in Ferroelectric/Dielectric Stack | [
"physics.app-ph",
"cond-mat.mes-hall",
"cond-mat.mtrl-sci"
] | The ferroelectric polarization switching in ferroelectric hafnium zirconium oxide (Hf0.5Zr0.5O2, HZO) in the HZO/Al2O3 ferroelectric/dielectric stack is investigated systematically by capacitance-voltage and polarization-voltage measurements. The thickness of dielectric layer is found to have a determinant impact on the ferroelectric polarization switching of ferroelectric HZO. A suppression of ferroelectricity is observed with thick dielectric layer. In the gate stacks with thin dielectric layers, a full polarization switching of the ferroelectric layer is found possible by the proposed leakage-current-assist mechanism through the ultrathin dielectric layer. Theoretical simulation results agree well with experimental data. This work clarifies some of the critical parts of the long-standing confusions and debating related to negative capacitance field-effect transistors (NC-FETs) concepts and experiments. | physics.app-ph | physics | On the Ferroelectric Polarization Switching of Hafnium
Zirconium Oxide in Ferroelectric/Dielectric Stack
Mengwei Si, Xiao Lyu, and Peide D. Ye*
School of Electrical and Computer Engineering and Birck Nanotechnology Center, Purdue
University, West Lafayette, Indiana 47907, United States
* Address correspondence to: [email protected] (P.D.Y.)
ABSTRACT
The ferroelectric polarization switching in ferroelectric hafnium zirconium oxide
(Hf0.5Zr0.5O2, HZO) in the HZO/Al2O3 ferroelectric/dielectric stack is investigated systematically
by capacitance-voltage and polarization-voltage measurements. The thickness of dielectric layer
is found to have a determinant impact on the ferroelectric polarization switching of ferroelectric
HZO. A suppression of ferroelectricity is observed with thick dielectric layer. In the gate stacks
with thin dielectric layers, a full polarization switching of the ferroelectric layer is found possible
by the proposed leakage-current-assist mechanism through the ultrathin dielectric layer.
Theoretical simulation results agree well with experimental data. This work clarifies some of the
critical parts of the long-standing confusions and debating related to negative capacitance field-
effect transistors (NC-FETs) concepts and experiments.
KEYWORDS: Ferroelectric, HZO, Fe-FET, Negative capacitance, Steep-slope.
1
Introduction
A ferroelectric material has two stable polarization states with different directions, which
are switchable by the external electric field, and thus is extensively explored for non-volatile
memory applications. Using ferroelectric field-effect transistors (Fe-FETs) as FET-type
ferroelectric memory is a promising ferroelectric memory architecture, because of its high density
and non-destructive readout.1 -- 4 Recently, using a ferroelectric-gated transistor as a negative
capacitance field-effect transistor (NC-FET) has attracted tremendous attention as a novel steep-
slope device.5 -- 9 In both Fe-FET and NC-FET, a ferroelectric (FE) insulator and linear dielectric
(DE) insulator bilayer stack10-17 is applied as the gate structure. The necessity of such a linear DE
layer is because an interfacial oxide layer between semiconductor channel and FE insulator is
required to improve the ferroelectric/semiconductor interface and meanwhile provide sufficient
capacitance matching if quasi-static negative capacitance (QSNC) concept is applied for the
development of NC-FETs.5 QSNC definition was introduced to distinguish the stabilized NC effect
and transient NC effect.18 One important fact, which has been overlooked in the past several years,
is that the FE/DE stack capacitor is fundamentally different from a FE capacitor and DE capacitor
in series.18,19 Therefore, the complete understanding of the impact of DE layer on the ferroelectric
properties of a FE/DE stack is crucial to study the ferroelectric switching mechanism in Fe-FETs
and NC-FETs.
Ferroelectric hafnium oxide, such as hafnium zirconium oxide (Hf0.5Zr0.5O2, HZO), has
been recently discovered as an ultrathin CMOS compatible high performance ferroelectric
insulator.20,21 Therefore, HZO is chosen as the FE insulator for this study and Al2O3 is chosen as
the linear DE insulator to study the ferroelectric polarization switching in the FE/DE stack. As is
well-known that ferroelectric HZO has a thickness-dependent remnant polarization (Pr) at about
2
10-30 μC/cm2,22,23 However, a conventional dielectric insulator cannot support such a large charge
density. For example, Al2O3 has a typical dielectric constant of 8 and a breakdown electric field
less than 1 V/nm.24 The calculated charge density at breakdown electric field is about 7 μC/cm2,
which is the maximum charge density (Qmax) for ideal Al2O3 to be able to support without
breakdown. Note that in reality, this value for Qmax is much smaller. As can be seen that, even in
ideal case, there is a big gap between the remnant polarization in HZO and the maximum charge
density in Al2O3. This fact can also be generally applied to other FE/DE stack systems. Thus, the
puzzle and confusion in the field is how can ferroelectric polarization switching happen in a FE/DE
stack without sufficient charge balance? Such charge difference can only be explained by
introducing the leakage current and interfacial charges. The impact of leakage current in FE/DE
bi-layer was previously reported, and the interfacial charging is believed to be important in the
ferroelectric switching process by a thermodynamic free energy model.10 In FE hafnium oxide
systems, the discussions are mostly focused on the impact of leakage current on the negative
capacitance effect.15-17,25 The hysteresis-free NC effect in HZO/Al2O3 are reported by fast pulse
measurement, the impacts of leakage current and charge trapping are minimized because of the
fast pulses.15-17,26
In this work, we provide a simple understanding on the ferroelectric polarization switching
process in FE/DE stack by introducing the leakage current through the thin DE layer and only
considering the electrostatics. The ferroelectric polarization switching of HZO in the HZO/Al2O3
FE/DE stack is investigated systematically by capacitance-voltage (C-V) and polarization-voltage
(P-V) measurements. The thickness of dielectric layer is found to have determinant impact on the
ferroelectric polarization switching of ferroelectric HZO. The suppression of ferroelectricity is
observed with thick dielectric layer. In the gate stacks with thin dielectric layers, a full polarization
3
switching of the ferroelectric layer is found possible by the proposed leakage-current-assist
mechanism through the ultrathin dielectric layer. It is confirmed that the charge needed for
ferroelectric polarization switching comes from the leakage current through the thin dielectric
layer. Without such leakage current to realize the charge balance, the FE HZO cannot be fully
polarized.
Experimental
Fig. 1 shows the experimental device structures. Four types of capacitor structures are used
in this work: (a) TiN/Al2O3/TiN (type A), (b) TiN/HZO/TiN (type B), (c) TiN/Al2O3/HZO/TiN
(type C), and (d) TiN/Al2O3/TiN/HZO/TiN (type D). The device fabrication process started with
the standard solvent cleaning of heavily p-doped Si substrates (resistivity < 0.005 Ω⋅cm). TiN was
deposited by atomic layer deposition (ALD) at 250 °C, using [(CH3)2N]4Ti (TDMAT, heated up
to 60 °C) and NH3 as the Ti and N precursors, respectively. All TiN layers are metallic and 30 nm
thick. Hf1−xZrxO2 film was deposited by ALD at 200 °C, using [(CH3)2N]4Hf (TDMAHf, heated
up to 60 °C), [(CH3)2N]4Zr (TDMAZr, heated up to 75 °C), and H2O as the Hf, Zr, and O precursors,
respectively. The Hf1−xZrxO2 film with x = 0.5 was achieved by controlling HfO2:ZrO2 cycle ratio
of 1:1. The ALD deposition of TiN and HZO were in two separated ALD chambers to avoid cross-
contamination. The two ALD chambers are connected externally by Ar environment in a glove
box to avoid the environmental contamination. After the deposition of type A-D structures, the
samples were annealed at 500 °C in N2 environment for 1 min by rapid thermal annealing. Then,
Ti/Au top electrodes were fabricated by photo-lithography, e-beam evaporation and lift-off process
(capacitor area=5024 μm2). CF4/Ar dry etching was done to remove top TiN layer for device
isolation for type A-C capacitors. For type D capacitors, BCl3/Ar dry etching was used to remove
the top Al2O3 layer and CF4/Ar dry etching was used to remove the top and middle TiN layers. All
4
electrical measurements were done at room temperature in a cascade summit probe station. C-V
measurement was performed using an Agilent E4980A LCR meter and P-V measurement was
carried out using a Radiant RT66C ferroelectric tester.
Results and Discussion
Fig. 2(a) shows the C-V measurement of a type A capacitor with 20 nm Al2O3, from 1 kHz
to 1 MHz. Fig. 2(b) shows the P-V measurement of the same type A capacitor, showing a linear
dielectric characteristic. Both measurements (small signal C-V, dP/dV in P-V) give consistent
capacitance values for the type A dielectric capacitor with a capacitance of ~0.33 μF/cm2 and a
corresponding dielectric constant of ~8. Fig. 3(a) shows the C-V measurement of a type B
capacitor with 20 nm HZO, from 1 kHz to 1 MHz. The C-V measurement of a type B capacitor
shows signature two peaks in the C-V hysteresis loop as the ferroelectric characteristics. The
different capacitances at different voltages in C-V are attributed to the different dielectric constant
due to the difference in atomic structures in different ferroelectric polarization states. The
corresponding dielectric constants are calculated, as also shown in the right axis. Fig. 3(b) shows
the P-V measurement of the same type B capacitor, showing a ferroelectric hysteresis loop.
Fig. 4(a) shows the C-V measurements of type C capacitors with 20 nm HZO and Al2O3
from 0 nm to 20 nm, measured at 10 kHz. The capacitances of type C capacitors decrease with
thicker Al2O3 as expected. The signature two capacitance peaks due to ferroelectricity in the C-V
hysteresis loop decrease and eventually disappear in 20 nm HZO/20 nm Al2O3 stack, suggesting
the reduction of ferroelectricity in thick DE layer and FE layer stack. This feature is even more
clearly presented in Fig. 4(b) which shows the P-V measurements of type C capacitors with 20 nm
HZO and Al2O3 from 4 nm to 20 nm. The applied voltage ranges are maximized in P-V
measurement before the leakage current has essential impacts. The significant decrease of remnant
5
polarization in P-V hysteresis loops is clearly observed with thicker DE layers. The C-V
measurements and P-V measurements consistently confirm that thick DE layer can suppress the
ferroelectricity in FE/DE stack. Fig. 4(c) shows the P-V characteristics of a FE/DE capacitor with
20 nm HZO and 6 nm Al2O3, measured at different voltage sweep ranges. The coercive voltage
and remnant polarization are found to be dependent on the sweep voltage range.
To further understand the physics behind the DE layer thickness dependence on the
ferroelectricity of FE/DE stack, a theoretical analysis is provided, as shown in Fig. 5. To
understand the dynamic process of ferroelectric switching, this process is plotted by a two-step
process: before ferroelectric polarization switching and after ferroelectric polarization switching.
As is well-known, the ferroelectric polarization switching is atom re-position within the unit cell,
so it is always slower than the electron re-distribution. Thus, the two-step assumption is valid. For
simplicity, it is assumed the FE layer has a dielectric constant of ϵFE (without considering
ferroelectric polarization) and a thickness of tFE; the DE layer has a dielectric constant of ϵDE and
a thickness of tDE. It also assumes that the FE layer has equal number of polarization up states and
polarization down states in the virgin state before the measurement so the net polarization is zero,
as in Fig. 5(a). This situation is similar to two high-k dielectric stack. We define VTOT to be the
voltage applied to the FE/DE stack, VDE to be the voltage across the DE layer and VFE to be the
voltage across the FE layer. Therefore, before the ferroelectric polarization switching, the voltages
across the DE layer and FE layer are
𝑽𝑫𝑬,𝒊𝒏𝒊𝒕 =
𝑽𝑻𝑶𝑻𝝐𝑭𝑬𝒕𝑫𝑬
𝝐𝑫𝑬𝒕𝑭𝑬+𝝐𝑭𝑬𝒕𝑫𝑬
𝑽𝑭𝑬,𝒊𝒏𝒊𝒕 =
𝑽𝑻𝑶𝑻𝝐𝑫𝑬𝒕𝑭𝑬
𝝐𝑫𝑬𝒕𝑭𝑬+𝝐𝑭𝑬𝒕𝑫𝑬
(1)
(2)
There are totally three different cases according to the different DE thickness. Here, we
first assume the leakage current is zero and then discuss the impact of leakage current. Firstly, if
6
tDE is very thick, then VFE,init can be sufficiently small so that it is smaller than the coercive voltage
(Vc) of the FE layer, according to eqn. (2). Thus, no polarization switching can happen. So, the C-
V and P-E characteristics behave like a linear dielectric insulator. Secondly, if VFE,init > Vc but tDE
is sufficiently thick, the FE layer cannot be fully polarized. As the polarization switching happens,
VDE increases until VFE reaches Vc and the polarization process cannot continue. In this case, the
total charge in FE layer (QFE) can be approximated as ϵDE(VTOT-Vc)/tDE, where we have
ferroelectric polarization charge (PFE) < Pr and VDE,final=QFE/(ϵDE/tDE). Note that PFE≅QFE=QDE if
the PFE is significantly larger than the dielectric charge. Such assumption is made for the simplicity
of qualitative discussion and does not affect the conclusion. The numerical simulation including
the difference of PFE and QFE gives the same conclusion. Thirdly, if the DE layer is thin enough,
so that the second criterion does not meet anymore, we can have the FE layer fully polarized. So
VDE can be estimated as VDE,final=Pr/(ϵDE/tDE). It is clear that if QFE is larger than the maximum
charge density in DE layer, VDE,final will be larger than the breakdown voltage (VBD) of the DE
layer, which of course cannot happen. What is really happening in this process (if VDE,final > VBD)
is when VDE rises from VDE,init to VDE,final, the DE layer first becomes leaky and these leakage
charges will balance the ferroelectric polarization charges so that VDE cannot reach VBD. Thus, all
the ferroelectric polarization charges are balanced by the charges from leakage current instead of
the charge in DE layer. Therefore, in thin DE limit, the ferroelectric polarization switching process
is a leakage-current-assist process. At the extremely leaky limit, it becomes almost as metal-FE-
metal structure. Here's a summary of all three cases,
Case 1: thick DE limit, no polarization switching
𝑽𝑭𝑬,𝒊𝒏𝒊𝒕 < 𝑽𝒄
(3)
Case 2: moderate DE thickness, partial switching (by dielectric charge or leakage)
7
𝑽𝑭𝑬,𝒊𝒏𝒊𝒕 > 𝑽𝒄 𝒂𝒏𝒅 𝑷𝒓 > 𝝐𝑫𝑬(𝑽𝑻𝑶𝑻 − 𝑽𝒄)/𝒕𝑫𝑬
(4)
𝑷𝑭𝑬 = 𝝐𝑫𝑬(𝑽𝑻𝑶𝑻 − 𝑽𝒄)/𝒕𝑫𝑬
𝑽𝑫𝑬,𝒇𝒊𝒏𝒂𝒍 = 𝑷𝑭𝑬/
𝝐𝑫𝑬
𝒕𝑫𝑬
Case 3: ultra-thin DE limit, leakage-current-assist switching
𝑷𝒓 < 𝝐𝑫𝑬(𝑽𝑻𝑶𝑻 − 𝑽𝒄)/𝒕𝑫𝑬
𝑷𝑭𝑬 = 𝑷𝒓
𝑽𝑫𝑬,𝒇𝒊𝒏𝒂𝒍 = 𝑷𝒓/
𝝐𝑫𝑬
𝒕𝑫𝑬
(5)
(6)
(7)
(8)
(9)
To calculate the leakage-current-assistant switching process as in case 3, a theoretical
model is developed. Fig. 5 shows the model of FE/DE stack and the charge distribution upon the
application of a positive external voltage (VTOT). If there is no leakage current, charge in the DE
layer (QDE) is always the same as the charge in the FE layer (QFE). But as discussed above, this no
leakage current assumption is not valid since QFE can be much larger than the maximum QDE at
DE breakdown. Therefore, the leakage current through the DE layer is unavoidable. Here, Eeffect is
defined as a critical electric field. For simplicity, Eeffect is assumed to be a constant without
thickness dependence. There is negligible leakage current below the Eeffect and above the Eeffect the
leakage current exists. The charge carried by the leakage current will be trapped at the FE/DE
interface as Qit. In the equilibrium condition, there is no charge transfer process with zero current.
As a result, the electric field in the DE layer will be pinned at the Eeffect, so that,
𝑸𝑫𝑬 = 𝝐𝑫𝑬𝑬𝒆𝒇𝒇𝒆𝒄𝒕
The charge balance equation becomes
𝑸𝑭𝑬 = 𝑸𝑫𝑬 + 𝑸𝒊𝒕
(10)
(11)
8
As we can see, it is critical to have enough Qit from leakage current to obtain a high QFE. So, the
polarization switching process must be leakage-assist-switching. Eqns. (10) and (11) are the key
formulas in the simulation of P-V hysteresis loop of FE/DE stack.
Fig. 6(a) shows the simulation of the minor loops of the FE HZO capacitor based on
numerical fitting to the experimental P-E curve. If there is not enough charge from Qit and QDE,
the FE layer always exhibits a minor loop with less Pr. Fig. 6(b) shows the simulation of the P-V
hysteresis loops in a 6 nm Al2O3/20 nm HZO capacitor at different voltage sweep ranges, assuming
no leakage current. Fig. 6(c) shows the simulation of the same structure but using the leakage-
assist-switching model with leakage current from DE layer. It is obvious that a significantly larger
Pr is obtained. The experimental P-V hysteresis loops in a 6 nm Al2O3/20 nm HZO capacitor (Fig.
4(c)) with certain level of unavoidable leakage current match well with the leakage-assist-
switching model presented in Fig. 6(c).
The thickness of DE layer, thus the leakage current, has a significant impact on the
ferroelectricity of FE/DE stack as studied experimentally. The leakage-assist-switching model can
also simulate the thickness-dependent behavior. Fig. 7(a) shows the simulation of P-V hysteresis
loop of FE/DE stack with 20 nm HZO and different Al2O3 thicknesses, assuming no leakage
current. Fig. 7(b) shows the simulation of the same structure but using the leakage-assist-switching
model. The specific voltage sweep ranges are selected according the experimental results, as
shown in Fig. 4(b). Modeling and experiments are in great agreement. It also successfully predicts
the loss of Pr in FE/DE stack with thick DE layer, because of the voltage division and no leakage
current. Fig. 7(c) shows the Pr versus thickness comparison in w/o leakage model and w/ leakage
model with different Eeffect, suggesting FE/DE stack with a leakier DE layer can have larger Pr. Fig.
7(d) compares the experimental results on Pr vs. Al2O3 thickness with simulation results, showing
9
the leakage-assist-switching model matches well with experimental results in terms of thickness
dependence. The above results provide new insights to understand the large amount of reported
experimental results in NC-FETs usually with FE/DE stacks. FE and FE/DE are fundamentally
different in terms of coercive field, Pr, switch speed, and many others.
The impact of internal metal is studied by comparing type C and type D capacitors (see
supplementary section 1). The FE/DE stacks with internal metal and without internal metal are
physically very different. If the internal metal gate becomes the externally connected metal wires
or the internal metal gate is physically connected to measurement equipment, the required balanced
charges can be provided even externally. All these facts are extremely important to understand and
interpret the experimental observation related to Fe-FETs and NC-FETs. The so-called interfacial
coupling effect or capacitance enhancement was observed in the previous reports19,27-30. This
interfacial coupling effect can improve the equivalent oxide thickness of FE/DE gate stack, which
is also observed in the HZO/Al2O3 FE/DE stack as shown in supplementary section 2. The above
understanding of ferroelectric switching process helps to just apply ferroelectric polarization
switching (sometimes calls transient negative capacitance effect) to explain the DC enhancement
of ferroelectric-gated transistors without invoking QSNC concept (see supplementary section 3).
The operation speed of such transistors could be eventually limited by the ferroelectric switching
speed31, which needs to be thoroughly investigated still.
Conclusion
In summary, the ferroelectric polarization switching in FE HZO in the HZO/Al2O3 FE/DE
stack is investigated systematically by C-V and P-V measurements. The thickness of dielectric
layer is found to have determinant impact on the ferroelectric polarization switching of HZO. The
suppression of ferroelectricity is observed with a thick linear dielectric layer. In the gate stacks
10
with thin dielectric layers, a full polarization switching of the ferroelectric layer is found possible
by the proposed leakage-current-assist mechanism through the ultrathin dielectric layer. The
numerical simulation using the leakage-assist-switching model matches very well with the
experimental results.
11
ASSOCIATED CONTENT
Supporting Information
Additional details for the impact of internal metal, interfacial coupling and the DC enhancement
of Fe-FETs are in the supporting information.
AUTHOR INFORMATION
Corresponding Author
*E-mail: [email protected]
Author Contributions
P.D.Y. conceived the idea of FE/DE stack and supervised the experiments. M.S. and X.L. did the
ALD deposition and device fabrication. M.S. and X.L. performed DC electrical measurements and
analysis. M.S. did the numerical simulation. P.D.Y. and M.S. proposed the idea of DC
enhancement on Fe-FETs. M.S. and P.D.Y. co-wrote the manuscript and all authors commented
on it.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGEMENTS
The authors would like to thank Muhammad Ashraful Alam and Suman Datta for valuable
discussions. The work was supported in part by the Semiconductor Research Corporation (SRC)
and DARPA.
REFERENCES
12
(1)
Ishiwara, H. Ferroelectric Random Access Memories. J. Nanosci. Nanotechnol. 2012, 12,
7619 -- 7627.
(2) Müller, J.; Böscke, T. S.; Schröder, U.; Hoffmann, R.; Mikolajick, T.; Frey, L. Nanosecond
Polarization Switching and Long Retention in a Novel MFIS-FET Based on Ferroelectric
HfO2. IEEE Electron Device Lett. 2012, 33, 185 -- 187.
(3) Yoo, H. K.; Kim, J. S.; Zhu, Z.; Choi, Y. S.; Yoon, A.; MacDonald, M. R.; Lei, X.; Lee, T.
Y.; Lee, D.; Chae, S. C.; et al. Engineering of Ferroelectric Switching Speed in Si Doped
HfO2 for High-Speed 1T-FERAM Application. In IEEE Intl. Electron Devices Meet.; 2017;
pp 481 -- 484.
(4) Dünkel, S.; Trentzsch, M.; Richter, R.; Moll, P.; Fuchs, C.; Gehring, O.; Majer, M.; Wittek,
S.; Müller, B.; Melde, T.; et al. A Fe-FET Based Super-Low-Power Ultra-Fast Embedded
NVM Technology for 22nm FDSOI and Beyond. In IEEE Intl. Electron Devices Meet.;
2017; pp 485 -- 488.
(5) Salahuddin, S.; Datta, S. Use of Negative Capacitance to Provide Voltage Amplification for
Low Power Nanoscale Devices. Nano Lett. 2008, 8, 405 -- 410.
(6) Li, K.-S.; Chen, P.-G.; Lai, T.-Y.; Lin, C.-H.; Cheng, C.-C.; Chen, C.-C.; Wei, Y.-J.; Hou,
Y.-F.; Liao, M.-H.; Lee, M.-H.; et al. Sub-60mV-Swing Negative-Capacitance FinFET
without Hysteresis. In IEEE Intl. Electron Devices Meet.; 2015; pp 620 -- 623.
(7) Lee, M. H.; Fan, S.-T.; Tang, C.-H.; Chen, P.-G.; Chou, Y.-C.; Chen, H.-H.; Kuo, J.-Y.;
Xie, M.-J.; Liu, S.-N.; Liao, M.-H.; et al. Physical Thickness 1.x Nm Ferroelectric HfZrOx
Negative Capacitance FETs. In IEEE Intl. Electron Devices Meet.; 2016; pp 306 -- 309.
(8) Si, M.; Su, C.-J.; Jiang, C.; Conrad, N. J.; Zhou, H.; Maize, K. D.; Qiu, G.; Wu, C.-T.;
Shakouri, A.; Alam, M. A.; et al. Steep-Slope Hysteresis-Free Negative Capacitance MoS2
13
Transistors. Nat. Nanotechnol. 2018, 13, 24 -- 29.
(9) Si, M.; Jiang, C.; Su, C.; Tang, Y.; Yang, L.; Chung, W.; Alam, M. A.; Ye, P. D. Sub-60
mV/Dec Ferroelectric HZO MoS2 Negative Capacitance Field-Effect Transistor with
Internal Metal Gate: The Role of Parasitic Capacitance. In IEEE Intl. Electron Devices
Meet.; 2017; pp 573 -- 576.
(10) Kim, Y. J.; Park, M. H.; Jeon, W.; Kim, H. J.; Moon, T.; Lee, Y. H.; Kim, K. D.; Hyun, S.
D.; Hwang, C. S. Interfacial Charge-Induced Polarization Switching in Al2O3/Pb(Zr,Ti)O3
Bi-Layer. J. Appl. Phys. 2015, 118, 224105.
(11) Kim, Y. J.; Yamada, H.; Moon, T.; Kwon, Y. J.; An, C. H.; Kim, H. J.; Kim, K. D.; Lee, Y.
H.; Hyun, S. D.; Park, M. H.; et al. Time-Dependent Negative Capacitance Effects in
Al2O3/BaTiO3 Bilayers. Nano Lett. 2016, 16, 4375 -- 4381.
(12) Kim, Y. J.; Park, M. H.; Lee, Y. H.; Kim, H. J.; Jeon, W.; Moon, T.; Kim, K. D.; Jeong, D.
S.; Yamada, H.; Hwang, C. S. Frustration of Negative Capacitance in Al2O3/BaTiO3 Bilayer
Structure. Sci. Rep. 2016, 6, 19039.
(13) Chang, S.-C.; Avci, U. E.; Nikonov, D. E.; Young, I. A. A Thermodynamic Perspective of
Negative-Capacitance Field-Effect-Transistors. IEEE J. Explor. Solid-State Comput.
Devices Circuits 2017, 3, 56 -- 64.
(14) Hoffman, M.; Pesic, M.; Slesazeck, S; Schroeder, U.; Mikolajick, T. On the Stabilization
of Ferroelectric Negative Capacitance in Nanoscale Devices. Nanoscale 2018, 10, 10891.
(15) Hoffmann, M.; Max, B.; Mittmann, T.; Schroeder, U.; Slesazeck, S.; Mikolajick, T.
Demonstration of High-speed Hysteresis-free Negative Capacitance in Ferroelectric
Hf0.5Zr0.5O2. In IEEE Intl. Electron Devices Meet.; 2018; pp 727 -- 730.
(16) Hoffmann, M.; Fengler, F. P. G.; Herzig, M.; Mittmann, T.; Max, B.; Schroeder, U.; Negrea,
14
R.; Lucian, P.; Slesazeck, S.; Mikolajick, T. Unveiling the Double-Well Energy Landscape
in a Ferroelectric Layer. Nature 2019, 565, 464 -- 467.
(17) Kim, K. D.; Kim, Y. J.; Park, M. H.; Park, H. W.; Kwon, Y. J.; Lee, Y. B.; Kim, H. J.;
Moon, T.; Lee, Y. H.; Hyun, S. D.; et al. Transient Negative Capacitance Effect in Atomic-
Layer-Deposited Al2O3/Hf0.3Zr0.7O2 Bilayer Thin Film. Adv. Funct. Mater. 2019, 1808228.
(18) Liu, Z.; Bhuiyan, M. A.; Ma, T. P. A Critical Examination of 'Quasi-Static Negative
Capacitance' (QSNC) Theory. In IEEE Intl. Electron Devices Meet.; 2018; pp 711 -- 714.
(19) Sun, F. C.; Kesim, M. T.; Espinal, Y.; Alpay, S. P. Are Ferroelectric Multilayers Capacitors
in Series? J. Mater. Sci. 2016, 51, 499 -- 505.
(20) Böescke, T. S.; Müller, J.; Bräuhaus, D.; Schröder, U.; Böttger, U. Ferroelectricity in
Hafnium Oxide: CMOS Compatible Ferroelectric Field Effect Transistors. In IEEE Intl.
Electron Devices Meet.; 2011; pp 547 -- 550.
(21) Muller, J.; Boscke, T. S.; Schroder, U.; Mueller, S.; Brauhaus, D.; Bottger, U.; Frey, L.;
Mikolajick, T. Ferroelectricity in Simple Binary ZrO2 and HfO2. Nano Lett. 2012, 12, 4318 --
4323.
(22) Park, M. H.; Kim, H. J.; Kim, Y. J.; Lee, W.; Moon, T.; Hwang, C. S. Evolution of Phases
and Ferroelectric Properties of Thin Hf0.5Zr0.5O2 films According to the Thickness and
Annealing Temperature. Appl. Phys. Lett. 2013, 102, 242905.
(23) Tian, X.; Shibayama, S.; Nishimura, T.; Yajima, T.; Migita, S.; Toriumi, A. Evolution of
Ferroelectric HfO2 in Ultrathin Region down to 3 nm. Appl. Phys. Lett. 2018, 112, 102902.
(24) Lin, H. C.; Ye, P. D.; Wilk, G. D. Leakage Current and Breakdown Electric-Field Studies
on Ultrathin Atomic-Layer-Deposited Al2O3 on GaAs. Appl. Phys. Lett. 2005, 87, 182904.
(25) Khan, A. I.; Radhakrishna, U.; Chatterjee, K.; Salahuddin, S.; Antoniadis, D. A. Negative
15
capacitance behavior in a leaky ferroelectric. IEEE Transactions on Electron Devices 2016,
63, 4416.
(26) Alam, M. A.; Si, M.; Ye, P. D A Critical Review of Recent Progress on Negative
Capacitance Field-Effect Transistors. Appl. Phys. Lett. 2019, 114, 090401.
(27) Tsang, C. H.; Chew, K.-H.; Ishibashi, Y.; Shin, F. G. Structure of Interfaces in Layered
Ferroelectrics of First and/or Second Order Transition. J. Phys. Soc. Jpn. 2004, 73, 3158 --
3165.
(28) Dawber, M.; Lichtensteiger, C.; Cantoni, M.; Veithen, M.; Ghosez, P.; Johnston, K.; Rabe,
K. M.; Triscone, J.-M. Unusual Behavior of the Ferroelectric Polarization in PbTiO3/SrTiO3
Superlattices. Phys. Rev. Lett. 2005, 95, 177601.
(29) Zhou, Y. Enhancement of Dielectric and Ferroelectric Properties in Ferroelectric
Superlattices. Solid State Commun. 2010, 150, 1382 -- 1385.
(30) Salev, P.; Mahayni, A.; Grigoriev, A. Polarization Coupling Transition
in
BaTiO3/PbZr0.2Ti0.8O3 Ferroelectric Bilayers. Phys. Rev. B 2016, 93, 041423.
(31) Chung, W.; Si, M.; Shrestha, P. R.; Campbell, J. P.; Cheung, K. P.; Ye, P. D. First Direct
Experimental Studies of Hf0.5Zr0.5O2 Ferroelectric Polarization Switching Down to 100-
picosecond in Sub-60mV/dec Germanium Ferroelectric Nanowire FETs. In VLSI Tech.
Dig.; 2018; pp 89 -- 90.
16
Figures
Figure 1. Capacitor structures used in this work (a) Al2O3 only (type A), (b) HZO only (type B),
(c) Al2O3 and HZO stack without internal metal (type C), and (d) Al2O3 and HZO stack with TiN
layer as the internal metal (type D).
Figure 2. (a) C-V measurement and (b) P-V measurement on a type A capacitor with 20 nm Al2O3.
17
(a)(b)HZOP+ SiTiNAl2O3Ti/AuTiNHZOP+ SiTiNAl2O3Ti/AuTiNTiN(c)(d)P+ SiTiNAl2O3Ti/AuTiNP+ SiTiNTi/AuTiNHZOType B Type A Type C Type D (a)(b)-8-6-4-202468-3-2-10123 Data Fitting P (C/cm2)Voltage (V)dP/dV=0.33 F/cm2-10-8-6-4-202468100.000.050.100.150.200.250.300.35k ~ 8 1 kHz 10 kHz 100 kHz 1 MHz30 nm TiN/20 nm Al2O3/30 nm TiN C (F/cm2)Voltage (V)
Figure 3. (a) C-V measurement and (b) P-V measurement of a type B capacitor with 20 nm HZO.
18
-8-6-4-20246810-20-1001020 P (C/cm2)Voltage (V)-10-8-6-4-202468100.40.60.81.0 1 kHz 10 kHz 100 kHz 1 MHz30 nm TiN/20 nm HZO/30 nm TiN C (F/cm2)Voltage (V)101520Dielectric Constant(a)(b)-10-505101520250.00.20.40.60.81.010 kHz 0 nm Al2O3 4 nm Al2O3 6 nm Al2O3 8 nm Al2O3 10 nm Al2O3 20 nm Al2O3 C (F/cm2)Voltage (V)30 nm TiN/x nm Al2O3/20 nm HZO/30 nm TiN-20-15-10-505101520-15-10-505101530 nm TiN/x nm Al2O3/20 nm HZO/30 nm TiN 4 nm Al2O3 6 nm Al2O3 8 nm Al2O3 10 nm Al2O3 20 nm Al2O3 P (C/cm2)Voltage (V)(a)(b)-15-10-5051015-20-15-10-5051015206 nm Al2O3/20 nm HZO P (C/cm2)Voltage (V)(c)
Figure 4. (a) C-V measurements and (b) P-V measurements on type C capacitors with different
Al2O3 thickness and 20 nm HZO. (c) P-V measurements on a type C capacitor with 6 nm Al2O3
and 20 nm HZO at different voltage sweep ranges.
Figure 5. Illustration of charge distribution in FE/DE stack in the following conditions. (a) VTOT=0
and FE layer is not polarized. (b) VTOT larger than positive coercive voltage Vc+ before polarization
19
FEDEVTOT< Vc-PFEDEVTOT< Vc-Qit(e)(f)LeakageFEPFEVDE,finalw/o LeakageQitw/ LeakageFEDEVTOT> Vc+PFEFEDEVTOT> Vc+PFE(c)(d)LeakageVDE,initVirgin StateBefore SwitchingFEDEVTOT=0VTOT> Vc+(a)(b)FEDEAfter SwitchingAfter Switchingw/o Leakagew/ LeakageAfter SwitchingAfter Switching
switching. VTOT larger than positive coercive voltage Vc+ after polarization switching (c) assuming
no leakage current though DE layer and (d) assuming the existence of leakage current though DE
layer. VTOT less than negative coercive voltage Vc- after polarization switching (e) assuming no
leakage current though DE layer and (f) assuming the existence of leakage current though DE layer.
Type of charges: black, ferroelectric polarization; blue, dielectric polarization; red, interfacial
trapped charge.
Figure 6. (a) Simulation of minor loops in a FE HZO capacitor. (b) Simulation w/o leakage current
of P-V hysteresis loops in a 6 nm Al2O3/20 nm HZO capacitor at different voltage sweep ranges.
20
-0.4-0.20.00.20.4-20-1001020 Forward Reverse Experiment P (C/cm2)E (V/nm)Minor P-E Loop-15-10-505101520-20-15-10-505101520w/o Leakage P (C/cm2)Voltage (V)6 nm Al2O3/20 nm HZO-15-10-5051015-20-15-10-505101520w/ Leakage P (C/cm2)Voltage (V)6 nm Al2O3/20 nm HZO(a)(b)(c) (c) Simulation w/ leakage current of P-V hysteresis loops in a 6 nm Al2O3/20 nm HZO capacitor
at different voltage sweep ranges.
Figure 7. Simulation on (a) w/o leakage current and (b) w/ leakage current of P-V hysteresis loops
of FE/DE capacitors with 20 nm HZO and different Al2O3 thicknesses. (c) Remnant polarization
versus Al2O3 thickness on both w/o leakage current and w/ leakage current assumptions at different
Eeffect. (d) Comparison of experimental results on Pr vs. Al2O3 thickness with simulation results,
w/ and w/o leakage-assist-switching.
21
-20-100102030-20-1001020Al2O3 Thicknessw/o Leakage 4 nm 6 nm 8 nm 10 nm 20 nm P (C/cm2)Voltage (V)-20-100102030-20-1001020w/ LeakageAl2O3 Thickness 4 nm 6 nm 8 nm 10 nm 20 nm P (C/cm2)Voltage (V)510152005101520No LeakageSymbol: Eeffect 0.7-1.5 V/nm Line: No Leakage Pr (C/cm2)Al2O3 Thickness (nm)510152002468101214 Experiment w/ Leakage w/o Leakage Pr (C/cm2)Al2O3 Thickness (nm)(a)(b)(c)(d) TOC
22
-20-15-10-505101520-15-10-5051015x nm Al2O3/20 nm HZO/ 4 nm 6 nm 8 nm 10 nm 20 nm P (C/cm2)Voltage (V)FEMetalDEMetal Supplementary Information for:
On the Ferroelectric Polarization Switching of Hafnium
Zirconium Oxide in Ferroelectric/Dielectric Stack
Mengwei Si, Xiao Lyu, and Peide D. Ye*
School of Electrical and Computer Engineering and Birck Nanotechnology Center, Purdue
University, West Lafayette, Indiana 47907, United States
* Address correspondence to: [email protected] (P.D.Y.)
1
1. Ferroelectric/Dielectric Stack with Internal Metal
Figure S1. (a) C-V measurement and (b) P-V measurement on a type D capacitor with 20 nm Al2O3 and 20 nm HZO.
(c) C-V measurements comparison of a type C and a type D capacitor with 20 nm Al2O3 and 20 nm HZO.
Fig. S1(a) shows the C-V measurement of a type D capacitor with 20 nm HZO, 20 nm
Al2O3 and 30nm TiN in between, measured from 1 kHz to 1 MHz. The two signature capacitance
peaks in the C-V hysteresis loop are observed. Fig. S1(b) shows the P-V measurement of a type D
capacitor with 20 nm HZO and 20 nm Al2O3, where a weak ferroelectric hysteresis loop is achieved.
Fig. S1(c) shows the comparison of C-V measurements of type C and type D capacitors with 20
nm HZO and 20 nm Al2O3. Although same thicknesses of HZO and Al2O3 are used, a type D
capacitor with 30 nm TiN exhibits weak ferroelectricity in C-V and P-V characteristics, suggesting
2
-15-10-50510150.200.220.240.26 1 kHz 10 kHz 100 kHz 1 MHz30 nm TiN/20 nm Al2O3/30 nm TiN/20 nm HZO/30 nm TiN C (F/cm2)Voltage (V)-15-10-5051015-4-2024 P (C/cm2)Voltage (V)-15-10-50510150.200.220.240.260.280.3020 nm Al2O3 + 20 nm HZO10 kHz w/o Internal TiN w Internal TiN C (F/cm2)Voltage (V)(a)(b)(c)
that the charge in the internal metal can assist the ferroelectric switching process, in great contrast
to the result from a type C capacitor. Fig. S1(c) concludes the FE/DE stacks with internal metal
and without internal metal are physically very different. If the internal metal gate becomes much
larger than the capacitor area by design or it is externally connected to metal wires through vias or
the internal metal gate is physically connected to the measurement equipment, the required
balanced charges can be provided even externally. All these facts are extremely important to
understand and interpret the experimental observation related to Fe-FETs and NC-FETs.
3
2. Interfacial Coupling in Ferroelectric/Dielectric Stack
Figure S2. (a) Capacitance-1 of type C capacitor versus Al2O3 thickness, by both experimental measurements (red
squares) and calculated total capacitance (blue triangles). The calculated total capacitance in series is based on the
experimental measurements of type A and type B capacitors, using the maximum capacitance in the measured C-V
curves. The capacitance values of measured type A capacitors with different Al2O3 thicknesses are presented as black
circles. (b) The ratio of experimental capacitances of type C capacitor over capacitance in series versus different Al2O3
thickness.
Fig. S2(a) shows the capacitance-1 versus Al2O3 thickness characteristics of three types of
capacitors, Al2O3 only (type A), Al2O3/20 nm HZO stack (type C) and the capacitance value of
measured Al2O3 (type A) and HZO (type B) capacitors in series. Experimentally, capacitance of
type C capacitor is lower than type A capacitor with same Al2O3 thickness. No obvious QSNC
effect is observed in HZO material system. But a capacitance enhancement of type C capacitor is
observed to be larger than the capacitance value in series, as shown in Fig. S2(b). Over 20%
capacitance enhancement is observed with 8 and 10 nm Al2O3/20 nm HZO stack. This result shows
from another aspect that the FE/DE stack with internal metal and without internal metal are
physically very different. It demonstrates the existence of interfacial coupling1-6 between the Al2O3
layer and HZO layer. This interfacial coupling effect can improve the equivalent oxide thickness
4
481216201.051.101.151.201.25x nm Al2O3 + 20 nm HZOCFE+DE/(C-1FE+C-1DE)-1Al2O3 Thickness (nm)(a)(b)51015200123410 kHz Al2O3 only x nm Al2O3 + 20 nm HZO Capacitance in series C-1 (cm2/F)Al2O3 Thickness (nm)
of FE/DE gate stack. The static capacitance enhancement by negative capacitance effect (CTOT >
CDE) is not directly achieved in this slow measurement. The intrinsic quasi-static negative
capacitance phenomenon might be masked by the charge trapping and de-trapping.10 So it is not
conclusive to claim the existence of negative capacitance or not in this work using HZO as the
ferroelectric stack.
5
3. DC Enhancement in Fe-FET
.
Figure S3. (a) Illustration of DC enhancement of a ferroelectric-gated FET.
It is clear that ferroelectric polarization switching can lead to the sub-60 mV/dec
subthreshold slope (SS) in ferroelectric-gated transistors. But hysteresis in transfer characteristics
is unavoidable, if not considering charge trapping process. Note that the concept of transient NC
effect in Fe-FETs7-9 is fundamentally different from the concept of QSNC effect in NC-FETs. But
it is unclear whether performance benefit is achievable or not with a hysteretic and sub-60 mV/dec
device. Here, the authors want to emphasize that ferroelectric polarization switching and
polarization charge in Fe-FETs can offer DC enhancement (IOFF reduction and ION enhancement
simultaneously), with the existence of manageable hysteresis and without incorporating QSNC
explanation. All transient effects of ferroelectric dynamic polarization switching10 -- 15 are negligible
in DC condition discussed in this work. Meanwhile, the hysteresis might not have serious impact
in logic circuits if it is controlled in between zero voltage and half of the supply voltage (VDD).
6
``` VupVdownNo polarizationPolarized after negative biasPolarized after positive biasIONenhancementIOFFsupressionIDVG
Thus, this work addresses an important fact that ferroelectric field-effect transistors can offer DC
enhancement from the perspective of ferroelectric polarization switching only. The potential of
using ferroelectric-gated transistor for low-power logic applications is limited by the speed of the
devices.
In a CMOS logic circuit, a lower off-state current (IOFF) and higher on-state current (ION)
is preferred. The DC enhancement here is defined as at the same IOFF and a given supply voltage
(VDD), the transistor can have higher ION. Whether a small hysteresis exists or not is not important
if lower IOFF and higher ION can be achieved simultaneously. For circuit applications, hysteresis
window of the devices should be controlled less than half of the VDD. As shown in Fig. S3, the
black line is the transfer characteristics of the baseline FET without ferroelectric polarization. If a
high gate voltage is applied, the transfer curve shifts to the left as the green curve. If a low gate
voltage is applied, the transfer curve shifts to the right as the blue curve. The amount of threshold
voltage shift (ΔVT) is determined by the remnant polarization, the capacitance of dielectric layer
(CDE) and the ratio of ferroelectric capacitor area (AFE) and the dielectric capacitor area (ADE)
(assuming the existence of an internal metal layer). Note that the conclusion is still valid in FE/DE
stack without internal metal, but the area ratio of AFE and ADE becomes one. The transfer
characteristics of the Fe-FET switches between the polarization up and polarization down transfer
curves and the switching voltages (Vup, Vdown) are determined by the coercive voltages. The
coercive voltages can be tuned by the thickness of the FE layer, so Vup and Vdown can be tuned
accordingly. Therefore, if we plot the full bi-directional transfer characteristics, as the red line in
Fig. S3, a reduction in IOFF and an enhancement in ION are achieved simultaneously. This exactly
shows the DC enhancement can be achieved using a Fe-FET structure. The difficulty in realization
of such performance is that the Pr in conventional ferroelectric insulator material is so high that
7
the hysteresis window become too large for logic applications. However, by using a DE layer for
capacitance matching and using an internal metal gate to modulate the area ratio of AFE and ADE
if it is needed, we can effectively reduce the hysteresis window, achieve DC enhancement in Fe-
FET.
Such experimental structure and experimental results were already reported in our previous
publication with AFE/ADE ~ 100, as shown in Ref. 16. It is a MoS2 ferroelectric-gate FET with
internal metal gate structure. Subthreshold slope (SS) of 37.6 mV/dec in forward sweep and SS of
42.2 mV/dec in reverse sweep are achieved. More importantly, a clear ION enhancement is achieved
with same IOFF so that this is an obvious DC enhancement. From the perspective of ferroelectric
polarization switching, such DC enhancement can be explained without invoking QSNC concept.
REFERENCES
(1) Sun, F. C.; Kesim, M. T.; Espinal, Y.; Alpay, S. P. Are Ferroelectric Multilayers Capacitors
in Series? J. Mater. Sci. 2016, 51, 499 -- 505.
(2) Tsang, C. H.; Chew, K.-H.; Ishibashi, Y.; Shin, F. G. Structure of Interfaces in Layered
Ferroelectrics of First and/or Second Order Transition. J. Phys. Soc. Jpn. 2004, 73, 3158 --
3165.
(3) Dawber, M.; Lichtensteiger, C.; Cantoni, M.; Veithen, M.; Ghosez, P.; Johnston, K.; Rabe,
K. M.; Triscone, J.-M. Unusual Behavior of the Ferroelectric Polarization in PbTiO3/SrTiO3
Superlattices. Phys. Rev. Lett. 2005, 95, 177601.
(4) Zhou, Y. Enhancement of Dielectric and Ferroelectric Properties in Ferroelectric
Superlattices. Solid State Commun. 2010, 150, 1382 -- 1385.
8
(5) Salev, P.; Mahayni, A.; Grigoriev, A. Polarization Coupling Transition
in
BaTiO3/PbZr0.2Ti0.8O3 Ferroelectric Bilayers. Phys. Rev. B 2016, 93, 041423.
(6) Hoffman, M.; Pesic, M.; Slesazeck, S; Schroeder, U.; Mikolajick, T. On the stabilization of
ferroelectric negative capacitance in nanoscale devices. Nanoscale 2018, 10, 10891.
(7) Khan, A. I.; Bhowmik, D.; Yu, P.; Kim, S. J.; Pan, X.; Ramesh, R.; Salahuddin, S.
Experimental Evidence of Ferroelectric Negative Capacitance
in Nanoscale
Heterostructures. Appl. Phys. Lett. 2011, 99, 113501.
(8) Appleby, D. J. R.; Ponon, N. K.; Kwa, K. S. K.; Zou, B.; Petrov, P. K.; Wang, T.; Alford,
N. M.; O'Neill, A. Experimental Observation of Negative Capacitance in Ferroelectrics at
Room Temperature. Nano Lett. 2014, 14, 3864 -- 3868.
(9) Gao, W.; Khan, A.; Marti, X.; Nelson, C.; Serrao, C.; Ravichandran, J.; Ramesh, R.;
Salahuddin, S. Room-Temperature Negative Capacitance in a Ferroelectric-Dielectric
Superlattice Heterostructure. Nano Lett. 2014, 14, 5814 -- 5819.
(10) Hoffmann, M.; Max, B.; Mittmann, T.; Schroeder, U.; Slesazeck, S.; Mikolajick, T.
Demonstration of High-speed Hysteresis-free Negative Capacitance in Ferroelectric
Hf0.5Zr0.5O2. In IEEE Intl. Electron Devices Meet.; 2018; pp 727 -- 730.
(11) Van Houdt, J.; Roussel, P. Physical Model for the Steep Subthreshold Slope in Ferroelectric
FETs. IEEE Electron Device Lett. 2018, 39, 877 -- 880.
(12) Kittl, J. A.; Obradovic, B.; Reddy, D.; Rakshit, T.; Hatcher, R. M.; Rodder, M. S. On the
Validity and Applicability of Models of Negative Capacitance and Implications for MOS
Applications. Appl. Phys. Lett. 2018, 113, 042904.
(13) Kim, Y. J.; Park, H. W.; Hyun, S. D.; Kim, H. J.; Kim, K. D.; Lee, Y. H.; Moon, T.; Lee,
Y. B.; Park, M. H.; Hwang, C. S. Voltage Drop in a Ferroelectric Single Layer Capacitor
9
by Retarded Domain Nucleation. Nano Lett. 2017, 17, 7796 -- 7802.
(14) Saha, A. K.; Datta, S.; Gupta, S. K. "Negative Capacitance" in Resistor-Ferroelectric and
Ferroelectric-Dielectric Networks: Apparent or Intrinsic? J. Appl. Phys. 2018, 123, 105102.
(15) Obradovic, B.; Rakshit, T.; Hatcher, R.; Kittl, J. A.; Rodder, M. S. Ferroelectric Switching
Delay as Cause of Negative Capacitance and the Implications to NCFETs. In VLSI Tech.
Dig.; 2018; pp 51 -- 52.
(16) Si, M.; Jiang, C.; Su, C.; Tang, Y.; Yang, L.; Chung, W.; Alam, M. A.; Ye, P. D. Sub-60
mV/Dec Ferroelectric HZO MoS2 Negative Capacitance Field-Effect Transistor with
Internal Metal Gate: The Role of Parasitic Capacitance. In IEEE Intl. Electron Devices
Meet.; 2017; pp 573 -- 576.
10
|
1710.00677 | 2 | 1710 | 2018-01-16T00:40:46 | Effects of optical attenuation, heat diffusion and acoustic coherence in photoacoustic signals produced by nanoparticles | [
"physics.app-ph",
"physics.med-ph",
"physics.optics"
] | Behavior of the photoacoustic signal produced by nanoparticles as a function of their concentration was studied in detail. As the concentration of nanoparticles is increased in a sample, the peak-to-peak photoacoustic amplitude increases linearly up to a certain value, after which an asymptotic saturated behavior is observed. To elucidate the mechanisms responsible for these observations, we evaluate the effects of nanoparticles concentration, the optical attenuation and the effects of heat propagation from nano-sources to their surroundings. We found that the saturation effect of the photoacoustic signal as a function the concentration of nanoparticles is explained by a combination of two different mechanisms. As has been suggested previously, but not modeled correctly, the most important mechanism is attributed to optical attenuation. The second mechanism is due to an interference destructive process attributed to the superimposition of the photoacoustic amplitudes generated for each nanoparticle, this explanation is reinforced through our experimental and simulations results; based on this, it is found that the linear behavior of the photoacoustic amplitude could be restricted to optical densities $\le0.5$. | physics.app-ph | physics | Effects of optical attenuation, heat diffusion and acoustic coherence in
photoacoustic signals produced by nanoparticles
J. E. Alba -- Rosales,1 G. Ramos -- Ortiz,2 L. F. Escamilla -- Herrera,3 B. Reyes -- Ram´ırez,2 L. Polo -- Parada,4, a) and G.
Guti´errez -- Ju´arez1, b)
1)Divisi´on de Ciencias e Ingenier´ıas, Universidad de Guanajuato. Le´on, Gto.,
M´exico.
2)Centro de Investigaciones en ´Optica. Le´on, Gto., M´exico.
3)Instituto de Ciencias Nucleares, UNAM, CDMX, M´exico.
4)Dalton Cardiovascular Research Center, Department of Medical Pharmacology and Physiology,
University of Missouri-Columbia, Columbia, MO, USA.
(Dated: 7 November 2018)
Behavior of the photoacoustic signal produced by nanoparticles as a function of their concentration was
studied in detail. As the concentration of nanoparticles is increased in a sample, the peak-to-peak photoa-
coustic amplitude increases linearly up to a certain value, after which an asymptotic saturated behavior is
observed. To elucidate the mechanisms responsible for these observations, we evaluate the effects of nanopar-
ticles concentration, the optical attenuation and the effects of heat propagation from nano-sources to their
surroundings. We found that the saturation effect of the photoacoustic signal as a function the concentration
of nanoparticles is explained by a combination of two different mechanisms. As has been suggested previously,
but not modeled correctly, the most important mechanism is attributed to optical attenuation. The second
mechanism is due to an interference destructive process attributed to the superimposition of the photoacous-
tic amplitudes generated for each nanoparticle, this explanation is reinforced through our experimental and
simulations results; based on this, it is found that the linear behavior of the photoacoustic amplitude could
be restricted to optical densities ≤ 0.5.
PACS numbers: 78.20.Hp
Keywords: Photoacoustics, Nanoparticles, Interference, Saturation
In recent years, pulsed laser-induced ultrasound (US),
better known as the Photoacoustic (PA) effect, has had
a major resurgence because its wide range of applica-
tions, mainly in the biological and medical areas1, for
instance, PA imaging2 and as monitoring method in
thermo-therapy of cancer3. Further the analogies be-
tween optical and acoustic phenomena, led to advance-
ments in confocal PA microscopy4, creations of new
methodologies to detect US5 and generation of new ma-
terials to achieve thermal and/or acoustic contrasts6. PA
effect is produced by the absorption of CW modulated
pulsed optical radiation by a medium. This absorption
raises non-radiative decays that increase the temperature
and causes mechanical waves typically in the range of US.
The major advantages for PA techniques are their sensi-
bility to distinguish different optical contrast and the US
penetration in the tissue7.
Nowadays metallic nanoparticles (NPs) play an im-
portant role as enhancers of the PA signal, the design
and application of these materials is subject to the type
of applications desired, as well as to the available laser
source.1,8,9.
Previous reports have shown that the PA amplitude
is not always proportional to materials concentration;
for example, the first reports of PA saturation are de-
a)Electronic mail: [email protected]
b)Electronic mail: [email protected]
scribed in the work of J. W. Pin10 and J. Wang11; in
the later, authors employ a continuous absorption model
to explain the saturation in red ink, but such model de-
pends of multiple parameters (without physical mean-
ing) adjustment to correctly fit the experimental data.
Notwithstanding this deficiency, this reference has been
the explanation for the PA saturation obtained in dyes,
pigments and tissue. On the other hand, despite the dif-
ferent physical properties exhibit by NPs, a similar satu-
ration effect is observed. In several cases a linear depen-
dence between amplitude and concentration has been ob-
tained; for example, gold (Au) nano-vesicles3, Au nano-
rods9, Au bio-conjugated nano-spheres12,13, silica-coated
Au nano-rods14, Au nano-carbon-tubes and Fe2O3 nano-
spheres15; while, in many other examples a saturated be-
havior is observed; Au nano-spheres16, Au nano-cages17
and Au nano-beacons18. The origin of this discrepancy
has been never analyzed in detail and there are few re-
ports where the PA response from a discretized system
in micro scale was obtained19 -- 21.
Also, the PA signal shape at low concentrations has
been observed that is symmetric whereas at high con-
centrations became asymmetric; this behavior has been
extensively reported in the literature4,13,19,21 -- 28. There-
fore, the goal of this study is to explain the causes of
these concentration-dependent effects. Herein it is pro-
posed that the key characteristics of the nonlinear satu-
rated behavior in the peak-to-peak (P-P) PA amplitude
and its asymmetry at high concentrations are explained
8
1
0
2
n
a
J
6
1
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
2
v
7
7
6
0
0
.
0
1
7
1
:
v
i
X
r
a
by means of a photo-thermo-acoustic model29. Sigrist22
previously reported a model that explained the PA signal
produced by a continuous media (liquids); in this paper,
the model is extended to a discrete case from continuous
media to NPs. The NPs are represented by single point
that absorbs the incident radiation instantaneously30.
With this model a simulation code was developed tak-
ing in account the heat source size, the light attenu-
ated by the NPs and the coherence of a single PA signal
source29,31. The light losses, by absorption and scatter-
ing, can be described by the well-known Beer-Lambert
(BL) law, which corresponds to an exponential decay in
the amount of light. We hypothesize that the asymmet-
ric shape of the PA signals is due to the interference of
the individual PA signals generated by the NPs. To verify
this, it is theoretically estimated the spatial region where
interference of two PA waves occurs. These results are
of great importance in applications where the NPs are
used as a PA enhancer, since they allow to identify a dy-
namic range for PA amplitude generation and an optimal
contrast agent concentration at which the maximum PA
signal contrast is possible.
For this, PA experiments were conducted with 5, 10
and 100 nm Au NPs (spheres, nanoComposix) in 2 mM
sodium citrate dihydrate aqueous solution (Detailed in-
formation about the samples characteristics can be found
in supplementary information). The suspensions opti-
cal density (OD)32,33 were measured as a function of the
NPs concentration for each aliquot by using an UV-VIS
spectrometer (Lambda 900 UV/VIS/NIR, Pekin Elmer).
The experimental setup for the detection of PA signals is
shown in Figure 1. The second harmonic from a Nd:YAG
pulsed laser (Brilliant, Quantel) was employed to provide
532 nm light with a pulse duration of 10 ns and a repe-
tition rate of 10 Hz. This beam was focused into the PA
cell using a couple of lenses such that a large Rayleigh
waits length was obtained to be approximately constant
inside the cell (0.7 mm of diameter). The energy per
pulse was set at 1 mJ (±5%, SD). The laser beam was
set perpendicular to the transducer 2 mm away. The
laser energy was monitored using a thermopile (1917-
R, Newport) coupled to a power meter (818P-030-19,
Newport). PA signals were detected with a homemade
transducer (20 MHz central response and a bandwidth
of 20 MHz)34 and displayed by a 200-MHz oscilloscope
(TDS5104B, Tektronix, Wilsonville, OR) triggered by a
photo-diode (DET10A; Thorlabs, Newton, NJ) with a 1-
nanoseconds rise time. The signals were amplified with a
gain of 25 dB via 500 MHz amplifier (ZFL-500LN-BNC+,
Mini-Circuits). NPs suspensions were diluted from stock
concentration (100%) in steps of 10% using a sodium cit-
rate aqueous solution at 2 M (6028, Karal).
Figure 2(a) shows the PA signals generated by the NPs
suspensions at stock concentration, these are 4 × 1013
for 5 nm, 5 × 1012 for 10 nm and 5 × 109 for 100 nm.
The signals generated by the NPs are observed at 1.3 µs.
For lower NPs concentrations, the PA signals exhibited
similar shape, but smaller amplitudes. Figures 2 (b) to
2
FIG. 1. Experimental setup.PD: photo-diode, AM: amplifier,
PM: power meter, TP: thermopile, ND: neutral density filters,
BS1: beam spliter 10:90, BS2: beam spliter 10:90, L1: plane-
convex lens with f = +300 mm, L2: plane-concave lens with
f = −50 mm, US-T: ultra sound transducer. The distance
between L1 and L2 is 250 mm.
2 (d) show the respective normalized P-P PA amplitudes
as a function of the NPs concentration for each NP size.
FIG. 2. (a) Measured PA signals obtained from different NPs
samples at stock concentration. (b)-(d) Normalized PA am-
plitude as a function of concentration for each NPs of 100 nm,
10 nm and 5 nm, respectively.
Here, it is assumed that the measured pressure p(R, t)
in r generated by a single nanoparticle (NP) at r(cid:48) is de-
scribed by 22,30,35,36:
(cid:34)
(cid:32)
(cid:33)2(cid:35)
p(R, t) = p0
ct − R
R
exp
−
ct − R
d0/2
;
(1)
with p0 = 4E0βc2/π3/2Cpd0
3 and R = r− r(cid:48). Here E0
β, c, and Cp, are the the energy per pulse, the thermal
expansion coefficient, the sound speed medium propaga-
tion and the heat capacity at constant pressure of the
fluid sample, respectively. In Sigrist paper, d0 was de-
fined as the spatial illumination profile of a Gaussian
beam22. However, we associate this parameter with the
thermal size of the object. This hypothesis is justified
from the assumption that a NP only can absorb radia-
tion, due to plasmonic effect8, meanwhile the surround-
ing fluid medium (water) does not. According to this, a
minimum value for d0 is the NP diameter (dNP); and as
maximum the quantity (dNP + dth) where:
dth = 4(χwτl)
1
2 .
(2)
Equation (2) is related the thermal diffusion length37;
for this expression χw is the water diffusivity (0.143 ×
106m2/s) and τl is the laser pulse (FWHM of 10 ns);
therefore, dth = 150 nm. This analytical approxima-
tion is equivalent to solve the coupling heat and pressure
equations considering the laser time profile.
A code was written in the
software Wolfram
MathematicaTM to emulate the experiments performed.
Considering the equation (1) and the optical attenuation,
the numerical expression employed was:
n,m(cid:88)
psim(Dij, t) =
10−ζj∆z × p(Dij, t);
(3)
i=1,j=1
with Dij = s − rij. Here corresponds to the extinction
coefficient and ζ is the sample concentration, which were
measured for each NPs aliquot by UV-VIS spectroscopy.
Likewise , s corresponds to the sensor position. Counters
i and j are used to label each NP and the cylinder sec-
tion, respectively, the length of the sections is ∆z. The
complete methodology details used to perform the simu-
lations can be found in the supplemental material.
In Figures 3a to 3c simulated PA signals for stock con-
centration and d0 = dNP are presented. A statistical
study shows that all signals are symmetric with well de-
fined maximum and minimum peaks, which always ap-
pear near the center temporal range. To obtain the P-P
PA amplitude it was sought the higher and lower peaks
values for each individual simulation. Same behavior was
observed when the BL law is considered, nonetheless, the
maximum amplitude is diminished, as expected, being
60% less when compared to the case without attenua-
tion.
In Figures 3d-3f, a comparison between the experimen-
tal and simulated P-P PA amplitudes as a function of
the NPs concentration is shown; all data were normal-
ized to the respective maximum amplitude. Simulations
for d0 = dNP without BL law predict completely the ex-
perimental trend, but when the optical attenuation is
considered it fails; contrary to the expected results, the
inclusion of BL law in the model did not predict the ex-
periments. Trying to understand this discrepancy, the
numerical Fourier transform can be performed to the sim-
ulated signals, when doing this a broad spectra are pre-
dicted with high central frequencies for d0 = dNP. The
frequencies are at 100 GHz for the 5 nm and 10 nm sam-
ples and at 10 GHz for 100 nm; however, this is not in
3
FIG. 3. (a)-(c) Simulated PA signals for d0 = dNP at stock
concentration for NPs of 5 nm, 10 nm and 100 nm respec-
tively; the gray lines are the simulations without BL law and
the black ones with BL. (d)-(f) Corresponding comparison
between simulated and experimental P-P PA amplitudes as
a function of concentration, squares for experimental data,
circles for simulations without BL law and triangles for sim-
ulations with BL law. All data is normalized.
agreement with the actual spectral response of the sen-
sors that we used in our experiments, then the d0 = dNP
assumption must be discarded.
In Figures 4a to 4c, simulations supposing d0 =
dNP + dth are shown for the stock concentration of NPs.
There are two aspects that must be highlighted: First,
amplitude-shape for 5 and 10 nm samples are well-defined
for all simulated concentrations; however, for 100 nm NPs
it remains noisy, but its shape is more defined than for
the case d0 = dNP. Second, opposite to the above case,
PA signals are asymmetric for all samples, being like the
experimental results and the reported literature. Consid-
ering BL law still decreases the PA amplitude approxi-
mately at 40% of the non-attenuated value. Comparison
between experimental and simulated P-P PA amplitudes
as a function of the NPs concentration is presented in
Figures 4d to 4f. The simulated PA signals without BL
law are linear. When light attenuation is considered, be-
havior of experimental signals is well predicted; as in the
previous cases, the spectra for the simulated signals can
be calculated numerically, the obtained frequencies now
are between 1 and 100 MHz. For both sets of simulations,
optical attenuation, through BL law, defines the ampli-
tude of pressure wave and its effect is only to decrease
the total PA amplitude.
To better understand the consequences for choosing
a d0 value, three aspects must be considered. First,
from the PA power spectrum of a single NP it is found
√
that the maximum frequency value occurs at νmax =
c/
2π(d0/2). From this value can be calculated the spa-
other, it corresponds to λ ≡ √
tial region where pulse of one NP can interact with each
2πd0. Second, from the
4
uals PA pulses, considering the heat diffusion from the
NP volume to their surroundings, gives a PA signal with
high signal-to-noise ratio, asymmetric shape peaks and a
linear behavior in the P-P PA amplitude as a function of
the NPs number. When the optical attenuation is taking
into account the simulated PA amplitudes loss its linear
dependence with the NPs concentration and the signal
saturates, predicting properly the experimental results.
FIG. 4. (a)-(c) Simulated PA signals for d0 = dNP + dth at
stock concentration for NPs of 5 nm, 10 nm and 100 nm re-
spectively; the gray lines are the simulations without BL law
and the black ones with BL. (d)-(f) Corresponding compari-
son between simulated and experimental P-P PA amplitudes
as a function of concentration, squares for experimental data,
circles for simulations without BL law and triangles for sim-
ulations with BL law. All data is normalized.
specific volume of the NPs suspension, an average dis-
tance between NPs L, can be determined. Third, equa-
tion (1) is proportional to the time derivative of a Gaus-
sian function, which has a bipolar temporal profile i.e., it
is compose of a compression and rarefaction cycle. When
summation over two individual PA signals is performed
at the measuring point, there are three extreme possible
situations for the time delay (or acoustical path differ-
ence ∆l), namely: (i) it is equal to zero, then the PA
pulses match exactly and only constructive interference
appears.
(ii) it is equal to λ/2c; then, the rarefaction
of one pulse corresponds exactly to the compression of
other pulse, and therefore partial destructive interference
is produced. (iii) It is greater than λ/c, so they cannot
superimpose. These cases are displayed in Figures 5a to
5c. Using the above information, the ratios L/λ where
calculated for all d0 values and are show in Figure 5d.
When d0 = dNP, L/λ (cid:29) 1 for all NPs diameters and all
NPs concentrations, thus a high number of NPs cannot
be superimposed; therefore, the sum of the individual
signals, at the measurement point during a time inter-
val, looks noisy and symmetric. For d0 = dNP + dth,
we can see in Figure 5d that L/λ < 2 for dNP = 5 nm
and dNP = 10 nm, respectively; now summation over
individual signals produces well defined shape and asym-
metric PA signals with linear behavior of the P-P PA
amplitudes as a function of the NPs number. This is be-
cause at the measurement point, in the time interval, the
superposition of individual pressure waves occurs. For
d0 = 100 nm+dth the corresponding ratio is in the range
5 < L/λ < 12.5, then the interference is more probable
than the case d0 = 100 nm, but less when d0 = dNP + dth
for 5 nm and 10 nm. The consequence for adding individ-
FIG. 5.
(a)-(c) Present the tree extreme possible configu-
ration between the sources, totally constructive interference,
partial destructive interference and when the signals can be
superposed, respectively.
(d) Ratios L/λ, for all d0 values
used in the simulations.
It is important to remark that, if the results are dis-
played as a function of the NPs OD instead of their con-
centrations, the nonlinear behavior of the P-P PA ampli-
tudes is given for OD≥ 0.5. A meticulous review of the
references3,9,11 -- 21 is in concordance with this threshold
(see supplementary data).
In summary, when heat propagates beyond the indi-
vidual NPs volume and the optical attenuation of the
sample is ignored, the P-P PA amplitude as a function of
NPs concentration is linear; this extended PA source can
improve the interference between the single US pulses.
The simulations showed that asymmetric shape of the
PA signal is obtained under this condition. When the
heat is confined inside the NP dimensions, symmetric
signals are obtained and a nonlinear P-P PA amplitude;
the NP thermal confinement can be discarded through
frequency spectrum too. The saturated behavior of P-P
PA amplitude for the extended thermal source is cor-
rectly explained when the optical attenuation is consid-
ered. Finally, our simulations and experimental results
showed that no linear behavior appears for an OD≥ 0.5.
This threshold was well-matched with previous experi-
mental reports. This value can be taken as a point of
departure to obtain linear PA amplitudes as a function
of the concentration for NPs samples.
Acknowledgments. The computation for this work
was performed on the high-performance computing in-
frastructure provided by Research Computing Support
Services and in part by the National Science Foundation
under grant number CNS-1429294 at the University of
Missouri, Columbia Mo. The experimental part of this
work was financed with the 2nd edition of UG-CIO re-
search grants and CONACyT (grant number 5215708).
Authors thank to Martin Olmos and Enrique Noe Arias
for their support in this research and CONACyT.
1W. Li and X. Chen, "Gold nanoparticles for photoacoustic imag-
ing," Nanomedicine 10(2), 299 -- 320 (2015).
2T. Chaigne, J. Gateau, M. Allain, O. Katz, S. Gigan, A. Sen-
tenac, and E. Bossy, "Super-resolution photoacoustic fluctua-
tion imaging with multiple speckle illumination," Optica 3, 54 -- 57
(2016).
3P. Huang and et al, "Biodegradable gold nanovesicles with an
ultrastrong plasmonic coupling effect for photoacoustic imaging
and photothermal therapy," Angewandte Chemie International
Edition 52, 13958 -- 13964 (2013).
4L. Wang, C. Zhang, and L. V. Wang, "Grueneisen relaxation
photoacoustic microscopy," Phys. Rev. Lett. 113, 174301 (2014).
5C. Chao, S. Ashkenazi, S. Huang, M. O'Donnell, and L. Guo,
"High-frequency ultrasound sensors using polymer microring res-
onators," IEEE Transactions on Ultrasonics, Ferroelectrics, and
Frequency Control 54, 957 -- 964 (2007).
6F. Gao, X. Feng, and Y. Zheng, "Advanced photoacoustic and
thermoacoustic sensing and imaging beyond pulsed absorption
contrast," Journal of Optics 18, 074006 (2016).
7L. Wang, Photoacoustic Imaging and Spectroscopy (CRC Press,
Taylor & Francis Group, UK, 2009) p. xiii.
8T. H. D. M. W. Yujun Zhong, Shyamala Devi Malagari, "Review
of mid-infrared plasmonic materials," Journal of Nanophotonics
9, 9 -- 9 -- 21 (2015).
9C. L. Bayer, S. Y. Nam, Y.-S. Chen, and S. Y. Emelianov, "Pho-
toacoustic signal amplification through plasmonic nanoparticle
aggregation," Journal of Biomedical Optics 18, 16001 (2013).
10J. W. Lin and L. P. Dudek, "Signal saturation effect and analyt-
ical techniques in photoacoustic spectroscopy of solids," Analyt-
ical Chemistry 51, 1627 -- 1632 (1979).
11J. Wang, T. Liu, S. Jiao, R. Chen, Q. Zhou, K. K. Shung, L. V.
Wang, and H. F. Zhang, "Saturation effect in functional photoa-
coustic imaging," Journal of Biomedical Optics 15, 021317 1 -- 5
(2010).
12S. Mallidi, T. Larson, J. Tam, P. Joshi, A. Karpiouk, K. Sokolov,
and S. Emelianov, "Multiwavelength photoacoustic imaging and
plasmon resonance coupling of gold nanoparticles for selective
detection of cancer," Nano Letters 9, 2825 -- 2831 (2009).
13D. L. Chamberland, A. Agarwal, N. Kotov, J. B. Fowlkes, P. L.
and X. Wang, "Photoacoustic tomography of joints
Carson,
aided by an etanercept-conjugated gold nanoparticle contrast
agentan ex vivo preliminary rat study," Nanotechnology 19,
095101 (2008).
14J. V. Jokerst, M. Thangaraj, P. J. Kempen, R. Sin-
clair,
and S. S. Gambhir, "Photoacoustic imaging of mes-
enchymal stem cells in living mice via silica-coated gold
nanorods," ACS Nano 6, 5920 -- 5930 (2012), pMID: 22681633,
http://dx.doi.org/10.1021/nn302042y.
15E. I. Galanzha, E. V. Shashkov, T. Kelly, J. W. Kim, L. Yang,
and V. P. Zharov, "In vivo magnetic enrichment and multiplex
photoacoustic detection of circulating tumour cells," Nat Nan-
otechnol. 4, 855 -- 860 (2009).
16M. E. Khosroshahi and A. Mandelis, "Combined photoacous-
tic ultrasound and beam deflection signal monitoring of gold
nanoparticle agglomerate concentrations in tissue phantoms us-
ing a pulsed nd:yag laser," International Journal of Thermo-
physics 36, 880 -- 890 (2015).
17K. H. Song, C. Kim, C. M. Cobley, Y. Xia,
and L. V.
Wang, "Near-infrared gold nanocages as a new class of trac-
ers for photoacoustic sentinel lymph node mapping on a rat
model," Nano Letters 9, 183 -- 188 (2009), pMID: 19072058,
http://dx.doi.org/10.1021/nl802746w.
18D. Pan, M. Pramanik, A. Senpan, S. Ghosh, S. A. Wickline, L. V.
Wang, and G. M. Lanza, "Near infrared photoacoustic detection
5
of sentinel lymph nodes with gold nanobeacons," Biomaterials
31, 4088 -- 4093 (2010).
19A. B. Karpiouk, S. R. Aglyamov, S. Mallidi, J. Shah, W. G. Scott,
J. M. Rubin, and S. Y. Emelianov, "Combined ultrasound and
photoacoustic imaging to detect and stage deep vein thrombosis:
phantom and ex vivo studies." Journal of Biomedical Optics 13,
054061 (2008).
20R. K. Saha and M. C. Kolios, "A simulation study on
photoacoustic signals from red blood cells," The Journal of
the Acoustical Society of America 129, 2935 -- 2943 (2011),
http://dx.doi.org/10.1121/1.3570946.
21R. P. Solano, F. I. Ramirez-Perez, J. A. Castorena-Gonzalez,
E. A. Anell, G. Gutirrez-Jurez, and L. Polo-Parada, "An experi-
mental and theoretical approach to the study of the photoacous-
tic signal produced by cancer cells," AIP Advances 2, 011102
(2012), http://dx.doi.org/10.1063/1.3697852.
22M. W. Sigrist and F. K. Kneubhl, "Lasergenerated stress waves
in liquids," The Journal of the Acoustical Society of America 64,
1652 -- 1663 (1978), http://dx.doi.org/10.1121/1.382132.
23A. J. Dixon, S. Hu, A. L. Klibanov, and J. A. Hossack, "Oscilla-
tory dynamics and in vivo photoacoustic imaging performance of
plasmonic nanoparticle-coated microbubbles," Small 11, 3066 --
3077 (2015).
24A. Feis, C. Gellini, P. R. Salvi, and M. Becucci, "Photoacoustic
excitation profiles of gold nanoparticles," Photoacoustics 2, 47 --
53 (2014).
25D. Pan, X. Cai, C. Yalaz, A. Senpan, K. Omanakuttan, S. Wick-
line, L. Wang, and G. Lanza, "Photoacoustic sentinel lymph
node imaging with self-assembled copper neodecanoate nanopar-
ticles," ACS Nano 6, 1260 -- 1267 (2012).
26L. Puxiang, W. Lidai, W. T. Jian, and V. W. Lihong, "Photoa-
coustically guided wavefront shaping for enhanced optical focus-
ing in scattering media," Nat. Photon. 9, 126132 (2015).
27S. Y. Emelianov, P.-C. Li, and O. M., "Photoacoustics for molec-
ular imaging and therapy," Phys. Today 11, 34 -- 9 (2009).
28R. Zhang, D. Pan, X. Cai, X. Yang, A. Senpan, J. S. Allen, G. M.
Lanza, and L. V. Wang, "ανβ3-targeted copper nanoparticles
incorporating an sn 2 lipase-labile fumagillin prodrug for pho-
toacoustic neovascular imaging and treatment," Theranostics 5,
124 -- 133 (2015).
29Y. Tian, H. Tian, Y. L. Wu, L. L. Zhu, L. Q. Tao, W. Zhang,
Y. Shu, D. Xie, Y. Yang, Z. Y. Wei, X. H. Lu, T.-L. Ren, C.-
K. Shih, and J. Zhao, "Coherent generation of photo-thermo-
acoustic wave from graphene sheets," Scientific Reports 5, 1 -- 8
(2015).
30X. Chen, Y. Chen, M. Yan,
ond photothermal
ACS Nano
http://dx.doi.org/10.1021/nn2050032.
effects
2550 -- 2557
6,
and M. Qiu, "Nanosec-
in plasmonic nanostructures,"
22356648,
(2012),
pMID:
31T. Lee, Q. Li, and L. J. Guo, "Out-coupling of longitudinal pho-
toacoustic pulses by mitigating the phase cancellation," Scientific
Reports 6, 1 -- 9 (2016).
32J. M. Weber, Handbook of optical materials (CRC Press, 2003).
33C. D. Mobley, "The optical properties of water," in Handbook of
Optics: Fundamentals, techniques and design, Vol. 1, edited by
M. Bass (Optical Society of America) 2nd ed.
34B. Reyes-Ram´ırez, C. Garc´ıa-Segundo,
and A. Garc´ıa-
Valenzuela, "An examination of polyvinylidene fluoride capac-
itive sensors as ultrasound transducer for imaging applications,"
Measurement Science and Technology 25, 055109 (2014).
35P. M. Morse and K. U. Ingard, Theoretical Acoustics (McGraw-
Hill, New York, 1968) pp. 281 -- 283.
36B. Wu, C. Frez,
and G. J. Diebold, "Photoacous-
tic transients produced by laser generated, ultrahigh ther-
mal gradients," Applied Physics Letters 103, 124105 (2013),
http://dx.doi.org/10.1063/1.4821739.
37H. Carslaw and J. Jaeger, Conduction of Heat in Solids (Oxford
University Press, UK, 1959) p. 257.
|
1704.07362 | 1 | 1704 | 2017-04-21T18:16:23 | The Effects of Rolling Deformation and Annealing Treatment on Damping Capacity of 1200 Aluminium Alloy | [
"physics.app-ph",
"cond-mat.mtrl-sci"
] | Annealing treatment is an important step of rolling deformation that contributes to microstructural evolution and leads to the significant changes in damping capacity. Damping capacities were analyzed in the parallel to rolling direction at 1 and 10 Hz respectively. It was found that severe plastic deformation at 40 percent reduction has lower damping capacity compared to that of 30 percent and 20 percent reductions respectively. The microstructural results show that the grains of as rolled alloys were changed to almost equiaxed structures after a rolling reduction at 40 percent reduction. | physics.app-ph | physics | Please cite this article as: Mazlee, M. N., Jamaludin, S. B., Yasmin, Y., Shamsudin, S. R., Risby, M.
S., & Afendi, M. (2015). The Effects of Rolling Deformation and Annealing Treatment on Damping
Capacity of 1200 Aluminium Alloy. In Materials Science Forum (Vol. 819, pp. 20-24). Trans Tech
Publications.
The Effects of Rolling Deformation and Annealing Treatment
on Damping Capacity of 1200 Aluminium Alloy
*M. N. Mazlee1,a, J. B. Shamsul1,b, Y. Yasmin2,c ,
S. R. Shamsudin2,d, M. S. Risby3,e, M. Afendi4,f
1Sustainable Engineering Cluster, School of Materials Engineering,
Universiti Malaysia Perlis, 01000 Kangar, Perlis, MALAYSIA
2School of Materials Engineering, Universiti Malaysia Perlis,
02600 Arau, Perlis, MALAYSIA
3Faculty of Engineering, Universiti Pertahanan Nasional Malaysia (UPNM)
Kem Sg. Besi, 57000 Kuala Lumpur, MALAYSIA
4School of Mechatronics Engineering, Universiti Malaysia Perlis,
Pauh, 02600 Arau, Perlis, MALAYSIA
[email protected], [email protected], [email protected],
[email protected], [email protected], [email protected]
Keywords: Rolling deformation, damping capacity, annealing treatment, 1200 aluminium alloy.
Abstract. Annealing treatment is an important step of rolling deformation that contributes to
microstructural evolution and leads to the significant changes in damping capacity. Damping
capacities were analyzed in the parallel to rolling direction at 1 and 10 Hz respectively. It was found
that severe plastic deformation at 40 percent reduction has lower damping capacity compared to that
of 30 percent and 20 percent reductions respectively. The microstructural results show that the
grains of as rolled alloys were changed to almost equiaxed structures after a rolling reduction at 40
percent reduction.
Introduction
Progress in technology and industry is based on developments in materials and the related
heat treatment processes involved. The damping capacity of a material is determined by evaluating
the energy dissipated in the material during mechanical vibration. High damping materials, which
have the ability to dissipate mechanical vibration energy, are valuable to be applied in the fields of
noise control and in stabilizing structures in order to suppress mechanical vibrations and attenuate
wave propagation [1-2]. Practical applications need low density materials that simultaneously
exhibit high damping capacity and good ductility. However, in metals these properties are usually
incompatible because of the dependence on microscopic mechanisms involved in strengthening and
damping [3].
The compatibility of high damping capacity with high strength has been considered to be
important for structural damping capacity of the aluminum severely deformed by materials
subjected to resonance loading. However, increases in damping by various methods have been
accompanied by decreases in strength [4]. It is established that severe plastic deformation is viable
to produce high strength metals with ultrafine grained microstructure [5, 6].
The severe plastic deformation is reliable also as a process to produce high damping
materials since the severely formed metals contain significant amount of lattice defects which give
rise to damping capacity. Zheng Ming Yi et al. have reported the high damping capacity of Mg-Cu-
Mn alloy severely deformed by equal channel angular press (ECAP) [7]. On the other hand, cold
rolling is the one of the severe plastic deformation processes applicable to continuous production of
large bulky materials [8, 9].
The increase of the damping capacity of the aluminium also can be achieved by the
application of precipitation hardening treatment [10, 11] and superheating treatment [12]. Choong
Do Lee reported that the precipitation hardening of coherent Mg2Si on T6 treatment in Al-7Si-
0.3Mg alloy play a fundamental role in the simultaneous enhancement of mechanical property and
damping capacity [10]. The purpose of this research is to study the effects of rolling deformation
and annealing treatment on damping capacity of 1200 aluminium alloy.
Experimental Procedure
The raw material used was as-received 1200 aluminium alloys in sheet form with 1.3 mm
thickness. The samples were cut into 70 mm length x 12 mm width dimensions for homogenization
treatment at 560°C for 4 hours in a normal atmosphere and then cooled in the furnace to room
temperature. Subsequently, the samples were undergone cold rolling process by using cold rolling
machine to produce 20, 30 and 40 percent reductions respectively. Then, the samples were annealed
at two different temperatures of 345°C and 400°C for 1 and 3 hours soaking times respectively.
A dynamic mechanical analyzer (Pyris Diamond DMA model, USA) was used to measure
the damping capacity. Dynamic mechanical analysis was carried out in the three point bending
mode using a dual cantilever system. The samples were prepared in the form of rectangular bars
with dimensions of 50 x 10 x 1.0 mm. The tested specimens were run at 5°C/minute heating rate
from 30 to 400ºC with 100 µm strain at 10 Hz vibration frequency in a flowing purified nitrogen
gas.
The microstructures after homogenization process, rolling reduction and annealing process
were analyzed by using optical microscope. Specimens were prepared by the standard
metallography methods of cutting and mounting followed by wet grinding on a series of SiC papers.
Finally, the specimens were polished with 6 µm, 3 µm and 1 µm diamond suspension using napless
cloth. The etchant used was 1.0% HF in order to reveal the microstructures.
Results and Discussion
i) Damping Capacity
Figures 1, 2, 3 and 4 display the damping capacities as a function of temperature after
homogenized (H) at 560°C for 4 hours, rolling at various percent of rolling reductions (RR) and
followed by annealed (A) at 345°C (1 hour and 3 hours soaking times) and 400°C (1 hour and 3
hours soaking times) respectively and also homogenized sample.
T2
H, 20% RR, A
H, 30% RR, A
H, 40% RR, A
H only
T1
Figure 1 Damping capacities as a function of temperature for various
percents of rolling reduction after annealed at 345°C for one hour
soaking time and homogenized sample.
T2
H, 20% RR, A
H, 30% RR, A
H, 40% RR, A
H only
T1
Figure 2 Damping capacities as a function of temperature for various
percents of rolling reduction after annealed at 345°C for three hours
soaking time and homogenized sample.
T2
H, 20% RR, A
H, 30% RR, A
H, 40% RR, A
H only
T1
Figure 3 Damping capacities as a function of temperature for various
percents of rolling reduction after annealed at 400°C for one hour
soaking time and homogenized sample.
T2
H, 20% RR, A
H, 30% RR, A
H, 40% RR, A
H only
T1
Figure 4 Damping capacities as a function of temperature for various
percents of rolling reduction after annealed at 400°C for three hours
soaking time and homogenized sample.
Basically, the damping capacity generally exhibited an increase at elevated temperatures for
all samples. Two significant transition points were observed in Figure 1 and Figure 2 which termed
as T1 and T2 respectively. T1 transition point took place at about 150°C in Figure 1 meanwhile in
between 100 to 150°C in Figures 2, 3 and 4. In all Figures 1 to 4, T2 took place at about 340°C in
homogenized samples meanwhile T2 took place at about 350°C in all homogenized, rolled
reduction and annealed samples.
In general, 30 percent and 40 percent rolling reductions in Figures 1 and 2 started to
decrease just after 350°C temperature. It have been reported that the grain boundary for high rolling
reduction were difficult to slip at high temperature and cause the damping capacity decreases [13]
and also believe to be depending also on the annealing parameters used. In this study, it also can be
seen that homogenized sample cannot maintain the damping capacity and drop drastically after
achieved the peak just after 350°C temperature. The same trend of decreasing of damping capacity
in as-rolled magnesium sheets also have been observed after achieved the peak at around 225°C
[14].
In this study, the promising results in the increasing of damping capacity values at higher
temperatures (above 350°C) has been achieved by homogenization at 560°C for 4 hours followed
by 20 percent rolling reduction annealing at 345°C for 1 hour relatively. Previous study by Ning Ma
et al. have reported that the increase of damping capacity is attributed by the decreasing of
annealing temperature [14]. However, in composite system combined with roll bonding, the results
indicate that by increasing the percentage of reinforcing phase, the damping capacity increases. It is
obvious that the damping capacity of the composite is higher than that of Al6061 alloy. Results
show that the increase of damping by internal friction is due to the presence of SiC particles [3].
ii) Microstructural Evolution
The microstructural evolution of the surface layer along the rolling direction of 1200
aluminium alloy after homogenization, rolling reduction and annealing the processes are presented
in Figure 5. From Figure 5, it could be said that the increase in percent reduction during cold rolling
procedure has led to a decrease of the average grain size of surface layers of the alloy. It can be
observed that when the rolling percent reduction was 40% (in Figure 5d), the equiaxed grains
became smaller and denser.
Zhang et al. reported that high damping capacity has been achieved in ultrafine-grained pure
aluminum L12 with a mean grain size of 1 μm was produced by equal channel angular pressing
(ECAP) and annealing at 150°C for 2 hours [5]. Koizumi et al. also reported that pure aluminum
that refined to ultrafine grains, smaller than 1 μm which possess a high damping capacity after
accumulative roll bonding for five cycles [4].
The following conclusions can be drawn from this study:
Conclusions
i ) The decreasing of damping capacity in 30 percent and 40 percent reductions just after 350°C
temperature was due to difficulty of grain boundary to slip at high temperature.
ii ) A lower annealing temperature at 375°C for 1 hour has produced increasing of damping
capacity values after 350°C in 20 percent rolling reduction relatively.
iii) The combination of homogenization, rolling reduction and annealing processes is viable to be
applied in order to sustain a stable damping capacity value at about 350°C in 1200 aluminium alloy.
Homogenization at 560°C for 4 hours followed by 20 percent rolling reduction and annealing at
345°C for 1 hour is the optimal treatment for 1200 aluminium alloy.
(a)
(c)
(b)
(d)
Figure 5 Optical micrographs show the microstructural evolution after
homogenization, rolling reduction and annealing at 50x magnification of
(a) 20 percent reduction, 345°C for 1 hour.
(b) 40 percent reduction, 345°C for 1 hour.
(c) 20 percent reduction, 400°C for 3 hours.
(d) 40 percent reduction, 400°C for 3 hours.
*Note : TD is tranverse direction, RD is rolling direction.
References
[1] Zhang J., Perez R. J. , Lavernia E. J., Documentation of damping capacity of metallic, ceramic
and metal-matrix composite materials, J Mater Sci 28(9) (1993) 2395-2404.
[2] Sugita I. K. G., Soekrisno R., Suyitno I. M. M., Mechanical and damping properties of silicon
bronze alloys for music applications. Int J Eng & Technol. 11(6) (2011) 98-105.
[3] E. Emadoddin, M. Tajally, M. Masoumi, Damping behavior of Al/SiCP multilayer composite
manufactured by roll bonding, Mater. Des. 42 (2012) 334-338.
[4] Yuichiro Koizumi, Masanori Ueyama, Nobuhiro Tsuji, Yoritoshi Minamino, Ken'ichi Ota, High
damping capacity of ultra-fine grained aluminum produced by accumulative roll bonding, J. Alloys
and Comp. 355 (2003) 47-51.
[5] Z. M Zhang, C. J. Xu, J. C. Wang, H. Z. Liu, Damping behavior of ultrafine-grained pure
aluminium L2 and the damping mechanism, Acta Metall. Sin., 19(3) (2006) 223-227.
[6] Laszlo S. Totha, Chengfan Gu, Ultrafine-grain metals by severe plastic deformation, Mat.
Charac., 92 (2014) 1-14.
[7] Zheng Ming-Yi, Fan Guo-Dong, Tong Li-Bo, Hu Xiao-Shi, Wu Kun, Damping behavior and
mechanical properties of Mg-Cu-Mn alloy processed by equal channel angular pressing, Trans.
Nonferrous Met. Soc. China, 18 (2008) 33-38.
[8] M. Wu, L. Hu , Y.C. Shao, Q.J. Zhou, Influence of the annealing cooling rate on the
icrostructure evolution and deformation behaviours in the cold ring rolling of medium steel, Mater.
Des., 32 (2011) 2292-2300.
[9] Håkan Hallberg, Influence of process parameters on grain refinement in AA1050 aluminum
during cold rolling, Int. J. Mech. Sci., 66 (2013) 260-272.
[10] C.Y. Xie, R. Schaller, C. Jaquerod, High damping capacity after precipitation in some
commercial aluminum alloys, Mater. Sci. Eng., A252 (1998) 78-84.
[11] Choong Do Lee, Damping properties on age hardening ofAl–7Si–0.3Mg alloy during T6
treatment, Materials Science and Engineering A, 394 (2005) 112-116.
[12] M. N. Mazlee, J. B. Shamsul, H. Kamarudin, Comparative dynamic mechanical properties of
non-superheated and superheated A357 alloys. Kov. Mate., 48, (2010) 1-6.
[13] Li Guo-Cong, Ma Yue, He Xiao-lei, Li Wei, Li Pei-Yong, Damping capacity of high strength-
damping aluminum alloys prepared by rapid solidification and powder metallurgy process. J. Trans.
Nonferrous Met. Soc. China., 22 (2012) 1112 – 1117.
[14] Ning Ma, Qiuming Peng, Junling Pan, Hui Li, Wenlong Xiao, Effect of microalloying with
rare-earth on recrystallization behavior and damping properties of Mg sheets, J. Alloys Comp., 592
(2014) 24–34.
|
1906.11822 | 2 | 1906 | 2019-08-05T21:00:42 | Design of plasmonic directional antennas via evolutionary optimization | [
"physics.app-ph",
"cond-mat.mes-hall"
] | We demonstrate inverse design of plasmonic nanoantennas for directional light scattering. Our method is based on a combination of full-field electrodynamical simulations via the Green dyadic method and evolutionary optimization (EO). Without any initial bias, we find that the geometries reproducibly found by EO, work on the same principles as radio-frequency antennas. We demonstrate the versatility of our approach by designing various directional optical antennas for different scattering problems. EO based nanoantenna design has tremendous potential for a multitude of applications like nano-scale information routing and processing or single-molecule spectroscopy. Furthermore, EO can help to derive general design rules and to identify inherent physical limitations for photonic nanoparticles and metasurfaces. | physics.app-ph | physics |
Design of plasmonic directional antennas via
evolutionary optimization
PETER R. WIECHA1,2,∗, CLÉMENT MAJOREL2, CHRISTIAN GIRARD2,
AURÉLIEN CUCHE2, VINCENT PAILLARD2, OTTO L. MUSKENS1, AND
ARNAUD ARBOUET2
1Physics and Astronomy, Faculty of Engineering and Physical Sciences, University of Southampton,
Southampton, UK
2CEMES-CNRS, Université de Toulouse, CNRS, UPS, Toulouse, France
∗[email protected]
Abstract: We demonstrate inverse design of plasmonic nanoantennas for directional light
scattering. Our method is based on a combination of full-field electrodynamical simulations via
the Green dyadic method and evolutionary optimization (EO). Without any initial bias, we find
that the geometries reproducibly found by EO, work on the same principles as radio-frequency
antennas. We demonstrate the versatility of our approach by designing various directional
optical antennas for different scattering problems. EO based nanoantenna design has tremendous
potential for a multitude of applications like nano-scale information routing and processing or
single-molecule spectroscopy. Furthermore, EO can help to derive general design rules and to
identify inherent physical limitations for photonic nanoparticles and metasurfaces.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
The vast opportunities of light-based nanotechnology lead to tremendous research efforts on
light-matter interaction at sub-wavelength scales, and in particular gave rise to the broad field of
plasmonics, as localized surface plasmon-polariton (LSP) resonances in metallic nanostructures
can strongly confine far-field radiation [1]. By variations of the particle geometry or its material
it is possible to tailor manifold optical properties like resonance positions [2], polarization
conversion [3], optical chirality [4], localized heat generation [5, 6] or nonlinear optical effects
[7, 8] with many applications like field enhanced spectroscopy [9] or refractive index sensing
[8, 10].
Another optical functionality that is in the focus of the plasmonics community is the directional
routing of light. Not only individual nanostructures were demonstrated for directional scattering
of far-field light [11 -- 13], color-routing [14], quantum emitter radiation steering [9, 15 -- 19],
electro-luminescence [20, 21] or directional non-linear emission [22]. Also metasurfaces for
directional scattering and color-routing have been proposed [23, 24]. Such photonic devices
imply applications like on-chip light routing and de-multiplexing [25] or sensing [9, 10].
To obtain functionalities such as directional scattering of light, a careful design of the nano-
antenna is crucial. In the conventional approach a reference geometry is chosen from qualitative
considerations whose properties are subsequently studied systematically. However, imposing
a geometry from the start risks excluding interesting solutions to the problem. A promising
alternative can be evolutionary optimization (EO) strategies. By mimicking natural selection
through a cycle of reproduction, evaluation and selection (see Fig. 1(a)), EO is able to find global
optima for complex non-analytical problems [26]. EO has recently been applied very successfully
to various problems in nano-optics like maximization of local field enhancement [27 -- 29] or
of optical resonances and structural color [6, 30 -- 34]. Other applications include the design
of carbon nanotube field-emission sources [35], dielectric anti-reflection coatings [36], optical
cloaking [37], multi focal-spot flat lenses [38] or photonic power-splitters [39].
In this paper we use EO to design directional optical antennas. We couple an evolutionary
Fig. 1. (a) Scheme illustrating evolutionary optimization. (b) Plasmonic antenna structure
model for EO of directional scattering. 40 gold-blocks (Bi), each 40× 40× 40nm3 large
are placed on a ns = 1.5 substrate (in the XY plane) within an area of 1× 1µm2. Plane
wave illumination at normal incidence (k along −z), with λ0 = 800nm, linearly polarized
along OX. (c) Sketch of the directionality problem: Maximize the ratio of scattered intensity
through a narrow window (green) and scattering to the remaining solid angle (red).
algorithm to full-field electro-dynamical simulations using the Green dyadic method (GDM).
With a randomly initialized and almost unconstrained geometric model, our approach repro-
ducibly yields antennas resembling the classical radio-frequency (RF) Yagi-Uda layout [40]. We
find, that the working principle of these plasmonic antennas is indeed equivalent to their RF
counter-parts. We demonstrate the versatility of our method on different directional scattering
problems such as varying angles of scattering, double-focusing, or quantum emitter steering. We
finally show that evolutionary optimization can yield general design rules and furthermore can
be used to identify and analyze physical limitations in nanoantenna design.
1. Optimization method, model and problem
Evolutionary optimization is a heuristic approach to search for an optimum solution of a complex
problem, which cannot be optimized by classical means such as steepest descent methods.
EO considers a certain number of randomly generated parametersets describing a problem.
Parameters may be for instance the positions or sizes of sub-constituents of a nanostructure. The
ensemble of parametersets is called the population. In a first step, each parameterset (also called
an individual) is evaluated, in our case through a numerical simulation of the optical scattering
(more details below). The evaluation criterion is in general referred to as the fitness function. In a
selection step, the best solutions of the population (i.e. with the highest fitness values) are chosen
for the following reproduction step. Therein, randomly chosen parameters are mutated (randomly
modified) and interchanged between individuals, mimicking evolutionary processes in nature
and yielding a new population of parametersets. This cyclic process is repeated multiple times
(see Fig. 1(a)) until some stop criterion like a timelimit is met, leading to a close-to-optimum
solution to the problem.
As EO algorithm we use the implementation of self-adaptive differential evolution ("jDE")
in the "pyGMO" toolkit [41]. It takes as input the directionality of scattering, calculated using
"pyGDM" [42], our own python implementation of the Green Dyadic Method (GDM) [43]. By
including a surface propagator we can take into account a glass substrate [44]. Details on our
EO-GDM technique can be found elsewhere [33, 42]. We want to note, that any numerical
approach for solving Maxwell's equations can be used together with EO (possible other methods
are for example finite difference time domain [28] or mode expansion techniques [32]). With the
Δy = 1000nmΔx = 1000nmB1blocks:40x40x40 nm3B5B10B40B30B20......b)xyzselection reproduction evaluation a)c)xyzΩdirectΩrest Fig. 2. (a) Fitness vs. EO iteration with a population of Npop = 500 individuals. Best and
average fitness are indicated by a green, respectively red line. (b-h) Best solutions at selected
iterations between random initialization (b) and final solution (h). Iteration numbers are
indicated at the bottom left. Scattering patterns in XZ-plane are shown in the top panels by
blue lines, a green segment indicates the optimization target. The corresponding gold cube
arrangements are shown in the panels below. (i) Phase of the electric field X component
relative to the driving element's center. (j) Functional components: feed (blue), reflector
(green) and director (red). The distances between the centers of gravity (dashed lines) are
indicated at the top. (k) Impact on the directionality ratio in decibel (10log10Rdirect) when
each block is toggled (gold block ↔ no gold block). Scale bars are 200nm.
GDM, we use a volume integral method which is particularly convenient when a relatively small
nanostructure is placed in a large domain, where the GDM provides fast convergence.
A scheme of the geometric model to be optimized is shown in Fig. 1(b). 40 gold blocks Bi,
each 40× 40× 40nm3 in size, are placed in air (nenv = 1) on a glass substrate (ns = 1.5). For
gold, we use the tabulated dispersion from Ref. [45]. Each of the blocks are modeled by 2×2×2
dipoles. A discussion on the discretization step and the elementary block size and shape can be
found in the appendix (Fig. 9). The optimization target parameters are the positions xi,yi of the
blocks, which are bound to an area of 1× 1µm2. Additionally, the positions of the blocks are
constrained to lie on a grid with steps of the block-size (40nm), to avoid the problem of partially
overlapping elements. This results in a total of 25× 25 possible positions. We note that blocks
with identical positions are treated as a single block, hence the amount of material is not strictly
constant. With 40 gold blocks, we have 80 free parameters and 25× 25!25× 25− 40! ≈ 10111
possible arrangements.
A plane wave with λ0 = 800nm, linearly polarized along OX, is normally incident onto the
gold blocks. Using the GDM, we calculate at first the average intensity per solid angle Idirect,
scattered through a solid angle Ωdirect of polar and azimuthal dimensions ∆ϕ = 30◦ and ∆θ = 45◦,
centered at ϕ = 45◦ and θ = 0◦ (diagonal between X and Z axis), as illustrated by the green
window in Fig. 1(c). Second we calculate Irest, the average intensity per solid angle, scattered
towards the remaining hemisphere (Ωrest, red in Fig. 1(c)). The goal of the EO is to maximize the
Rdirect = 10.8 dBIdirect = 0.0056Qscat = 0.057final solution:Qabs = 0.099a)b)c)d)e)f)g)h)i)j)#0#1#2#12#14#20#50zxyxyφx0.29k)significance of individual blocks ratio between scattering per solid angle to the target "window" and to the rest of the hemisphere:
(cid:0)
Ωdirect IΩdΩ(cid:1)Ωdirect
Ωrest IΩdΩ(cid:1)Ωrest
(cid:0)
Rdirect =
=
Idirect
Irest
.
(1)
2. Directional scattering of a plane wave
We found that for small populations (Npop (cid:28) 100), the optimization does not always converge to
the best observed directionality ratio of Rdirect ≈ 12 (10.8dB). Small populations seem to offer
insufficient genetic diversity and tend to get stuck in local extrema. The improvement during
50 generations of an evolution with a population size of Npop = 500 is shown in figure 2(a).
Blue dots correspond to the fitness of the individuals, the green and red lines indicate the best
and average fitness, respectively. The best solutions during successive iterations and their far-
field backscattering patterns are shown in Fig. 2(b)-2(g) and Fig. 2(h) shows the best solution
after the 50th generation. To verify the convergence and reproducibility, we repeated the same
optimization with random initial populations, yielding always similar antenna morphologies with
directionality ratios Rdirect ≈ 10.8dB (see Appendix, Fig. 7).
At first sight we observe that the optimized nano-antenna visually resembles a Yagi-Uda RF
antenna [40]. An analysis of the phase of the electric field Ex component reveals a difference of
π between the central and the outer parts of the nanostructure (Fig. 2(i)). We identify three main
functional elements, as shown in Fig. 2(j): A driving feed in the antenna center (blue), a reflector
on the left (green) and a director on the right (red). As we explain in more detail in the appendix
(see also Fig. 8), we want to note that this is not a categorical analogy to RF antenna design, but
rather a simplified picture giving an idea about the roles of the different parts of the structure.
The intrinsic phase differences from resonance de-tuning of the elements in combination with
the adequate distances between reflector, feed and director, result in cancellation of the scattered
fields in backward direction and in constructive interference in forward direction. In fact, this
working principle is equivalent to the RF Yagi-Uda antenna, except that in our case the scattering
is occurring at an angle with respect to the antenna axis. Actually it has been shown earlier, that
RF antenna concepts can work also at optical frequencies, despite large plasmonic losses [46,47].
Nano Yagi-Uda antennas have been demonstrated at several occasions [15, 19, 22]. Through
an analysis of the impact of a removal or insertion of a gold block at every possible position
(Fig. 2(k)) we find that the possible further increase in directionality is small. Furthermore we
observe that fusing single blocks in the reflector and director, or cutting through the feed element
kills the directionality. This is because changing the length of an element would result in a
spectral shift of its LSP resonance, which would break the correct relative phases and scattered
field amplitudes required for directional scattering.
2.1. Different directional scattering objectives
In a second step, we run several optimizations for directional scattering problems, shown in
figure 3. The top row polar plots indicate the target solid angles Ωdirect by green color, which
are of varying size and position. The bottom plots show top views of the optimized gold block
arrangements. The center plots show the scattering radiation patterns of the optimization outcome.
In order to demonstrate that directionality is not obtained mainly through the index contrast
at the substrate, we also show optimizations with different target solid angles for a vacuum
environment (nenv = 1, Figs. 3(a)-3(d)). Figs. 2, 3(e) and 3(f) show optimizations for plasmonic
structures on a dielectric substrate (ns = 1.5), where in 3(e) and 3(f) furthermore scattering into
the substrate was maximized. In all other cases scattering goes towards the air hemisphere.
While EO yields good directionality ratios in all cases, we observe that the scattering intensity
and scattering efficiency Qscat (ratio of scattering and geometric cross sections) are generally
very low. The highest scattering intensities are obtained for large target solid angles and small
Fig. 3. Examples of directional EO (maximize Rdirect) for various target solid angles (green
areas in top plots). Bottom plots show top views of the optimized geometries, where scale
bars are 200nm. Center plots show scattering patterns for the full hemisphere of a normally
incident plane wave at λ0 = 800nm, linearly X-polarized. (a) towards a loose focal area,
(b) towards 90◦, (c) bidirectionally and (d) towards a "donut" target area. (a-d) are set in
a vacuum environment and show backscattering. (e-f) are set on an ns = 1.5 dielectric,
hemispherical substrate in order to reflect the conditions in oil immersion microscopy. In
these two cases, forward scattering into the glass substrate is maximized for areas at the
critical angle.
deflection angles (Fig. 3(a)) or for scattering into a substrate at the critical angle (Fig. 3(e)). But
even those cases have very small scattering efficiencies well below 1. As a first conclusion we
record that it is a physically demanding task to steer an incident plane wave unidirectionally at
large angles via a planar gold nanostructure.
For comparison we ran EOs maximizing the absolute backscattered intensity Idirect towards
polar angles of 45◦ and 90◦, shown in Figs. 4(a) and 4(b), respectively. In both cases, the largest
absolute scattered intensity towards the target solid angle is found for a structure built of several
resonant gold elements, giving 2-3 orders of magnitude higher overall scattering efficiencies. But
most of the light is backscattered with no deflection, directionality is completely lost. Still, the
scattered intensity Idirect towards the target solid angles in Fig. 4 is several times higher compared
to Fig. 2 and Fig. 3(b), where the ratio Rdirect was maximized. The latter antennas are built from
detuned elements, resulting in a very weak absolute scattering. This supports our conclusion that
using planar plasmonic nanostructures, it is physically almost impossible to redirect a normally
incident plane wave towards the direction of its linear polarization. Large scattering intensities
seem to imply that almost no directionality is obtained.
We note that Fig. 4 shows structures on a substrate. We also performed the same optimization
in a vacuum environment, yielding similar results.
We provide a visualization online (Visualization 1), showing the progress of EO of a large-area
(5× 5µm2), intensity optimized scatterer (analog to Fig. 4(a)). It results in an assembly of
several resonant nanoantennas, arranged in a grating-like way. Despite the large allowed area
and consequently much more degrees of freedom, also this optimization fails to combine a good
directionality ratio with high scattering intensity.
a) loose focusb) 90°c) double focusd) donute) tight focusf) double focus(a-d) backscattering in vacuum(e-f) scatter into glass substratemin.max.xyφθ Fig. 4. EO for maximization of the absolute backscattered intensity Idirect to the target solid
angle centered at a polar angle of (a) 45◦ and (b) 90◦. Target solid angles, backscattering
radiation patterns and top views of the optimized geometries are shown respectively from
top to bottom. Scale bars are 200nm. Normally incident plane wave, linearly X-polarized
with λ0 = 800nm, the structures lie in vacuum on an n = 1.5 substrate. A visualization of
the optimization convergence for case (a) can be found online (Visualization 1).
2.2. Discussion: Weak scattering due to phase difference requirement
The weak scattering at high deflection angles is a direct result of the physical properties of the
plane wave illumination. The phase of the illumination field is constant across the entire planar
structure, which means that it is not possible to obtain phase differences only through positioning
of plasmonic elements. On the other hand, the directional antenna requires a proper phase
difference between antenna feed and director elements. To achieve this, the EO algorithm choses
very small blocks as director elements, which are driven by the plane wave below resonance
(ωres (cid:29) ω0), resulting in a phase of −π2 between illumination and response. As driving feed the
optimizer choses a large rod having its LSP resonance in the infrared (ωres (cid:28) ω0), which gives
a phase of +π2 with respect to the driving plane wave. In consequence, the feed and director
structures have a relative phase of π with respect to each other but are also entirely off-resonance
and thus scatter only very weakly. This results in an overall small scattering intensity, so most
of the incident light passes just through the plasmonic structure without interaction. The weak
interaction with the incident plane wave is confirmed by the very low absorption efficiencies of
the optimized antennas in Figs 2 and 3.
As we see in Figs. 4(a) and 4(b), if the intensity is target of the optimization, the resulting
structures look fundamentally different: The dispersed small blocks, that give the directionality-
optimized antennas the fragmented look, disappear. The EO algorithm now uses all the available
material to construct as many resonant nano-rods as possible, yielding a strong scattering cross
section but very weak directionality.
We finally want to note, that a drawback of the GDM with regards to fabrication robustness
is, that diagonally neighboring metallic blocks are electrically isolated in the simulations. In
an actual sample however, there are high chances that diagonal neighbor blocks are electrically
connected at their touching point due to unavoidable imperfections in the fabrication process.
Such electric connections can have a significant impact on localized surface plasmon resonances
and in consquence have the potential to change the optical response of the structure. To assess
the robustness of our results against fabricational imperfections, we ran simulations, connecting
diagonally touching elements in the optimized antennas using additional dipoles, which didn't
a) 45°maximize Idirectb) 90°min.max.xyφθ Fig. 5. Optimized antennas for dipole emitter directional emission. The quantum emitter is
oriented along OY , radiates at λ0 = 800nm and has a minimum distance to the gold structure
of 20nm. Emission is maximized at the critical angle of scattering inside the ns = 1.5
substrate. (a) maximize directionality ratio Rdirect. (b) maximize absolute intensity Idirect at
target solid angle. (c) simulation of reference antenna from Ref. [15]. (d) spectrum of total
dipole emission intensity, normalized to the intensity Idp of an isolated dipole emitter (no
antenna). The inset shows directionality spectra for the three antennas. Radiation patterns
are evaluated on a hemispherical screen inside the substrate. Scale bars are 200nm.
have a severe impact in these cases. To generally overcome this shortcoming, dielectric /
plasmonic hybrid structures may be designed which can yield inherent phase differences without
requiring precise tailoring of the relative LSP resonances [48].
3. Quantum emitter directional antenna
As found above, plane wave directional scattering at high scattering efficiency is physically
difficult to obtain. On the other hand, it has been demonstrated that plasmonic nanostructures
can be used as efficient directional antennas for quantum emitters, being also able to boost the
decay of a quantum transition thanks to the Purcell effect [9, 15 -- 17, 49]. In fact, in the case of a
local dipole transition as light source, there is not the problem that the illumination excites every
part of the antenna simultaneously and with constant phase. This makes it significantly easier to
find a solution.
Following the configuration reported in [15], we replace the plane wave illumination by a
dipole transition along OY , oscillating at λ0 = 800nm, positioned at the origin of the coordinate
system. We prohibit the EO algorithm to put a gold block at this location, which results in a
minimum distance of 20nm between gold and emitter. The emission is to be maximized towards
OX along the critical angle into the ns = 1.5 substrate. The EO converges towards an antenna
with a directionality ratio as high as Rdirect ≈ 13.9dB, shown in Fig. 5(a). Ref. [15] uses a
slightly different definition of directionality, therefore we simulate the antenna "YU145" of
Ref. [15] with the GDM, using the same emitter-feed distance of 20nm as for the EO. We find a
emitter pos.a) maximize Rdirectb) maximize Idirectc) referenced)min.max.xyφθ Fig. 6. Symmetry analysis, showing optimization results of directional quantum emitter gold
antennas in vacuum environment (nenv = 1) for different orientations of the dipole transition.
(a) and (d): dipole emitter along OX, (b) and (e): dipole emitter along OY , (c) and (f): dipole
emitter along OZ. In (a-c), the geometry is unconstrained, in (d-f), the same optimizations
were repeated, imposing an axial mirror symmetry at the x-axis for the geometry of the
antenna. Emitter placed at x = 0nm,y = 0nm,z = 20nm, indicated by a red arrow. Scale
bars are 200nm.
directionality of Rdirect, ref. ≈ 9.2dB (Fig. 5(c)). While this is lower compared to the optimized
antenna, we see in Fig. 5(d) that the emission intensity of the optimized antenna is around two
orders of magnitude weaker. We re-run the EO using as optimization target the absolute intensity
of directed emission (Fig. 5(b)). While directionality is moderately reduced to 11.9dB, the signal
from the intensity-optimized directional nanoantenna (orange line in Fig. 5(d)) is now almost
one order of magnitude higher than that of the reference. Having a closer look at the geometry
in Fig. 5(b), we observe that the EO algorithm designed the antenna such that the emitter is
surrounded by gold to maximize the Purcell effect, which strongly boosts the emitted intensity.
Finally we find that an increased directionality comes at the cost of a reduced spectral full width
at half maximum (FWHM) of Rdirect, hence leads to a narrower operation window (see inset
Fig. 5(d)).
3.1. Dipole emitter orientation
To study the role of the orientation of the quantum emitter's dipole transition, we run antenna
optimizations for X, Y and Z oriented dipoles, shown in Figs. 6(a)-6(c). To keep the symmetry
of the system as high as possible, we assume a vacuum environment and maximize scattering
towards OX (solid angle as used in Fig. 3(b)). In case of X- and Y -oriented dipoles (Figs. 6(a)-
6(b)), the optimization organizes the gold blocks in a compact arrangement, forming a strongly
directional antenna. On the other hand, for a dipole pointing out of the structure plane (Z
Rdirect=11.1dBRdirect=11.7dBRdirect=4.1dBmirrorsymmetryRdirect=2.5dBmirrorsymmetryRdirect=11.6dBmirrorsymmetryRdirect=4.1dBa) dipole Xb) dipole Yc) dipole Zd) dipole Xf) dipole Ze) dipole Y(a-c) no symmetry - unconstrained optimization(d-f) impose axial symmetry in optimizationmin.max.xyφθ direction), the limited height of the planar gold structure (H = 40nm) impedes obtaining a
significant optical response of the gold blocks at the emission wavelength of λ0 = 800nm, hence
directionality remains weak (Fig. 6(c)).
3.2. Antenna symmetry
In contrast to the intuitive expectation that symmetric antennas should perform best, EO finds
mostly rather fragmented antennas of low symmetry. To assess the role of symmetry, we re-run
the quantum emitter antenna optimizations for different dipole-orientation (Figs. 6(a)-6(c)),
imposing a mirror symmetry around the X axis. The results shown in Fig. 6(d)-6(f) suggest, that
symmetric antennas give a similar performance in the cases of dipole transitions aligned along Y
and Z, as depicted in Fig. 6(e), respectively (f), even though the symmetric designs do not yield
noticeable improvement over an unconstrained geometric model.
In the case of an X-oriented dipole however, the symmetric EO completely fails to design
a directional antenna (Fig. 6(d)). By inspecting the antenna in Fig. 6(a), we notice that it is
designed such that the polarization of the X-oriented dipole emitter is converted by a gold
element to a plasmonic current along Y , effectively generating an accordingly oriented plasmonic
dipole emitter. The scattered light of this LSP dipolar resonance is then guided towards the
target solid angle via a Yagi-Uda-type assembly of structures around the feed. By imposing an
axial symmetry at the dipole position, this local polarization conversion is physically impossible.
Instead, in Fig. 6(d) the EO algorithm designs a quadrupolar feed element, directing the light
close to the desired solid angle. However, due to symmetry, emission is zero towards X.
4. Conclusions
In conclusion, we demonstrated how evolutionary optimization can be used to design directional
plasmonic nanoantennas. In our approach, within the limits of possible geometries built with
40 equal gold blocks, the design of the structures is completely free. We showed, that EO
automatically finds a nanostructure layout which corresponds to well known design-principles
from radio-frequency antennas. By comparing the performance of directionality- and intensity-
optimized planar gold nanostructures, we found that high directionality ratios for scattering of
normal incidence plane waves to large angles comes necessarily at the cost of weak scattering
intensities. We demonstrated that in contrast to weak directional plane wave scattering, EO can
design very efficient directional nanoantennas for quantum emitters. Via an appropriate fitness
function, EO finds an antenna which not only yields very good directionality but also exploits
at best the Purcell effect for quantum emitter signal enhancement. Through an analysis of the
EO results for different fitness functions we furthermore found, that high directionality ratios
come at the cost of spectrally less broadband working windows. Finally, through an EO-based
analysis of symmetry conditions we found that, dependent on the orientation of the quantum
emitter dipole transition, it can be necessary to design and asymmetric nano-antenna for good
performance.
We conclude, that from the results of EO general design rules can be derived for the tailoring
of photonic nanostructures. EO can also be used as a benchmark tool in nano-photonic design
and can reveal intrinsic physical limitations of nano-optical configurations. We foresee that EO
has manifold further promising applications beyond directional scattering, for examples in field
enhanced spectroscopy or non-linear nanophotonics.
Supplementary materials
Visualization 1: Movie showing the convergence process of EO for an intensity-optimized, plane
wave illuminated, directional optical antenna.
Appendix
Reproducibility
To test the convergence and the reproducibility of the EO, we re-run the optimization shown
in main text Fig. 2 several times with different, randomly initialized starting populations. As
shown in Fig. 7, the final antennas converge always to the same maximum directionality ratio of
Rdirect ≈ 12. The geometries of the solutions are in all cases similar, leading to the same phase
distribution.
Fig. 7. Demonstration of reproducibility and convergence: Top row shows antenna geome-
tries and their directivity ratios for 5 independent runs of the same optimization, with random
initial EO populations. Bottom row shows the relative phase of the Ex field component.
Functional elements of plane wave directional antenna
It is clear that the parasitic elements around the feed lead to the directional scattering effect in
the plane-wave antennas (as demonstrated in Fig. 8(a) and 8(b) below). On the other hand it is
difficult to ascribe to specific clusters of parasitic elements the unambiguous role of a "reflector"
or a "director". To assess whether it is justified to call the agglomeration of blocks on the left of
the feed in Fig. 2 an equivalent of a "reflector" and the agglomeration of blocks on the right a
"director", we simulated the feed alone, the feed + "reflector" and the feed + "director", shown in
Fig. 8(b)-8(d).
Fig. 8. Analysis of functional elements for the antenna in Fig. 2. Top row: Top view of
the considered sub-part of the full antenna geometry. Bottom row: backscattering pattern.
Scale bars are 100nm, the colorscale represents the scattering intensity in arbitrary units
(normalized to the peak intensity of the full structure).
While we observe no directional scattering for the isolated feed (Fig. 8(b)), the reflector as
well as director alone suffice to impose a directional effect on the feed element (Figs. 8(c)-8(d)),
Rdirect=12.13Rdirect=12.17Rdirect=12.45Rdirect=12.21Rdirect=12.44−ππphaseEx(rad)100nma) full structure0.51.0100nmb) feed alone0.250.500.75100nmc) "reflector" + feed0.250.500.75100nmd) feed + "director"0.20.4100nme) feed + lower"reflector" part0.250.50 in analogy to what would be expected in an RF Yagi-Uda antenna. On the other hand, if only a
part of for example the reflector cluster is used, the direction of scattering becomes tilted with
respect to the target angle (Fig. 8(e)). The full ensemble of "reflector" blocks is necessary to
obtain the desired scattering effect. We conclude that it is adequate to consider the left and right
agglomerations of parasitic elements equivalent to a "reflector" and a "director".
As a final note, we want to stress that this is not a categorical analogy. A direct adaptation of
the structure to RF antennas is not easily possible, we rather consider our analogy a simplified
description of the role of the complex arrangements of parasitic elements.
Discretization, size and shape of elementary blocks
In principle any kind of constituents can be used with the EO-GDM method to compose the
structure. However, larger building blocks or elementary blocks of some complex shape constrain
the generality of possible geometries. On the other hand, small blocks allow to perform a more
complete search of the solution space, but at the cost of slower convergence. The choice of size
(and shape) of the elementary blocks is therefore always a compromise between how flexible the
geometric model should be and the speed of convergence of the optimization. If constraints to
the attainable diversity of geometries are acceptable, only few but larger building blocks and/or
blocks of more specific shape like cuboids, rods or polygonal structures might be chosen as
elementary unit. In view of a possible experimental realization of the structures, it makes also
sense to use building blocks which are not smaller then the structural resolution that can be
obtained in fabrication.
Fig. 9. Backscattering patterns of an identical gold nanostructure but using different dis-
cretization steps (geometry shown on the right, same as in Fig. 2). λ = 800nm plane wave
illumination from the top, linear polarized along X. Each elementary block in the structure
was discretized from left to right by 2× 2× 2 dipoles with 20nm step, 4× 4× 4 dipoles
with 10nm step and 6× 6× 6 dipoles with 6.67nm step. The results are very similar, hence
we can conclude that a step of 20nm is a good approximation.
Concerning the discretization, we used steps of 20nm which gives good optimization speed
because the full structure consists of only 40× 2× 2× 2 = 320 dipoles. In order to verify that
this rather coarse discretization is a good approximation, we re-calculated the structure shown
in figure 2 with finer step-sizes. As shown in Fig. 9, simulations with steps of 20nm (2× 2× 2
dipoles per block), 10nm (4× 4× 4 dipoles per block) and 6.67nm (6× 6× 6 dipoles per block)
yield very similar results, so our approximation using steps of 20nm is good.
Funding
This work was supported by Programme Investissements d'Avenir under the program ANR-11-
IDEX-0002-02, reference ANR-10-LABX-0037-NEXT and by the computing facility center
CALMIP of the University of Toulouse under grant P12167. PW received funding by the German
Research Foundation (DFG) through a Research Fellowship (WI 5261/1-1). OM acknowledges
support through EPSRC grant EP/M009122/1.
dipoles per block:scatteringpattern:geometry:xyzxφθmin.max. 6. C. Girard, P. R. Wiecha, A. Cuche, and E. Dujardin, "Designing thermoplasmonic properties of metallic metasurfaces,"
7. M. Kauranen and A. V. Zayats, "Nonlinear plasmonics," Nat. Photonics 6, 737 -- 748 (2012).
8. M. Mesch, B. Metzger, M. Hentschel, and H. Giessen, "Nonlinear Plasmonic Sensing," Nano Lett. 16, 3155 -- 3159
Appl. Phys. Lett. 94, 153109 (2009).
J. Opt. 20, 075004 (2018).
(2016).
Acknowledgments
We gratefully thank N. Bonod for fruitful discussions. All data supporting this study are openly
available from the University of Southampton repository (DOI: 10.5258/SOTON/D0992).
References
1. P. Mühlschlegel, H.-J. Eisler, O. J. F. Martin, B. Hecht, and D. W. Pohl, "Resonant Optical Antennas," Science 308,
1607 -- 1609 (2005).
2. S. J. Tan, L. Zhang, D. Zhu, X. M. Goh, Y. M. Wang, K. Kumar, C.-W. Qiu, and J. K. W. Yang, "Plasmonic Color
Palettes for Photorealistic Printing with Aluminum Nanostructures," Nano Lett. 14, 4023 -- 4029 (2014).
3. P. R. Wiecha, L.-J. Black, Y. Wang, V. Paillard, C. Girard, O. L. Muskens, and A. Arbouet, "Polarization conversion
in plasmonic nanoantennas for metasurfaces using structural asymmetry and mode hybridization," Sci. Reports 7,
40906 (2017).
4. V. K. Valev, J. J. Baumberg, C. Sibilia, and T. Verbiest, "Chirality and Chiroptical Effects in Plasmonic Nanostructures:
Fundamentals, Recent Progress, and Outlook," Adv. Mater. 25, 2517 -- 2534 (2013).
5. G. Baffou, R. Quidant, and C. Girard, "Heat generation in plasmonic nanostructures: Influence of morphology,"
9. D. Vercruysse, X. Zheng, Y. Sonnefraud, N. Verellen, G. Di Martino, L. Lagae, G. A. E. Vandenbosch, V. V.
Moshchalkov, S. A. Maier, and P. Van Dorpe, "Directional Fluorescence Emission by Individual V-Antennas
Explained by Mode Expansion," ACS Nano 8, 8232 -- 8241 (2014).
10. M. Wersäll, R. Verre, M. Svedendahl, P. Johansson, M. Käll, and T. Shegai, "Directional Nanoplasmonic Antennas
for Self-Referenced Refractometric Molecular Analysis," The J. Phys. Chem. C 118, 21075 -- 21080 (2014).
11. D. Vercruysse, Y. Sonnefraud, N. Verellen, F. B. Fuchs, G. Di Martino, L. Lagae, V. V. Moshchalkov, S. A. Maier,
and P. Van Dorpe, "Unidirectional Side Scattering of Light by a Single-Element Nanoantenna," Nano Lett. 13,
3843 -- 3849 (2013).
12. A. Abass, P. Gutsche, B. Maes, C. Rockstuhl, and E. R. Martins, "Insights into directional scattering: From coupled
dipoles to asymmetric dimer nanoantennas," Opt. Express 24, 19638 (2016).
13. K. Yao and Y. Liu, "Controlling Electric and Magnetic Resonances for Ultracompact Nanoantennas with Tunable
14. T. Shegai, S. Chen, V. D. Miljkovi´c, G. Zengin, P. Johansson, and M. Käll, "A bimetallic nanoantenna for directional
Directionality," ACS Photonics (2016).
colour routing," Nat. Commun. 2, 481 (2011).
15. A. G. Curto, G. Volpe, T. H. Taminiau, M. P. Kreuzer, R. Quidant, and N. F. van Hulst, "Unidirectional Emission of a
Quantum Dot Coupled to a Nanoantenna," Science 329, 930 -- 933 (2010).
16. I. S. Maksymov, I. Staude, A. E. Miroshnichenko, and Y. S. Kivshar, "Optical Yagi-Uda nanoantennas," Nanophoton-
ics 1, 65 -- 81 (2012).
17. I. M. Hancu, A. G. Curto, M. Castro-López, M. Kuttge, and N. F. van Hulst, "Multipolar Interference for Directed
Light Emission," Nano Lett. 14, 166 -- 171 (2014).
18. D. Dregely, K. Lindfors, M. Lippitz, N. Engheta, M. Totzeck, and H. Giessen, "Imaging and steering an optical
wireless nanoantenna link," Nat. Commun. 5, 4354 (2014).
19. M. Ramezani, A. Casadei, G. Grzela, F. Matteini, G. Tütüncüoglu, D. Rüffer, A. Fontcuberta i Morral, and
J. Gómez Rivas, "Hybrid Semiconductor Nanowire -- Metallic Yagi-Uda Antennas," Nano Lett. 15, 4889 -- 4895 (2015).
20. S. P. Gurunarayanan, N. Verellen, V. S. Zharinov, F. James Shirley, V. V. Moshchalkov, M. Heyns, J. Van de Vondel,
I. P. Radu, and P. Van Dorpe, "Electrically Driven Unidirectional Optical Nanoantennas," Nano Lett. 17, 7433 -- 7439
(2017).
21. R. Kullock, P. Grimm, M. Ochs, and B. Hecht, "Directed emission by electrically driven optical antennas," in
Quantum Sensing and Nano Electronics and Photonics XV, vol. 10540 (2018), pp. 1054012 -- 10540 -- 9.
22. X. Y. Z. Xiong, L. J. Jiang, W. E. I. Sha, Y. H. Lo, and W. C. Chew, "Compact Nonlinear Yagi-Uda Nanoantennas,"
Sci. Reports 6, 18872 (2016).
23. K. Lindfors, D. Dregely, M. Lippitz, N. Engheta, M. Totzeck, and H. Giessen, "Imaging and Steering Unidirectional
Emission from Nanoantenna Array Metasurfaces," ACS Photonics 3, 286 -- 292 (2016).
24. C. Yan, K.-Y. Yang, and O. J. F. Martin, "Fano-resonance-assisted metasurface for color routing," Light. Sci. & Appl.
6, e17017 (2017).
25. R. Guo, M. Decker, F. Setzpfandt, X. Gai, D.-Y. Choi, R. Kiselev, A. Chipouline, I. Staude, T. Pertsch, D. N. Neshev,
and Y. S. Kivshar, "High -- bit rate ultra-compact light routing with mode-selective on-chip nanoantennas," Sci. Adv.
3, e1700007 (2017).
26. S. Sivanandam and S. Deepa, Introduction to Genetic Algorithms (Springer, Heidelberg, 2008).
27. C. Forestiere, M. Donelli, G. F. Walsh, E. Zeni, G. Miano, and L. Dal Negro, "Particle-swarm optimization of
broadband nanoplasmonic arrays," Opt. Lett. 35, 133 (2010).
28. T. Feichtner, O. Selig, M. Kiunke, and B. Hecht, "Evolutionary Optimization of Optical Antennas," Phys. Rev. Lett.
29. C. Forestiere, Y. He, R. Wang, R. M. Kirby, and L. Dal Negro, "Inverse Design of Metal Nanoparticles' Morphology,"
109, 127701 (2012).
ACS Photonics 3, 68 -- 78 (2016).
30. P. Ginzburg, N. Berkovitch, A. Nevet, I. Shor, and M. Orenstein, "Resonances On-Demand for Plasmonic Nano-
Particles," Nano Lett. 11, 2329 -- 2333 (2011).
31. F. Bigourdan, F. Marquier, J.-P. Hugonin, and J.-J. Greffet, "Design of highly efficient metallo-dielectric patch
antennas for single-photon emission," Opt. Express 22, 2337 (2014).
32. A. Mirzaei, A. E. Miroshnichenko, I. V. Shadrivov, and Y. S. Kivshar, "Superscattering of light optimized by a
genetic algorithm," Appl. Phys. Lett. 105, 011109 (2014).
33. P. R. Wiecha, A. Arbouet, C. Girard, A. Lecestre, G. Larrieu, and V. Paillard, "Evolutionary multi-objective
optimization of colour pixels based on dielectric nanoantennas," Nat. Nanotechnol. 12, 163 -- 169 (2017).
34. A. K. González-Alcalde, R. Salas-Montiel, H. Mohamad, A. Morand, S. Blaize, and D. Macías, "Optimization of
all-dielectric structures for color generation," Appl. Opt. 57, 3959 -- 3967 (2018).
35. P. Y. Chen, C. H. Chen, J. S. Wu, H. C. Wen, and W. P. Wang, "Optimal design of integrally gated CNT field-emission
devices using a genetic algorithm," Nanotechnology 18, 395203 (2007).
36. B. L. Good, S. Simmons, and M. Mirotznik, "General optimization of tapered anti-reflective coatings," Opt. Express
24, 16618 (2016).
37. A. Mirzaei, A. E. Miroshnichenko, I. V. Shadrivov, and Y. S. Kivshar, "All-Dielectric Multilayer Cylindrical
Structures for Invisibility Cloaking," Sci. Reports 5, 9574 (2015).
38. J. Hu, X. Ren, A. N. Reed, T. Reese, D. Rhee, B. Howe, L. J. Lauhon, A. M. Urbas, and T. W. Odom, "Evolutionary
Design and Prototyping of Single Crystalline Titanium Nitride Lattice Optics," ACS Photonics 4, 606 -- 612 (2017).
39. S. Molesky, Z. Lin, A. Y. Piggott, W. Jin, J. Vuckovi´c, and A. W. Rodriguez, "Inverse design in nanophotonics," Nat.
40. H. Yagi, "Beam Transmission of Ultra Short Waves," Proc. Inst. Radio Eng. 16, 715 -- 740 (1928).
41. F. Biscani, D. Izzo, and C. H. Yam, "A Global Optimisation Toolbox for Massively Parallel Engineering Optimisation,"
Photonics 12, 659 (2018).
arXiv:1004.3824 [cs, math] (2010).
42. P. R. Wiecha, "pyGDM -- A python toolkit for full-field electro-dynamical simulations and evolutionary optimization
of nanostructures," Comput. Phys. Commun. 233, 167 -- 192 (2018).
43. C. Girard, E. Dujardin, G. Baffou, and R. Quidant, "Shaping and manipulation of light fields with bottom-up
plasmonic structures," New J. Phys. 10, 105016 (2008).
44. C. Girard, A. Dereux, O. J. F. Martin, and M. Devel, "Generation of optical standing waves around mesoscopic
surface structures: Scattering and light confinement," Phys. Rev. B 52, 2889 -- 2898 (1995).
45. P. B. Johnson and R. W. Christy, "Optical Constants of the Noble Metals," Phys. Rev. B 6, 4370 -- 4379 (1972).
46. H. F. Hofmann, T. Kosako, and Y. Kadoya, "Design parameters for a nano-optical Yagi -- Uda antenna," New J. Phys.
9, 217 (2007).
47. L. Novotny, "Effective Wavelength Scaling for Optical Antennas," Phys. Rev. Lett. 98, 266802 (2007).
48. T. Zhang, J. Xu, Z.-L. Deng, D. Hu, F. Qin, and X. Li, "Unidirectional Enhanced Dipolar Emission with an Individual
Dielectric Nanoantenna," Nanomaterials 9, 629 (2019).
49. G. C. des Francs, J. Barthes, A. Bouhelier, J. C. Weeber, A. Dereux, A. Cuche, and C. Girard, "Plasmonic Purcell
factor and coupling efficiency to surface plasmons. Implications for addressing and controlling optical nanosources,"
J. Opt. 18, 094005 (2016).
|
1908.03056 | 1 | 1908 | 2019-08-08T13:20:54 | Electric selective activation of memristive interfaces in TaO$_x$-based devices | [
"physics.app-ph",
"cond-mat.mes-hall"
] | The development of novel devices for neuromorphic computing and non-traditional logic operations largely relies on the fabrication of well controlled memristive systems with functionalities beyond standard bipolar behavior and digital ON-OFF states. In the present work we demonstrate for Ta$_2$O$_5$-based devices that it is possible to selectively activate/deactivate two series memristive interfaces in order to obtain clockwise or counter-clockwise multilevel squared remanent resistance loops, just by controlling the (a)symmetry of the applied stimuli and independently of the nature of the used metallic electrodes. Based on our thorough characterization, analysis and modeling, we show that the physical origin of this electrical behavior relies on controlled oxygen vacancies electromigration between three different zones of the active Ta$_2$O$_{5-x}$ layer: a central -- bulk -- one and two quasi-symmetric interfaces with reduced TaO$_{2-h(y)}$ layers. Our devices fabrication process is rather simple as it implies the room temperature deposition of only one CMOS compatible oxide -- Ta-oxide -- and one metal, suggesting that it might be possible to take advantage of these properties at low cost and with easy scability. The tunable opposite remanent resistance loops circulations with multiple -- analogic -- intermediate stable states allows mimicking the adaptable synaptic weight of biological systems and presents potential for non-standard logic devices. | physics.app-ph | physics |
Electric selective activation of memristive interfaces in TaOx-based devices
C. Ferreyra 1, M. J. S´anchez 2, M. Aguirre3,4,5, C. Acha6, S. Bengi´o2, J. Lecourt7, U. Luders7, D. Rubi1
1 GIyA and INN-CONICET, CNEA, Av. Gral Paz 1499 (1650),San Mart´ın, Buenos Aires, Argentina.
2 INN-CONICET, Centro At´omico Bariloche and Instituto Balseiro, (8400) San Carlos de Bariloche, Argentina.
3 Departmento de F´ısica de Materia Condensada,
Universidad de Zaragoza, Pedro Cerbuna 12 50009 Zaragoza - Spain.
4 Laboratorio de Microscop´ıas Avanzada (LMA),
Instituto de Nanociencia de Arag´on (INA)-Universidad de Zaragoza,
C/Mariano Esquillor s/n. 50018 Zaragoza, Spain.
5 Instituto de Ciencias de Materiales de Arag´on (ICMA), Universidad de Zaragoza, Zaragoza, Spain.
6 Depto. de F´ısica, FCEyN, Universidad de Buenos Aires and IFIBA,
UBA-CONICET, Pab I, Ciudad Universitaria, Buenos Aires (1428), Argentina.
7 CRISMAT, CNRS UMR 6508, ENSICAEN, 6 Boulevard Marchal Juin,14050 Caen Cedex 4, France.
(Dated: August 9, 2019)
The development of novel devices for neuromorphic computing and non-traditional logic oper-
ations largely relies on the fabrication of well controlled memristive systems with functionalities
beyond standard bipolar behavior and digital ON-OFF states. In the present work we demonstrate
for Ta2O5-based devices that it is possible to selectively activate/deactivate two series memristive
interfaces in order to obtain clockwise or counter-clockwise multilevel squared remanent resistance
loops, just by controlling the (a)symmetry of the applied stimuli and independently of the nature
of the used metallic electrodes. Based on our thorough characterization, analysis and modeling,
we show that the physical origin of this electrical behavior relies on controlled oxygen vacancies
electromigration between three different zones of the active Ta2O5−x layer: a central -bulk- one and
two quasi-symmetric interfaces with reduced TaO2−h(y) layers. Our devices fabrication process is
rather simple as it implies the room temperature deposition of only one CMOS compatible oxide
-Ta-oxide- and one metal, suggesting that it might be possible to take advantage of these properties
at low cost and with easy scability. The tunable opposite remanent resistance loops circulations
with multiple -analogic- intermediate stable states allows mimicking the adaptable synaptic weight
of biological systems and presents potential for non-standard logic devices.
INTRODUCTION
Non-volatile resistive switching (RS) devices, usually
called memristors, are intensively studied due to their
potential as resistive random access memories (ReRAM)
[1], novel logic units [2] and neuromorphic computing [3]
devices. RS has been found for a large number of simple
and complex transition metal oxides, including TaOx [4].
Outstanding memristive figures have been reported in
this system, including endurances larger than 1012 cycles
[5], ON-OFF ratios up to 106 and retention times greater
than 10 years [6]. RS in simple oxides such as TaOx has
been reported as bipolar and the physical mechanism was
usually attributed to the creation and disruption of con-
ducting oxygen vacancies (OV) nano-filaments [4]. This
process usually takes place in a region close to a metal
electrode with a high work function such as Pt, which
in contact with a n-type insulating oxide forms a Schot-
tky barrier [7]. The drift diffusion of OV to and from
the interface under the action of the external electrical
field modifies the Schottky barrier and modulates its re-
sistance. The application of negative voltage to the elec-
trode attracts positively charged OV to the interface and
triggers the transition from high resistance (HR) to low
resistance (LR), usually called SET process. The op-
posite transition is achieved with positive voltage and
is usually called RESET process.
In this way, the cir-
culation of the current-voltage (I-V) curve is clockwise
in the positive current-positive voltage quadrant, while
the remanent resistance vs. voltage curve (usually called
Hysteresis Switching Loop, HSL [8]) displays a counter-
clockwise behavior.
The possibility of opposite (counter-clockwise) I-V and
(clockwise) HSL circulations was also shown for TiO2/Pt
interfaces, and attributed to oxygen exchange between Pt
and TiO2 at the local position of the filament [9]. Op-
posite switching polarities were also reported for TaOx
devices and attributed to a competition between ionic
motion and electronic - i.e charge trapping- effects [10],
to the formation of filaments with either conical or hour-
glass shapes [11, 12] or to volumetric oxygen vacancies
exchange between TaOx layers with different stochiome-
tries [13]. The coexistance of two bipolar memristive
regimes with opposite polarities in a single device allows
to increase or decrease the resistance of the device with
stimulus of different amplitudes but the same polarity,
which has potential for the development of beyond von
Neumann novel computing devices [9].
Asymmetric devices (i.e.
an insulating oxide sand-
wiched between two electrodes of different metals) with
an active interface and an ohmic non-active one were re-
ported to display squared HSL [14, 15]. On the contrary,
symmetric systems presenting two similar metal/oxide
memristive active interfaces in series were reported to
display HSLs with the so-called "table with legs" (TWL)
shape [8, 16]. In these cases, the two interfaces behave in
a complementary way: when one switches from low resis-
tance (LR) to high resistance (HR), the other switches in-
versely. Symmetric Pt/TiO2/Pt devices stimulated with
symmetric stimuli displayed complex I-V curves with
multiple transitions, which was interpreted as bipolar
switching taking place simultaneously at both Pt/TiO2
interfaces [17].
In the present paper we show for Ta2O5-based de-
vices that a careful tuning of the stimuli protocol allows
to selectively activate/deactivate the contribution of two
memristive interfaces to the overall RS behavior, allow-
ing to obtain both clockwise/counter-clockwise multilevel
squared HSL or a "table with legs" behavior, where both
interfaces are active simultaneously.
We show that the (a)symmetry of the electrical re-
sponse can be controlled by appropriately tuning the elec-
troforming process and the subsequent excitation pro-
tocol, independently of the electrodes symmetry. From
a combination of I-V curves analysis and modeling,
togheter with numerical simulations based on the Voltage
Enhanced Oxygen Vacancy migration (VEOV) model [8],
adapted for binary oxides-based devices [14], we achieve
a thorough understanding of the observed phenomenol-
ogy, which is rationalized in terms of oxygen vacancies
electromigration between a central (bulk) zone of the ac-
tive Ta2O5 layer and its interfaces with metallic TaOx
(x < 2). Finally, the potential technological implications
of the observed tunable opposite HSL circulations with
multiple intermediate stable states are discussed.
RESULTS
We have fabricated and characterized Ta-oxide bilayers
grown on platinized silicon at different oxygen pressures.
We will first describe the performed transmission elec-
tron microscopy characterization. Figure 1(a) shows a
STEM-HAADF cross section, evidencing that the thick-
nesses of the Ta-oxide layers grown at 0.01 (layer B) and
0.1 (layer A) mbar of O2 are 35nm and 15nm, respec-
tively. Both layers are amorphous and differ in their
contrast. A brighter image is found for the B layer, in-
dicating a higher relative concentration of Ta (see the
blow up displayed in Figure S1 of the Supp.
Informa-
tion), consistently with a more reduced oxide and the
lower oxygen pressure used for its growth. Figure 1(b)
shows an STEEM-HAADF-EELS map of the bilayer in-
dicating a higher relative concentration of oxygen at the
top A layer. Performed quantifications, as the one shown
in Figure1 (c), indicate average oxygen concentrations of
≈ 70 %at and ≈ 60%at for A and B layers, respectively.
This suggests average stochiometries for A and B layers
of Ta2O4.70 and TaO1.67, indicating in the latter case a
mixture of ≈ 67 % TaO2 and ≈ 33 % TaO. Consistent
results are obtained from STEEM-HAADF-EDX quan-
tification, as shown in Figure S1 of the Supp. Information
[18].
2
FIG. 1: (a) STEM-HAADF cross-section corresponding
to a TaOx bilayer grown at 0.1 (15nm, layer A) and
0.01 (35nm, layer B) mbar of O2; (b)
STEEM-HAADF-EELS line scan (colour map) of
oxygen concentration at a zone close to the interface
between both layers. The line scan starts in the more
oxidized layer and ends in the more reduced one. See
text for details; (c) Oxygen concentration quantification
from data presented in (c); (d) Sketch showing the
geometry and average chemical composition of the
pristine devices. The bottom Pt electrode was grounded
and electrical stimuli was applied to the top one.
XPS measurements were performed to obtain informa-
tion of the films surface chemical composition and Ta ox-
idation states. Both Ta-4f and O-1s spectra are shown
in Figure 2 (a) and (b) for TaOx films grown at 0.1 mbar
and 0.01 mbar of O2, respectively. The spectra were fit-
ted using a Voight function for each peak plus a Shirley-
type background. The total fitted intensities along with
the experimental ones are shown in each spectrum. Ta-
4f spectra components are characerized by two identical
peaks corresponding to the spin-orbit split between 4f7/2
and 4f5/2 levels, with relative intensities of 3:4. The Ta-
4f spectrum of the film grown at 0.1mbar of O2 display
one component corresponding to Ta2O5 and for the film
grown at 0.01mbar of O2 three components were iden-
tified: a major one at binding energy 26.7 eV ascribed
to Ta2O5 and two minor ones at binding energies 25.9
eV and 24eV, ascribed to TaO2 and TaO respectively.
Besides, there is a O-2s component at 27.8eV. The pres-
ence of some Ta2O5 at the spectra of the film grown at
low O2 pressure, absent in the case of layer B (buried) in
the bilayer analyzed by STEM-HAADF, is related to the
ex-situ character of the XPS experiments and indicates
surface re-oxidation upon exposure of the film to ambient
pressure. It is worth noting that no sputtering process
was performed on the films, as it is a possible source of
vacancies creation and formation of metastable suboxides
[19]. The O−1s spectra display a major component at
530.8 eV, ascribed to the metal oxide and a minor com-
ponent at 532.5 eV, attributed to adsorbed molecules.
3
Based on combined STEM-HAADF and XPS
as
sketch
shown in the
virgin devices
our
characterize
analysis, we
Pt/Ta2O4.70/TaO1.67/Pt,
displayed on Figure 1(d).
as
FIG. 3: I-V response of a device with 28x103 µm2 area,
obtained for voltage excursions between Vmax = +1.5 V
and Vmin = -2 V (a), and for Vmax =2 V and Vmin=-1.5
V (b), where the circulation of the I-V curve is inverted.
(c),(d) Fittings -in dashed lines- performed on I-V
curves recorded for a 6.4x103 µm2 device, for the high
and low resistance states, labeled respectively HR and
LR. The used equivalent circuit is described in the text
and Supp. Information. The insets show the associated
γ vs V1/2 representation for the LR ans HR. For better
visualization, the LR state was not included in the inset
of panel (c). The dashed line corresponds to the
theoretical γ curve, extracted from the I-V curves
fittings.
FIG. 2: (a) X-ray photoemission Ta-4f and O-1s
spectra recorded on a TaOx film grown at 0.1 mbar of
O2; (b) Same spectra for a TaOx film grown at
0.01mbar of O2. The references for Ta2O5 at 26.7 eV,
TaO2 at 25.9 eV, TaO at 24 eV, lattice oxygen at 530.8
eV and adsorbed O2 at 532.5 eV are shown as vertical
lines. The areas under each component of the spectra
are displayed with different colours. See text for details.
We describe now the electrical characterization. The
bottom Pt electrode was grounded and the electrical
stimulus (voltage) was applied to the top Pt electrode.
The devices were initially in a low resistance state (100
Ω); an electroforming process with negative stimulus (i.e.
-3 V pulses, with a 1 ms width) was necessary to in-
crease the devices resistance to the range of kΩ and ac-
tivate their memristive behaviour. After forming, dy-
namic current-voltage (I-V) curves were obtained by ap-
plying a sequence of voltage pulses of different amplitudes
(0→ Vmax → −Vmin → 0, with a time-width of 1ms and
a step of 100 mV), with the current measured during the
application of the pulse. Additionally, after each voltage
pulse a small reading voltage was applied to allow the
current to be measured and the remnant resistance state
determined, obtaining the HSL. We have found that the
devices electrical response is highly dependant on both
Vmax and Vmin.
Figures 3 (a) and 4 (a) display respectively, the dy-
namic I-V curve and HSL obtained for a device with
28x103 µm2 area, for Vmax = +1.5 V and Vmin = -2
V. It is found that the transition from HR2 to LR2 (SET
process, the notation of the resistive states is chosen to
be consistent with the fittings and simulations to be de-
scribed below) is achieved with positive stimulus (VS ≈
+1V) while the opposite transition (RESET) is obtained
with negative voltage (VR ≈ -1.2 V). The LR2 and HR2
states are 0.8 kΩ and 2.5 kΩ respectively, giving an ON-
OFF ratio of 3.1. The stability and reproducibility of the
curves is remarkable, as shown in Figure S2 of the Suppl.
Information for 200 consecutive cycles. In addition, re-
tention times up to 104 s were checked for both resistive
states.
From these measurements, it is established that the
HSL present a clockwise (CW) circulation. Remarkably,
this circulation is inverted to a counter-clockwise (CCW)
behaviour when the maximum voltage excursion are in-
verted to Vmax = +2 V and Vmin = -1.5 V, as shown in
Figures 3 (b) and 4 (d), respectively. In this case, the
transition from HR1 to LR1 (SET process) is achieved
with negative stimulus (VS ≈ -1V) while the opposite
transition (RESET) is obtained with positive voltage (VR
≈ +1.5 V). The LR1 and HR1 states in this case are 1
kΩ and 3.3kΩ respectively, giving an ON-OFF ratio of
3.3. Again, an excellent stability upon consecutive cy-
cling is found (200 cycles), as shown in Fig. S2 of the
Supp. Information, and retention times for up 104s were
also checked. These results clearly show that the circu-
lation of the HSL can be controlled and tuned on the
same device by properly choosing the voltage excursions
of applied stimuli. The inverse dependance of LR1 and
LR2 states with the device area (shown in Figure S3 of
the Supp.
Information) indicates the existance of con-
ducting paths comprising the complete device area.
Interestingly, if the voltage excursions are enlarged and
symmetrized to Vmax = −Vmin=2 V, the HSL change
from squared to a TWL-like shape, as shown in Figure
4 (c). The squared HSLs of Figures 4 (a) and (b) cor-
respond to a single active interface while the TWL of
Figure 4 (c) corresponds to two complementary active
interfaces; that is, when one switches from low to high
resistance the other one changes inversely [8]. These re-
sults clearly indicate that our two memristive series in-
terfaces can be selectively decoupled and activated by
means of the application of proper electrical stimuli, and
goes against previous claims about the "simultaneous"
memristive behavior of two series interfaces [17].
We have found that for both (squared) HSL circula-
tions multilevel states are possible. Figure 5 (a) shows,
for a device with 11.3x103 µm2 area, that different non-
volatile resistance levels can be achieved for CW HSLs
upon increasing VS from 1.3 to 1.9 V, while keeping VR
fixed at -1.9 V. In a similar way, Figure 5 (b) shows a
multi-level behavior for a CCW HSL upon changing VS
from -1.3 V to -1.9 V while keeping VR=1.2 V fixed. We
note that this device presents higher remanent resistance
states due to its lower area in relation to the previously
described devices of Figures 3 and 4.
The multilevel resistance states show that our devices
behave in an analogic way, which is a key feature to mimic
the adaptable sinaptic weight of biological synapsis.
In order to understand the described behavior, we start
with the forming process. Virgin devices are found in a
low resistance state, indicating that the more oxidated A
layer (with an average stochiometry Ta2O4.70, according
to STEEM-HAADF experiments) presents a substantial
higher conductivity that the one expected for the insulat-
ing stochiometric compound. This can be explained both
by the presence of strong oxygen deficiency and to some
material inhomogeneity at the nanoscale (see Figure S1 of
the Supp. Information), which allow the presence of per-
colative low resistance Ta-rich paths bridging the nearby
layers. However, as both TaO2 and TaO present at the
4
FIG. 4: HSL's with (a) CW circulation and (b) CCW
circulation, for asymmetric voltage excursions; (c) TWL
HSL obtained for symmetric voltage excursions,
showing two complementary active interfaces. The same
device than in Figures 3 (a) and (b) was used ; (d)
Voltage protocol used to obtain dynamic I-V curves and
HSL. Vmax (Vmin) is the maximum (minimum) voltage
excursion for the write pulses. Notice the different
values of Vmin and Vmax for the CW and CCW HSL's
of panels (a) y (b), respectively.
B layer are metallic [4], the A layer still dominates the
resistance of the oxide bilayer. The application of neg-
ative voltage to the top (left in Figure 6 (a)) electrode
triggers an electroforming process that increases the re-
sistance of the device by inducing the migration of OVs
within the A layer, as the electrical field acting on this
layer is higher that the one acting on the more reduced B
layer. Oxygen ions are pushed down within the A layer
(from left to right in the scheme of Figure 6 (a)), lead-
ing to the formation of a nearly stochiometric Ta2O5−x,
resistive, layer and a strongly reduced, metallic, layer in
contact with the Pt top (left zone in Figure 6 (a)) elec-
trode. The presence of this reduced layer is infered from
the symmetric and non-rectifying behavior observed in
the post-forming I-V curves (Figures 3 (a) and (b)) which
rule out the presence of Ta2O5−x in contact with the left
Pt electrode, as the Pt/Ta2O5−x interface is known to
present rectifying (Schottky) behavior [5, 20].
Interestingly, we have found that if the forming pro-
cess is done with positive voltage, a rectifying behavior
is observed in the I-V curves, indicating that in that case
the formed Ta2O5−x layer is in contact with the top Pt
electrode, consistently with the inversion of the forming
polarity (see Figure S4 of the Supp. Information). We
return now to the negative forming scenario.
Unlike the case of Ti-oxides, which display a large num-
ber of stable sub-stochiometric phases with less oxygen
content than TiO2 [21], the commonly accepted Ta-oxide
phase diagram only allows stable Ta2O5 and Ta(O) -
that is metallic Ta with some diluted oxygen- phases [22].
TaO2 has been considered as a metastable phase [23], but
recent reports claim that it might be stable with an even
lower potencial energy than Ta2O5 [24, 25]. This sug-
gests that the as-grown A layer, with nominal Ta2O4.70
composition, might be a metastable phase that is phase
separated during forming, due to Joule heating [23], into
nearly stochiometric stable Ta2O5−x and TaO2−h layers
(x,h << 1), as depicted in Figure 6 (b). If we assume
that the forming process only involves oxygen redistri-
bution within the A layer, it can be estimated that the
thickness of Ta2O5−x and TaO2−h layers layers are about
90 % and 10 % of the initial A layer thickness.
We remark that in the post forming state the Ta2O5−x
resistive layer is sandwiched between two reduced metal-
lic TaO2−h and TaO2−y layers, giving a quasi-symmetric
geometry, as shown in the zoomed sketch of Figure 6 (b),
which explains the HSL with TWL-like shape obtained
for symmetric stimuli (Figure 4 (c)), typical of symmetric
devices.
FIG. 5: Multilevel HSL recorded on a device with
11.3x103 µm2 area for (a) CW circulation and (b) CCW
circulation. In (a) Vmax is fixed while Vmin is changed;
in (b) Vmin is fixed while Vmax is changed.
The scenario proposed above is supported by the fol-
lowing analysis and fittings performed on the post form-
ing I-V curves, considering the γ = dLn(I)/dLn(V) pa-
rameter representation [26]. This method proved to be
useful in order to reveal the presence of a mixture of con-
duction mechanisms [27, 28], as usually found in oxide-
5
based devices [29 -- 31]. As can be observed in the insets
of Figures 3 (c) and (d), associated with both CCW and
CW HSLs of devices with 6.4x103 µm2 area, the depen-
dence of γ vs V1/2 indicates the existence of an ohmic
conduction (γ (cid:39) 1 at low voltages) in parallel with a mild
non-linear space charge limited current (SCLC) conduc-
tion (γ increases smoothly with V and remains <2).
FIG. 6: (a) Schematic diagram of the post-formed
device. The initial asymmetric geometry turns into a
quasi-symmetric Pt/TaO2−h/Ta2O5−x/TaO2−y/Pt
stack after forming; (b) Sketch of the post-forming
active region for the memristive behavior and scheme of
the 1D chain of nanodomains assumed for the numerical
simulations with the VEOV model. See text for details.
The simplest schematic circuit representation derived
from the voltage dependence of the γ parameter con-
sists in two parallel SCLC-ohmic conducting channels in
series. The experimental I-V curves were fitted by con-
sidering the equivalent circuit and related equations in-
cluded in the Supp. Information. The performed fits give
an excellent reproduction of the experimental I-V char-
acteristics as well as the associated γ vs. V1/2 curves,
as shown in Figures 3 (c) and (d). We notice that no
good fitting was obtained if Schottky emission is assumed
for the non-linear element of the equivalent circuit. This
fact, together with the already discussed non-rectifying
behavior observed in the I-V curves (Figures 3 (a) and
(b)) rule out the presence of Ta2O5−x in contact with
the left Pt electrode, giving consistency to the proposed
post-forming scenario.
1
The memristive and non-linear behaviours are nicely
reflected in the extracted values of the equivalent circuit
elements, listed in Table I. As can be observed, the elec-
tric response associated with the CCW HSL is dominated
by R1 and RN L
elements, as the R2 resistor is highly
metallized, shortcircuiting the RN L
non-linear element.
The remnant resistance measured al low-voltage bias (0.1
V) is essentially the equivalent resistance Req
1 between
R1 and RN L
. When the electric response switches to
the CW HSL, the situation is reversed: R1 shortcircuits
the non-linear RN L
element and the resistance of the de-
vice is dominated by the parallel combination between
2
1
1
TABLE I: Values in kΩ of the equivalent-circuit elements obtained by fitting the experimental I-V characteristics
with the corresponding equations (see Supp. Information). Values at 0.1 V and 1 V are indicated for the non-linear
SCLC element.
6
(1V) Req
≤0.8
≤0.1
8.4
5.1
70
25
-
-
2 RN L
2
(0.1V) RN L
2
(1V) Rrem
-
-
770
500
-
-
8
50
34.3
5.3
8.4
5.1
HSL STATE Req
1 RN L
1
(0.1V) RN L
1
CCW HR
LR
34.2
5.2
CW HR ≤0.1
≤0.1
LR
1700
620
-
-
R2 and the non-linear RN L
element. As can be seen
in Table I, the determined equivalent resistances at low
voltage (100mV) perfectly match the measured remanent
values Rrem.
2
1
The physical origin of the observed memristive behav-
ior can be understood in terms of OVs electromigration
between a (bulk) central zone, the Ta2O5−x layer, and
the interfacial zones close to TaO2−h and Ta2−y layers,
respectively. To achieve a behavior consistent with the
migration of OVs induced by the electric field, the CCW
HSL should be represented by an active SCLC-ohmic
zone, represented by the parallel R1 and RN L
circuit ele-
ments, which includes the left interface with TaO2−h plus
the central Ta2O5−x zone (we recall that SCLC is a bulk
conduction mechanism) where the memristive effects es-
entially modify this sector, while the ohmic element R2
represents the right interface with TaO2−y, character-
ized by a rather low resistance. On the contrary, for the
CW HSL, the increase of the amplitude of the negative
applied pulses switches the active element to a zone com-
prising the right interface with TaO2−y plus the central
bulk zone, both represented by R2 and RN L
, increasing
its resistance and concomitantly decreasing the resistance
of the R1 ohmic element, which now short-circuits the
SCLC RN L
element (see Supp. Information for a sketch
depicting these situations).
2
1
In the next section we will carefully address by means
of simulations the OV dynamics linked to this behavior.
NUMERICAL SIMULATIONS
Based on the previous description, we analize here the
electrical response obtained with the voltage enhanced
oxygen vacancies drift (VEOV) model, which allows sim-
ulating OV dynamics at the nanoscale,
linked to the
memristive behavior of our devices. We start by review-
ing the main equations, and we refer the readers to Ref-
erences 8 and 14, for further details.
We consider that the active memristive region is the
Ta2O5−x layer, which is assumed to consist of three
zones: a central (bulk) one labelled as C and two zones,
left (L) and right (R), localized close to the interfaces
with the metallic TaO2−h(y) layers, which present differ-
ent properties due to the presence of interfacial disorder
or defects. A sketch of the assumed memristive active
zone is shown in Figure 6 (b). For the modelling we sup-
pose a 1D chain of N = N L + N C + N R total sites.
The first N L sites correspond to the L layer, the fol-
lowing N C sites are assigned to the central C layer and
the last N R sites are linked to the R layer and we as-
sume N C > N L = N R. The sites, characterized by
their resisitivity, physically represent small domains of
(sub)nanoscopic dimensions with an initial OVs concen-
tration that we assume correspond to the post forming
state.
An universal feature of oxides is that their resistivity
is affected by the precise oxygen stoichiometry. In par-
ticular, Ta2O5 behaves as an n-type semiconductor in
which oxygen vacancies reduce its resistivity. Therefore,
we write the resistivity ρi of each site i as a linear de-
crescent (most simple) function of the local OV density
δi, namely:
ρi = ρ0(1 − Aiδi),
(1)
where ρ0 is the residual resistivity for negligible OV con-
centration (δi = 0) and Ai is a factor that changes be-
tween C, L and R layers, satisfying Aiδi < 1 ∀i. We con-
sider AL, AR < AC, which implies that the resistivity of
the interfacial zones is less sensitive to the presence of OV
due to the presence of disorder or defects. Also, the co-
efficients Ai can be taken for each layer either smoothly
dependent on the site position or as constants (as we
do for simplicity), without affecting the qualitative be-
haviour of the simulated results [18].
Following Equation(1), the total resistivity of the sys-
tem is given by:
ρ ≡ ρs − ρ0{ N L(cid:88)
ALδi − N−N R(cid:88)
i=1
i=N L+1
N(cid:88)
i=N−N R+1
ACδi −
being ρs ≡ N ρ0.
ARδi},
(2)
Given an external voltage V (t) applied at time t,
is updated for each simula-
the OV density at site i
tion step according to the rate probability pij = δi(1 −
δj) exp(−Vα + ∆Vi), for a transfer from site i to a near-
est neighbor j = i ±1. Notice that pij is proportional
step it is satisfied that(cid:80)N
to the OV density present at site i, and to the available
concentration at the neighbour site j.
In order to re-
strict the dynamics of OVs to the active region, we take
p01 = p10 = pN N +1 = pN +1N = 0. In addition as the to-
tal density of vacancies is conserved, for each simulation
i=1 δi = N δ0, being δ0 the OV
density per site for an uniform distribution (assumed as
known, see Supp. Information section).
In the Arrhenius factor, exp(−Vα + ∆Vi), ∆Vi is the
local potential drop at site i defined as ∆V i(t) = Vi(t)−
Vi−1(t) with Vi(t) = V (t)ρi/ρ and Vα the activation en-
ergy for vacancy diffusion in the absence of external stim-
ulus. We consider values of Vα = VL, VC and VR for the
L, C and R layers respectively. All the energy scales are
taken in units of the thermal energy kBT [18].
According to standard RS experiments, we chose the
stimulus V (t) as a linear ramp following the cycle 0 →
Vm1 → −Vm2 → 0. At each simulation time step tk
we compute the local voltage profile Vi(tk) and the lo-
cal voltage drops ∆V i(tk) and employing the probability
rates pij we obtain the transfers between nearest neigh-
boring sites. Afterwards the values δi(tk) are updated to
a new set of densities δi(tk+1), with which we compute, at
time tk+1, the local resistivities ρi(tk+1), the local volt-
age drops under the applied voltage V (tk+1), and finally
from Eq.(2) the total resistivity ρ(tk+1), to start the next
simulation step at tk+1. Notice that as de VEOV is a 1D
model, the conversion from resistivity to resistance is a
trivial scale factor. We refer to the Supp. Information
[18] for further details on the numerical values of the pa-
rameters employed in the simulations.
As it was already described in Sec., the negative form-
ing sets the device in a high resistance state, associ-
ated to which we define an initial OV density profile,
δi(t0) ∀i = 1..N , to start with the numerical implementa-
tion of the VEOV model. This initial OVs profile is cho-
sen to guarantee the post forming high resistance state,
in which the C zone contributes with the dominant resis-
tance, while L and R layers present both a lower resistiv-
ity due to a large density of OVs, and contribute little to
the total resistance (see Figure7 III) (a) for a scketch of
the initial post forming state).
Figure 7 panel I) shows the TWL-like HSL obtained
with the VEOV model simulations, for a symmetric volt-
age protocol V (t) following the cycle 0 → Vm1=2.1 V →
-Vm2 =2.1 V → 0. We start from an initial OVs distri-
bution compatible with the post forming state, scketched
in panel III) (a). This initial profile gives a resistance of
3KΩ, in perfect agreement with the reported experimen-
tal value. The positive stimulus moves OVs from the L
layer into the C layer (see panel III)(b)), initially tending
to reduce the resistance. However, as the applied voltage
is increased a strong electric field develops, moving OVs
further away to accumulate finally in the R layer. This
gives the RESET transition to the HR1 state displayed
in the HSL of panel I). The associated OVs density pro-
7
FIG. 7: I) Simulated HSL's. The TWL-like shape is
obtained for symmetric stimuli (black arrows indicate
the circulation) and squared CW and CCW HSLs for
asymmetric ones (blue and orange arrows indicate the
circulation in the respective HSLs). The HR and LR
states are labeled as described in the text; II) Colormap
of the OVs density per site (total density normalized to
1) for different resistance states, as labeled respectively
in the HSL of panel I); III) Scheme of the L, C and R
regions defined in Figure 6 (b), where the colors
qualitatively show the different total number of OVs in
each region for: (a) post forming (PF) state, (c)-(f)
HR1, LR1, HR2 and LR2 states, as labeled in panel I).
file along the active region is shown in the colour map of
panel II) whilst the total density of OVs in the L, C and
R regions respectively, is sketched in panel III) (c). The
HR1 state is mantained in the range Vm1 → 0, until the
polarity of the stimulus changes and consequently OVs
reverse their motion. For V = -1.4 V a SET transition
to the LR1 state takes place. The associated OVs distri-
bution is shown in panel II) (row labeled LR1) with the
color level in each of the L, C and R regions scketched
in panel III)(d), proportional to the total OV density re-
spectively. Notice that the R layer is mostly depleted
from OVs which accumulate in the C layer. In spite of
this accumulation, zone C still dominates the device resis-
tance in the low resistance state LR1. Increasing further
the intensity of the negative voltage, produces the elec-
tromigration of OV from the C to the L layer, almost
voiding of OVs the C layer and promoting a second RE-
SET transition to the HR2 state at a voltage V = −2 V
(see also row labeled HR2 in panel II) and panel III) (e)).
The OVs density profile for this case results slightly dif-
ferent from the previous one obtained for the HR1 state,
giving HR2 < HR1, in fully agreement with the experi-
mental TWL-like HSL. The HR2 state is mantained until
the voltage changes to possitive values and the OVs elec-
tromigrate from the L to the C layer, attaining the LR2
state at V = 1.2 V.
The agreement between the simulated and the experi-
mental TWL-like HSL show in Figure 4 (c) is remarkable,
denoting the predictive power of the VEOV model. For
completness, OV's density profiles numerically obtained
for other intermediate resistance states are also displayed
in different rows of panel II).
By changing to a non-symmetrical voltage protocol we
go from the TWL-like HSL to squared HSL's, emulating
the experimental response. The CW squared HSL (blue)
shown in Figure 7 panel (I) was obtained for a voltage
protocol 0 → Vm1=1.4 V→ −Vm2 =2.1 V → 0, whilts the
CCW HSL (orange) corresponds to a voltage protocol
0 → Vm1=2.1 V→ −Vm2=1.4 V → 0.
DISCUSSION AND CONCLUSIONS
The experiments and modeling described in the
present work show that it is possible to selectively ac-
tivate/deactivate two series memristive interfaces in or-
der to obtain squared CW, CCW or a TWL-like HSLs
just by controlling the maximum (Vmax >0) and min-
imum (Vmin <0) applied voltage excursions. This be-
haviour is independent of the nature of the used elec-
trodes: we found it in Pt/TaO2−h/Ta2O5−x/TaO2−y/Pt
devices, but also in similar systems where the top Pt elec-
trode was replaced by Au, as reported in Fig. S5 of the
Suppl. Information.
A key factor that allows the obtained selective re-
sponse is a forming process that changes the initial asym-
metric Ta-oxide bilayer, grown at different oxygen pres-
sures, to a quasi-symmetric TaO2−h/Ta2O5−x/TaO2−y
stack, excluding the Pt(Au)/TaO2−h(y) metal-metal in-
terfaces from contributing to the memristive behavior.
The finding of CCW and CW HSLs in a single quasi-
symmetric device goes against the common wisdom that,
for interface-related memristors, "(a)symmetric systems
give (a)symmetric electrical response". We propose,
based on our thourough analysis and modeling, that the
physical origin of the obtained electrical behavior relies
on OVs electromigration between bulk and interfacial
zones of the Ta2O5−x layer, sandwiched between metal-
lic TaO2−h(y). From the combination of experimental
results and simulations it can be stated that for CCW
HSL the OVs dynamics is constrained to C and L zones,
while for CW HSL the OV exchange is limited between
C and R layers. For symmetric stimuli all L, C and R
are involved and the memristive response becomes sym-
metric (TWL-like HSL).
It can be concluded that OV transfer between one in-
8
terface (i.e. R) and C zone does not start until the other
interface (i.e. L) is almost completely drained of OV.
This 2-steps process is the core of the observed behav-
ior. It is worth to remark that this rich phenomenology
was found for a system with a rather simple room tem-
perature fabrication process that includes only one oxide
-Ta-oxide- and one metal, suggesting that it might be
possible to take advantage of these properties at low cost
and with easy scability.
Regarding possible applications, the control of the
symmetry of the electric response allows optimizing, for
example, ON-OFF ratios for RRAM memories; however,
the results presented here have potentially more implica-
tions for the development of disruptive electronics such
as neuromorphic computing or novel logic devices. The
observation of multilevel resistance states indicates that
our devices can mimic the (analogic) adaptable synaptic
weight of biological synapsis. Importantly, the multilevel
states were found for both CCW and CW HSLs, indicat-
ing that the synaptic weight could be either potentiated
or depressed with electrical stimulus of the same polar-
ity. This might have implications for novel, beyond von
Neummann, devices [9].
EXPERIMENTAL SECTION
Ta-oxide thin films were deposited on commercial pla-
tinized silicon by pulsed laser deposition at room tem-
perature from a single phase Ta2O5 ceramic target. We
used a 2-steps deposition process with oxygen pressures
of 0.01 and 0.1mbar, resulting in 35nm and 15nm layers,
respectively, as shown in the main text. The laser fluence
was fixed at 1.5 J/cm2. For spectroscopic measurements,
we also deposited single TaOx layers at 0.01 and 0.1mbar
of O2, respectively. A Dual Beam Helios 650 was used
to acquire scanning electron microscopy images and pre-
pare FIB lamellas to observe cross-sections. High resolu-
tion scanning transmission electron microscopy was per-
formed using a FEI Titan G2 microscope at 300 kV with
probe corrector and in situ EELS spectrum acquisition
with a Gatan Energy Filter Tridiem 866 ERS. The sur-
face composition analysis obtained by X-Ray Photoelec-
tron Spectroscopy (XPS) used a standard Al/Mg twin
anode X-ray gun and a hemispherical electrostatic elec-
tron energy analyzer. The base chamber pressure was 10-
9 mbar. Top Pt electrodes, 20nm thick, were deposited
by sputtering and microstructured by optical lithogra-
phy. Top electrode sizes ranged between 6.4x103 and
78.5x103 µm2. Electrical characterization was performed
at room temperature with a Keithley 2612 source-meter
connected to a Suss probe station.
ACKNOWLEDGEMENTS
[12] T. H. Park, H. J. Kim, W. Y. Park, S. G. Kim, B. J.
9
We acknowledge support from INN-CNEA, UNCuyo
(06/C455), ANPCyT (PICT2014-1382, PICT2016-0867,
PICT2017-1836, PICT 2017-0984), CONICET (PIP
11220150100653CO) and Univ. Buenos Aires (UBACyT
20020170100284BA). M.A. also acknowledges financial
support of H2020-MSCA-RISE-2016 SPICOLOST Grant
No. 734187 to perform TEM studies at LMA-INA, Uni-
versity of Zaragoza. S.B. thanks technical support from
N. Cortes.
[1] D. Ielmini and R. Waser, Resistive Switching: From Fun-
damentals of Nanoionic Redox Processes to Memristive
Device Applications (Wiley-VCH, 2016).
[2] J. Borghetti, G. S. Snider, P. J. Kuekes, J. J. Yang, D. R.
Stewart, and R. S. Williams, Nature 464, 873 (2010).
[3] S. Yu, Neuro-inspiring computing using resistive synaptic
devices (Springer International Publishing, 2017).
[4] A. Prakash, D. Jana, and S. Maikap, Nanoscale Research
Letters 8, 418 (2013).
[5] M.-J. Lee, C. B. Lee, D. Lee, S. R. Lee, M. Chang, J. H.
Hur, Y.-B. Kim, C.-J. Kim, D. H. Seo, S. Seo, U.-I.
Chung, I.-K. Yoo, and K. Kim, Nature Materials 10,
625 (2011).
[6] K. X. Shi, H. Y. Xu, Z. Q. Wang, X. N. Zhao, W. Z. Liu,
J. G. Ma, and Y. C. Liu, Applied Physics Letters 111,
223505 (2017), https://doi.org/10.1063/1.5002571.
[7] G. H. Baek, A. R. Lee, T. Y. Kim, H. S. Im,
and
J. P. Hong, Applied Physics Letters 109, 143502 (2016),
https://doi.org/10.1063/1.4963884.
[8] M. J. Rozenberg, M. J. S´anchez, R. Weht, C. Acha,
and P. Levy, Phys. Rev. B 81,
F. Gomez-Marlasca,
115101 (2010).
[9] H. Zhang, Y. Sijung, S. Menzel, C.-F. W. D. J. H. C.
S. W. R. Funck, Carsten, and S. Hoffmann-Eifert, ACS
Applied Materials and Interfaces 10, 29766 (2018).
[10] C. La Torre, A. Kindsmller, D. J. Wouters, C. E. Graves,
G. A. Gibson, J. P. Strachan, R. S. Williams, R. Waser,
and S. Menzel, Nanoscale 9, 14414 (2017).
[11] T. H. Park, S. J. Song, H. J. Kim, S. G. Kim, S. Chung,
B. Y. Kim, K. J. Lee, K. M. Kim, B. J. Choi, and C. S.
Hwang, Scientific Reports 5, 15965 (2015).
Choi, and C. S. Hwang, Nanoscale 9, 6010 (2017).
[13] Y. Yang, P. Sheridan, and W. Lu, Applied Physics Let-
ters 100, 203112 (2012).
[14] N. Ghenzi, M. J. S´anchez,
and P. Levy, Journal of
Physics D: Applied Physics 46, 415101 (2013).
[15] N. Ghenzi, M. J. S´anchez, D. Rubi, M. J. Rozenberg,
C. Urdaniz, M. Weissman, and P. Levy, Applied Physics
Letters 104, 183505 (2014).
[16] X. Chen, N. J. Wu, J. Strozier,
Ig-
natiev, Applied Physics Letters 87, 233506 (2005),
https://doi.org/10.1063/1.2139843.
and A.
[17] D. S. Jeong, H. Schroeder, and R. Waser, Nanotechnol-
ogy 20, 375201 (2009).
[18] See Supporting Information for details..
[19] R. Simpson, R. G. White, J. F. Watts, and M. A. Baker,
Applied Surface Science 405, 79 (2017).
[20] V. Y.-Q. Zhuo, Y. Jiang, M. H. Li, E. K. Chua, Z. Zhang,
J. S. Pan, R. Zhao, L. P. Shi, T. C. Chong, and J. Robert-
son, Applied Physics Letters 102, 062106 (2013).
[21] H. Okamoto, Journal of Phase Equilibria 22, 515 (2001).
[22] S. P. Garg, N. Krishnamurthy, A. Awasthi,
and
M. Venkatraman, Journal of Phase Equilibria 17, 63
(1996).
[23] J. J. Yang, M.-X. Zhang, J. P. Strachan, F. Miao, M. D.
Pickett, R. D. Kelley, G. Medeiros-Ribeiro, and R. S.
Williams, Applied Physics Letters 97, 232102 (2010).
[24] Z. Wei, Y. Kanzawa, K. Arita, Y. Katoh, K. Kawai,
S. Muraoka, S. Mitani, S. Fujii, K. Katayama, M. Iijima,
T. Mikawa, T. Ninomiya, R. Miyanaga, Y. Kawashima,
K. Tsuji, A. Himeno, T. Okada, R. Azuma, K. Shi-
makawa, H. Sugaya, T. Takagi, R. Yasuhara, K. Horiba,
H. Kumigashira, and M. Oshima, Tech. Dig. - Int. Elec-
tron Devices Meet. 293 (2008).
[25] Y. C. Yang, C. Chen, F. Zeng, and F. Pan, Journal of
Applied Physics 107, 093701 (2010).
[26] C. Acha, Journal of Applied Physics 121 (2017),
10.1063/1.4979723.
[27] C. Acha, A. Schulman, M. Boudard, K. Daoudi,
and T. Tsuchiya, Applied Physics Letters 109 (2016),
10.1063/1.4955204.
[28] W. Acevedo Rom´an, C. Acha, M. Sanchez, P. Levy,
and D. Rubi, Applied Physics Letters 110 (2017),
10.1063/1.4975157.
[29] C. Acha, Journal of Physics D: Applied Physics 44,
345301 (2011).
[30] M. Cerchez, H. Langer, M. El Achhab, T. Heinzel, D. Os-
termann, H. Lder, and J. Degenhardt, Applied Physics
Letters 103, 033522 (2013).
[31] J. Blasco, N. Ghenzi, J. Su, P. Levy, and E. Miranda,
Microelectronics Reliability 55, 1 (2015).
|
1810.06946 | 1 | 1810 | 2018-10-16T12:10:38 | Dynamics of Ultrafast Laser Ablation of Water | [
"physics.app-ph"
] | Ultrafast laser ablation is an extremely precise and clean method of removing material, applied in material processing as well as medical applications. And due to its violent nature, it tests our understanding of the interplay between optics, condensed matter physics and fluid dynamics. In this manuscript, we experimentally investigate the femtosecond laser induced explosive vaporization of water at a water/gas interface on the micron-scale through several time-scales. Using time-resolved microscopy in reflection mode, we observe the formation of a hot electron plasma, an explosively expanding water vapor and a shockwave propelled into the surrounding gas. We study this fs-laser induced water vapor expansion dynamics in the presence of different atmospheres, i.e. Helium, air and tetrafluoroethane. We use the Sedov-Taylor model to explain the expansion of the water vapor and estimate the energy released in the process. | physics.app-ph | physics |
Dynamics of Ultrafast Laser Ablation of Water
JAVIER HERNANDEZ-RUEDA1,2 AND DRIES VAN OOSTEN1,3
1Debye Institute for Nanomaterials Science and Center for Extreme Matter and Emergent Phenomena,
Utrecht University, Princetonplein 5, 3584 CC Utrecht, The Netherlands
[email protected]
[email protected]
Abstract: Ultrafast laser ablation is an extremely precise and clean method of removing material,
applied in material processing as well as medical applications. And due to its violent nature,
it tests our understanding of the interplay between optics, condensed matter physics and fluid
dynamics. In this manuscript, we experimentally investigate the femtosecond laser induced
explosive vaporization of water at a water/gas interface on the micron-scale through several
time-scales. Using time-resolved microscopy in reflection mode, we observe the formation of a
hot electron plasma, an explosively expanding water vapor and a shockwave propelled into the
surrounding gas. We study this fs-laser induced water vapor expansion dynamics in the presence
of different atmospheres, i.e. Helium, air and tetrafluoroethane. We use the Sedov-Taylor model
to explain the expansion of the water vapor and estimate the energy released in the process.
, 2018
References
1. Claudiu A. Stan, Despina Milathianaki, Hartawan Laksmono, Raymond G. Sierra, Trevor A. McQueen, Marc
Messerschmidt, Garth J. Williams, Jason E. Koglin, Thomas J. Lane, Matt J. Hayes, Serge A. H. Guillet, Mengning
Liang, Andrew L. Aquila, Philip R. Willmott, Joseph S. Robinson, Karl L. Gumerlock, Sabine Botha, Karol Nass,
Ilme Schlichting, Robert L. Shoeman, Howard A. Stone, and Sébastien Boutet. Liquid explosions induced by x-ray
laser pulses. Nature Physics, 12:966, May 2016.
2. Yaron Silberberg. Laser science: Physics at the attosecond frontier. Nature, 414(6863):494, 2001.
3. Chris B. Schaffer, Nozomi Nishimura, Eli N. Glezer, Albert M.-T. Kim, and Eric Mazur. Dynamics of femtosecond
laser-induced breakdown in water from femtoseconds to microseconds. Opt. Express, 10(3):196 -- 203, Feb 2002.
4. B.D. Strycker, M.M. Springer, A.J. Traverso, A.A. Kolomenskii, G.W. Kattawar, and A.V. Sokolov. Femtosecond-
laser-induced shockwaves in water generated at an air-water interface. Opt. Express, 21(20):23772 -- 23784, Oct
2013.
5. Hao Zhang, SA Wolbers, DM Krol, JI Dijkhuis, and D Van Oosten. Modeling and experiments of self-reflectivity
under femtosecond ablation conditions. JOSA B, 32(4):606 -- 616, 2015.
6. Rafael R. Gattass and Eric Mazur. Femtosecond laser micromachining in transparent materials. Nature Photonics,
2:219, April 2008.
8.
7. Hao Zhang, Denise M Krol, Jaap I Dijkhuis, and Dries van Oosten. Self-scattering effects in femtosecond laser
nanoablation. Optics letters, 38(23):5032 -- 5035, 2013.
Javier Hernandez-Rueda, Jasper Clarijs, Dries van Oosten, and Denise M Krol. The influence of femtosecond laser
wavelength on waveguide fabrication inside fused silica. Applied Physics Letters, 110(16):161109, 2017.
9. Uday K. Tirlapur and Karsten König. Targeted transfection by femtosecond laser. Nature, 418:290, July 2002.
10. Norbert Linz, Sebastian Freidank, Xiao-Xuan Liang, Hannes Vogelmann, Thomas Trickl, and Alfred Vogel.
Wavelength dependence of nanosecond infrared laser-induced breakdown in water: Evidence for multiphoton
initiation via an intermediate state. Phys. Rev. B, 91:134114, Apr 2015.
11. Hazel A. Collins, Mamta Khurana, Eduardo H. Moriyama, Adrian Mariampillai, Emma Dahlstedt, Milan Balaz,
Marina K. Kuimova, Mikhail Drobizhev, Victor X. D. Yang, David Phillips, Aleksander Rebane, Brian C. Wilson,
and Harry L. Anderson. Blood-vessel closure using photosensitizers engineered for two-photon excitation. Nature
Photonics, 2:420, May 2008.
12. Todd Ditmire, J Zweiback, VP Yanovsky, TE Cowan, G Hays, and KB Wharton. Nuclear fusion from explosions of
femtosecond laser-heated deuterium clusters. Nature, 398(6727):489 -- 492, 1999.
13. M Hentschel, R Kienberger, Ch Spielmann, Georg A Reider, N Milosevic, Thomas Brabec, Paul Corkum, Ulrich
Heinzmann, Markus Drescher, and Ferenc Krausz. Attosecond metrology. Nature, 414(6863):509, 2001.
14. Mehmet Fatih Yanik, Hulusi Cinar, Hediye Nese Cinar, Andrew D. Chisholm, Yishi Jin, and Adela Ben-Yakar.
Functional regeneration after laser axotomy. Nature, 432:822, December 2004.
15. Samuel X. Guo, Frederic Bourgeois, Trushal Chokshi, Nicholas J. Durr, Massimo A. Hilliard, Nikos Chronis, and
Adela Ben-Yakar. Femtosecond laser nanoaxotomy lab-on-a-chip for in vivo nerve regeneration studies. Nature
Methods, 5:531, April 2008.
16. Alfred Vogel and Vasan Venugopalan. Mechanisms of pulsed laser ablation of biological tissues. Chemical Reviews,
103(2):577 -- 644, 2003. PMID: 12580643.
17. J. Charles Williamson, Jianming Cao, Hyotcherl Ihee, Hans Frey, and Ahmed H. Zewail. Clocking transient chemical
changes by ultrafast electron diffraction. Nature, 386:159, March 1997.
18. SK Sundaram and E Mazur. Inducing and probing non-thermal transitions in semiconductors using femtosecond
laser pulses. Nature Materials, 1:217 -- 224, 2002.
19. K. Sokolowski-Tinten, J. Bialkowski, A. Cavalleri, D. von der Linde, A. Oparin, J. Meyer-ter Vehn, and S. I. Anisimov.
Transient states of matter during short pulse laser ablation. Phys. Rev. Lett., 81:224 -- 227, Jul 1998.
20. A Douhal, SK Kim, and AH Zewail. Femtosecond molecular dynamics of tautomerization in model base pairs.
21. D von der Linde, K Sokolowski-Tinten, and J Bialkowski. Laser -- solid interaction in the femtosecond time regime.
Nature, 378(6554):260, 1995.
Applied Surface Science, 109-110:1 -- 10, 1997.
May 2004.
22. C. W. Siders, A. Cavalleri, K. Sokolowski-Tinten, Cs. Toth, T. Guo, M. Kammler, M. Horn von Hoegen, K. R.
Wilson, D. von der Linde, and C. P. J. Barty. Detection of nonthermal melting by ultrafast x-ray diffraction. Science,
286:1340, November 1999.
23. Vasily V. Temnov, Klaus Sokolowski-Tinten, Ping Zhou, and Dietrich von der Linde. Ultrafast imaging interferometry
at femtosecond-laser-excited surfaces. J. Opt. Soc. Am. B, 23(9):1954 -- 1964, Sep 2006.
24. Filkorn T Sarayba M. Nagy Z, Takacs A. Initial clinical evaluation of an intraocular femtosecond laser in cataract
surgery. J Refract Surg. 2009; 25, 2009.
25. C Sarpe, J Köhler, T Winkler, M Wollenhaupt, and T Baumert. Real-time observation of transient electron density in
water irradiated with tailored femtosecond laser pulses. New Journal of Physics, 14(7):075021, 2012.
26. Ronald R. Krueger, Jerzy S. Krasinski, Czeslaw Radzewicz, Karl G. Stonecipher, and J. James Rowsey. Photography
of shock waves during excimer laser ablation of the cornea: Effect of helium gas on propagation velocity. Cornea,
12(4):330 -- 334, 1993.
27. Tibor Juhasz, Xin H. Hu, Laszlo Turi, and Zsolt Bor. Dynamics of shock waves and cavitation bubbles generated by
picosecond laser pulses in corneal tissue and water. Lasers in Surgery and Medicine, 15(1):91 -- 98, 1994.
28. C. Sarpe-Tudoran, A. Assion, M. Wollenhaupt, M. Winter, and T. Baumert. Plasma dynamics of water breakdown at
a water surface induced by femtosecond laser pulses. Applied Physics Letters, 88(26):261109, 2006.
29. B. Rethfeld. Unified model for the free-electron avalanche in laser-irradiated dielectrics. Phys. Rev. Lett., 92:187401,
30. B. Rethfeld. Free-electron generation in laser-irradiated dielectrics. Phys. Rev. B, 73:035101, Jan 2006.
31. A. Vogel, S. Busch, and U. Parlitz. Shock wave emission and cavitation bubble generation by picosecond and
nanosecond optical breakdown in water. The Journal of the Acoustical Society of America, 100(1):148 -- 165, 1996.
32. I. Apitz and A. Vogel. Material ejection in nanosecond er:yag laser ablation of water, liver, and skin. Applied Physics
A, 81(2):329 -- 338, Jul 2005.
33. Xianzhong Zeng, Xianglei Mao, Samuel S. Mao, Sy-Bor Wen, Ralph Greif, and Richard E. Russo. Laser-induced
shockwave propagation from ablation in a cavity. Applied Physics Letters, 88(6):061502, 2006.
34. Sigurdur T Thoroddsen, K Takehara, TG Etoh, and C-D Ohl. Spray and microjets produced by focusing a laser pulse
into a hemispherical drop. Physics of Fluids, 21(11):112101, 2009.
35. Yoshiyuki Tagawa, Nikolai Oudalov, Claas Willem Visser, Ivo R Peters, Devaraj van der Meer, Chao Sun, Andrea
Prosperetti, and Detlef Lohse. Highly focused supersonic microjets. Physical review X, 2(3):031002, 2012.
36. Alexander L Klein, Wilco Bouwhuis, Claas Willem Visser, Henri Lhuissier, Chao Sun, Jacco H Snoeijer, Emmanuel
Villermaux, Detlef Lohse, and Hanneke Gelderblom. Drop shaping by laser-pulse impact. Physical review applied,
3(4):044018, 2015.
37. Silvestre Roberto Gonzalez Avila and Claus-Dieter Ohl. Fragmentation of acoustically levitating droplets by
laser-induced cavitation bubbles. Journal of Fluid Mechanics, 805:551 -- 576, 2016.
38. Thao T.P. Nguyen, Rie Tanabe, and Yoshiro Ito. Comparative study of the expansion dynamics of laser-driven plasma
and shock wave in in-air and underwater ablation regimes. Optics & Laser Technology, 100:21 -- 26, 2018.
39. Alfred Vogel, J Noack, G Hüttman, and G Paltauf. Mechanisms of femtosecond laser nanosurgery of cells and tissues.
Applied Physics B, 81(8):1015 -- 1047, 2005.
40. JM Liu. Simple technique for measurements of pulsed gaussian-beam spot sizes. Optics letters, 7(5):196 -- 198, 1982.
41. LV Keldysh et al. Ionization in the field of a strong electromagnetic wave. Sov. Phys. JETP, 20(5):1307 -- 1314, 1965.
42. J. Hernandez-Rueda, D. Puerto, J. Siegel, M. Galvan-Sosa, and J. Solis. Plasma dynamics and structural modifications
induced by femtosecond laser pulses in quartz. Applied Surface Science, 258(23):9389 -- 9393, 2012. EMRS 2011
Spring Symp J: Laser Materials Processing for Micro and Nano Applications.
43. J. Hernandez-Rueda, J. Siegel, M. Galvan-Sosa, A. Ruiz de la Cruz, M. Garcia-Lechuga, and J. Solis. Controlling
ablation mechanisms in sapphire by tuning the temporal shape of femtosecond laser pulses. J. Opt. Soc. Am. B,
32(1):150 -- 156, Jan 2015.
44. N. Arnold, J. Gruber, and J. Heitz. Spherical expansion of the vapor plume into ambient gas: an analytical model.
Applied Physics A, 69(1):S87 -- S93, Dec 1999.
45. Fumitaka Mafuné, Jun-ya Kohno, Yoshihiro Takeda, Tamotsu Kondow, and Hisahiro Sawabe. Formation and
size control of silver nanoparticles by laser ablation in aqueous solution. The Journal of Physical Chemistry B,
104(39):9111 -- 9117, 2000.
46. Christian Wagner and Noreen Harned. Euv lithography: Lithography gets extreme. Nature Photonics, 4(1):24, 2010.
Introduction
1.
In many applications in research [1 -- 5], technology [6 -- 8] and healthcare [9 -- 11], ultrafast lasers
are employed to cut and remove material as well as to locally modify the chemical, structural and
optical properties of the target. In even more extreme examples, pulsed lasers are used to trigger
nuclear fusion [12] and to generate high harmonics in attosecond science [13] or more practically,
to study nerve regeneration in vivo after fs-laser axotomy [14, 15]. The unprecedented spatial
resolution achieved in all these studies is partly due to the simultaneous ultrafast character and
non-linear nature of the light-matter interaction, which leads to a reduced heat affected zone and
to minimal collateral damage when compared to the outcome achieved using longer pulses [16].
These outstanding features allow researchers not only to accurately modify materials but also to
investigate the fast and ultrafast mechanisms involved during the process of modification, such
as phase transitions or bio-chemical and chemical reactions [1, 17 -- 19]. In this way, ultrashort
laser pulses, with a duration that ranges from a few to hundreds of femtoseconds (1 fs = 10−15 s),
can be used to probe the ultrafast chain of processes triggered by an excitation laser pulse in a
so-called pump and probe system [17,19 -- 23].
Particularly spectacular and very relevant for biological applications and laser-based surgery,
is the ablation of aqueous media [3, 9, 15, 16, 24]. Crucial steps in this inherently multi-scale
process are the absorption of laser-energy by the aqueous medium, resulting in local heating,
followed by the evaporation of the liquid, which in turn does work against the tissue in which it
is embedded [1,4,10,16,24 -- 28]. The use of fs-laser pulses is particularly interesting for such
applications, as the absorption in that case is very nonlinear and therefore is limited to a region
that is typically smaller than the focal volume [19,22,29]. The extreme optical nonlinearity makes
the description of the absorption of the ultra-short laser pulse very challenging, but on the other
hand, leads to a very attractive separation of timescales [29, 30]. In the first picosecond, a hot
electron plasma is created. Within the first 10 picoseconds, thus after the laser pulse is gone, this
plasma equilibrates in temperature with the surrounding liquid. As a result, the liquid becomes
superheated and explosively evaporates. The interaction dynamics of pulsed lasers with water
has been investigated using lateral imaging via time-resolved shadowgraphy or time-resolved
scatterometry, for instance looking at the propagation of laser-induced shockwaves and cavitation
bubbles [3,4,31 -- 38]. Lateral imaging provides a wealth of information of the aftermath for long
time-delays (ns-µs), at the expense of losing lateral spatial resolution to study the microscopic
features of the initial electron plasma under strong focusing conditions [3], relevant for instance
to fs-laser cell surgery and neurosurgery [9,39].
In this work, we study femtosecond laser ablation at a water/gas interface using time-resolved
imaging with submicron optical resolution in reflection mode, from 10 femtosecond to 100
nanosecond time-scales. The experiment is carried out inside three different gaseous atmospheres
with increasing molecular weights and densities at room temperature, i.e. Helium (4 g/mol), air
(18 g/mol) and tetrafluoroethane (102 g/mol). The already mentioned separation of time-regimes
allows us to compartmentalize the multi-scale problem and separately observe the light-water
interaction (fs-ps), the evaporation process (10 ps) and the supersonic vapor expansion dynamics
(ns). We interpret our experimental observations with a Sedov-Taylor model to extract an estimate
of the energy carried by the expanding vapor.
2. Experimental procedure
2.1. Experimental setup
Fig.1 (a) shows a detailed scheme of the collinear pump and probe setup. The laser source used
during the experiments is a femtosecond regenerative amplifier (Hurricane, Spectra-Physics) that
produces 150 fs laser pulses at a wavelength of 800 nm. Using a λ/2 waveplate and a polarized
beam cube, we split the 800 nm laser pulse into two sub-pulses. The most energetic pulse runs over
a fixed delay line to the microscope objective while the weaker fraction runs over an automated
delay line to tune the pump-probe delay time up to a maximum of 1.4 ns (Newport Co.). For
larger delays, we introduce longer detours in the probe optical path to obtain delays up to 100 ns.
Before the delay line, a BBO (beta barium borate) crystal is used to frequency double the probe
pulse and an edge filter (F) is used to block the remaining 800 nm light. The 400 nm probe light
is then spatially and temporally overlapped with the 800 nm pump light in a pellicle beam splitter
(PBS) before both collinearly enter an infinity corrected microscope objective (Nikon CFI60,
100X, NA = 0.8). Additional lenses in both optical paths are used to ensure that the 800 nm
pump (one-to-one telescope, T0) is strongly focused by the objective, while the 400 nm probe is
focused (L) in the back focal plane of the objective, resulting in wide-field illumination suitable
for imaging the surface of the sample over an area of 45 × 45 µm2 onto the EMCCD camera chip
(Andor, iXon 885).
Figure 1. (a) Schematic of the femtosecond time-resolved ablation setup. (b) Detail of the
sample cuvette.
The reflected light is collected by the objective and used to image the water/air interface
through a 300 mm tube lens (TL) as shown in Fig.1 (a). A filter (F) that only transmits the
400 nm light (∆λFWHM = 10 nm, λ0 = 400 nm) is used to prevent 800 nm pump light and plasma
emission from reaching the camera. The optical path between the tube lens/filter and the camera
chip is tubed in order to minimize the collection of stray ambient light. The delay control unit
of the Pockels cells is synchronized with two optomechanical shutters (S1, S2) to ensure single
shot experiments either with the probe pulse only (S1 open) or with both pump and probe pulses
(both S1 and S2 open). For each temporal delay, several images are taken, each corresponding to
a single shot of the laser while both pump and probe pulses are present. Additionally, several
reference images are acquired. It is worth noting that unlike in the case of solid targets, the water
surface self-restores a few milliseconds after strong laser excitation. This prevents incubation
effects and allows us to record several pump-probe images at the same spot without re-focusing
or moving the sample to a fresh area. The beam waist (1/e2) of the focused pump beam at the
sample surface was calibrated to be 3.1 µm using Liu method [40].
2.2. Sample preparation
In the experiment we use milli-Q demineralised water in a 25 mL beaker. This beaker is placed
inside a container in which the microscope objective is introduced through a closely fitting hole
Optical delay lineT1Pump (800 nm)T2BBOFλ/2Energy control2PDCross-correlationTLFλ/2λ/2PBCPBCS1S2LProbe (400 nm)Pump detourSample cuvetteEMCCDT0Fs-laserDelayprobepumpO-ringsMicroscope objectiveSaturated water vaporWaterGas returnGas supply(a)(b) in the lid that is sealed by using rubber O-rings (see figure 1 (b)). The isolated air inside is
then allowed to saturate with water vapor for a few minutes, to limit the speed with which the
water level lowers due to evaporation to less than 1 µm/hour, reducing the need to periodically
refocus the surface of the target. The gaseous atmosphere of the cuvette can be controlled using a
circulating system with a supply and an exhaust. For the experiments we use three different gasses,
namely Helium, air and 1,1,1,2 tetrafluoroethane. Table I shows relevant physical properties of
the gases.
TableI: Relevant gas properties at standard temperature and pressure.
1,1,1,2 Tetrafluoroethane
Air
Helium
Molar Mass (g/mol) Density (mg/cm3)
102.03
18.00
4.00
4.25
1.20
0.18
Speed of sound (m/s)
168.41
343.21
972.00
Figure 2. (a) Typical transient differential reflectivity maps of the water/air interface obtained
for a pump pulse fluence of 25 J/cm2. The images share the same lateral scale (20x20 µm2)
and are represented using the same color scale (which saturates the first four images). In the
first few picoseconds, we see a strong increase of the reflectivity. After 2 picoseconds, we
see the increase in reflectivity turn into a ring, which fades for longer pump-probe delays.
Starting at approximately 10 picoseconds, we see a reduced differential reflectivity starting
in the center. On the timescale of 100 ps, the dark region radially expands and a bright rim
develops around it. (b) Time-line during the ablation of water. The time-delays, at which the
images in (a) are taken, are indicated by the arrows labelled a1 to a18.
3. Results
3.1. Ultrafast ablation of water through the time-scales
Using the experimental setup, we investigate the transient reflectivity of a water/gas interface
during ultrafast laser ablation of water upon tight focusing conditions, from 10 fs to 100 ns.
100 ps500 ps1000 ps20 ps1 ps2 ps5 ps50 ps0.5 ps10 ps5 μm5000 ps10000 ps(a)0.2 ps200 ps2000 ps50000 ps20000 ps100000 ps femtosecondspicosecondsnanosecondsasμsfspsnsa1a3μsLaser pulsea7a9a11a13a15a5a17a18(b) Typical examples of such images, measured in an air atmosphere, are shown in Fig. 2 (a) whose
time delays are chosen to be equally spaced in a logarithmic scale, as presented in Fig. 2 (b).
Here, three main regimes are observed: excitation and relaxation of a dense electron plasma
(first row), evaporation onset (second row) and water vapor expansion (third row). We observe a
delay of few tens of picoseconds before the laser-induced water vapor noticeably starts to expand
laterally, which agrees with the expansion dynamics observed inside bulk water [3] and at the
surface of a water jet [28] under similar focusing conditions.
Figure 3. (a)-(e) Transient relative reflectivity maps representative of the different time
regimes. (f)-(j) Radial average of the above images. The blue curves share the vertical scale
of (f), the red curves share the vertical scale of (j). The color-bar indicating the grey-scale of
the images (a)-(e) is attached to the vertical axes of the graphs (f)-(j). (k)-(o) Schematics of
the physical processes, where (f)-(h) extends from 0 fs to 40 ps, (i)-(j) spreads from 40 ps
to 100 ns. Note that for a given column the images, profiles and schematics have the same
lateral scale.
In Figs. 3 (a)-(e), we show images representative of the different time-separated regimes,
combined with their radial averages (f)-(j) and schematics to discuss the corresponding physical
processes (k)-(o). Initially the energy of the laser is coupled into the irradiated water via strong
field ionization [41], i.e. multiphoton and tunneling ionization. Afterwards, seed electrons that are
excited via non-linear ionization processes gain kinetic energy by means of inverse bremsstrahlung
leading to impact ionization [10,28,30]. Consequently, during the first picosecond, we observe a
strong increase of the surface reflectivity that we atribute to the formation of a well-localized
dense electron plasma, as illustrated in Fig. 3 (k). The maximum reflectivity increase that we
observe experimentally is ∆R = 0.16, whereas the reflectivity of unexcited water is Ro = 0.02.
According to the Drude model for a free electron gas this corresponds to an electron density
on the order of 1022 cm−3. These findings are consistent with the transient reflectivity contrast
-20-1001020 Axis position (m)Bow-ShockOblique-ShockθwatervapourDensity-20-10010200.00.51.01.52.02.5 Relative reflectivity (a.u.) Axis position (m)-4-2024 Axis position (m)Water10 μm1 ns16 ps2 ps400 fs10 ns10 μm2 μm10 μmLasere-plasma diskEnergyredistributionPlane-waveexpansionShockwaveCone-likeexpansionLaser excitation50 μmVapourAir2 μm2 μm-4-2024024681012 Relative reflectivity (a.u.) Axis position (m)-4-2024 Axis position (m)(a)(b)(c)(d)(e)(f)(g)(h)(i)(j)(k)(l)(m)(n)(o)0 fs to 40 ps40 ps to 100 ns10 μm50 μm10 μm reported for solid crystalline and glassy dielectric targets [42, 43], whose electonic and linear
optical properties are comparable to those of water. After 2 picoseconds, the reflectivity in the
center starts to drop. We attribute this reduction in reflectivity to the fact that the water in the
center starts to evaporate, locally reducing the density of both the water and the electron plasma,
as illustrated in Fig. 3 (l). We therefore use the appearance of this reduction to estimate the
ablation threshold to be 8.1 J/cm2, considering the calibrated 1/e2 beam waist (see previous
section and [40]) and the Gaussian distribution of the laser. Curiously, we reproducibly observe
a periodic azimuthal structure in the ring of increased reflectivity for delays that range from 8
ps to 40 ps, as shown in Fig.3 (c). The orientation of this periodic structure remains unaltered
for different irradiation experiments at a given time-delay, indicating a deterministic behaviour.
As of yet, we have no explanation of the cause of this structure. As time goes on, water further
away from the center also evaporates, expanding upwards, as illustrated in Fig.3 (m), causing the
corrugated ring of increased reflectivity to fade away for longer pump-probe delays. Starting at
approximately 10 picoseconds, we see an overall reduced differential reflectivity in the affected
area. We attribute this to the fact that the hot water vapor above the water/air interface can act
as a local anti-reflection coating that absorbs the light of the probe beam [3, 28]. From 100
ps onwards, we observe that the area of reduced reflectivity radially expands and develops a
sharp bright rim. We can understand that this happens when the expanding vapor cloud becomes
larger in radius than the initially laser-excited area, as illustrated in Fig.3 (n). The bright edge in
Figs.3 (d),(e),(i),(j) can be interpreted as the projection of the contact discontinuity between the
supersonically expanding vapor and the surrounding air. The fact that the contact discontinuity is
sharply visible means it must be widest near the focal plane, i.e. the water/gas interface. This
strongly suggests that the expansion is cone-like, as illustrated in Fig.3 (n),(o), which in turn
indicates a highly supersonic expansion. This is supported by the fact that we observe a shockwave
in the surrounding air, as can be seen in Figs. 3 (e),(j).
3.2. Atmosphere influence during ultrafast ablation of water
The expansion of the water vapor also depends on the atmosphere that surrounds the irradiated
area. Fig. 4 (a) presents snapshots of the transient reflectivity during the ablation of water in the
presence of different gases, from 500 ps to 10 ns. As explained in Fig. 3 the dark disk in the
max ≈ 5.5 µm) is attributed to the rapidly expanding water vapor, whereas the outermost
center (rv
max ≈ 24 µm) corresponds to the shockwave propelled in the surrounding
dark and bright ring (rs
gas (see white arrows in Fig. 4 (a) "air" at 10 ns). The increment in the lateral size as well as
the sharpness of the vapor front are clearly influenced by the surrounding atmosphere. In the
presence of Helium and Air, the vapor resembles a homogeneous dark circle with a distinct edge,
while in a tetrafluoroethane atmosphere, the dark area is heterogeneous and lacks a sharp edge.
Although the images are similar in appearance, Fig. 4 (b) shows that the radial expansion in
air achieves a slightly larger radius than in He for long delays up to 100 ns. For the first two
nanoseconds, the same expansion velocity is observed in the presence of both gases (≈ 810 m/s),
which corresponds to a supersonic expansion for air (vs = 343 m/s, vs is the speed of sound) and
a subsonic expansion for Helium (vs = 972 m/s). This initial expansion speed explains why a
shockwave is experimentally observed in an air atmosphere but not in a Helium environment.
After 2 ns, the radial increase slows down more rapidly in a He environment than in the air
environment. We use a Sedov-Taylor formula to estimate the energy released during the expansion
of the water vapor in the presence of air and He (see Fig. 4 (b))
(cid:19)1/5
(cid:18) 3E
π ρ
r(t, E) =
t2/5
(1)
where ρ is the density of water and E stands for the energy. Considering a hemispherical
expansion, we estimate the energies released in an air and He atmospheres to be 618 pJ and
360 pJ, respectively. To test the validity of using the Sedov-Taylor model, we additionaly fit
a power-law formula r ∝ tα, retrieving an exponent of 0.30 and 0.26 for air and Helium,
respectively, as shown in Fig. 4 (b). This shows that the expansion is not fully self-similar. This
lack of self-similarity can be qualitatively understood, as we expect that in both atmospheres,
the superheated water initially undergoes a vertical supersonic expansion [39,44]. During the
expansion, the expanding vapor gathers mass by sweeping up the gas from the atmosphere,
eventually leading to a transition from a vertical expansion to a more isotropic expansion. In
an atmosphere with a low mass density (ρair ≈ 7ρHe), this transition will occur later. Thus the
system will favour vertical expansion over lateral expansion as compared to a system with an
atmosphere with a high mass density, in correspondence with the results shown in Fig. 4 (b).
Figure 4. (a) Representative snapshots of the ablation process during vapor expansion in
the presence of tetrafluoroethane, air and Helium. We use the same laser fluence employed
in section 3.1. (b) Radius of the front bright rim as a function of pump-probe delay for
experiments carried out in a water/air (dark blue) and a water/He (orange) interfaces. Each
experiment averages the radius of 20 images. The inset shows the data in a log-log scale. (c)
Radius of the shockwave in air (dark blue) and tetrafluoroethane (black) as function of time.
Fig. 4 (a) also shows that a gaseous atmosphere with higher density (ρTFE ≈ 4 · ρair ≈ 24 · ρHe,
see table I in the experimental section) presents a shock with higher optical visibility, i.e.
tetrafluoroethane. The graph in Fig. 4 (c) shows that the shock radius as a function of time has a
similar behaviour in both atmospheres.
4. Conclusion
In conclusion, we experimentally study the ultrafast laser ablation at a water/gas interface by
using time-resolved microscopy. We present the changes in the transient reflectivity for several
time-scales, from 10 fs to 100 ns. Overall, our work explores the initial laser-water interaction
(femtoseconds), the intermediate extreme thermodynamic state of the system (picoseconds) and
the subsequent compressible fluid dynamics (nanoseconds). As an outlook we propose to further
link and explore the long-run mechanical behaviour, i.e. surface waves, which will require delays
far into the microsecond regime. We additionally test the influence of the atmosphere during
the ablation dynamics, finding that the propulsion of a shock in the surrounding gas depends
on its speed of sound and molecular weight. Moreover, we find that the atmosphere influences
the way the laser induced vapor expands during the first 100 ns, following a Sedov-Taylor power
law expansion with different exponents, which indicates the process is not self similar. The
understanding of this concatenation of physical processes has high impact and tremendous
potential in the field of laser nano-surgery (i.e. cell, ocular and neuro surgery) [9,39], since they
are behind the behaviour of photomechanical damage. As the ablation of water involves less
024681048121620242832Radius (m) Shockwave in Air Shockwave in Tetrafluoroethane Delay time (ns)(a)500 ps2 ns5 nsAirHeliumTetrafluoroethane7 ns10 ns10 μm𝑟𝑚𝑎𝑥𝑣𝑟𝑚𝑎𝑥𝑠11010010 020406080100246810121416 Water in Air Water in He Power law fit with exponent 0.30 Power law fit with exponent 0.26 Radius (m)Delay time (ns)(c)(b) phase transitions than the ablation of solid samples, our study can contribute further research on
the ultrafast laser ablation of more complex systems, such as nanoparticle synthesis via ablation
of inmersed targets [45] or EUV-light generation for nanolithograpy via tin droplet ablation [46].
Funding
European Commission, Horizon 2020, Marie Skłodowska-Curie Action Individual Fellowship
(703696 ADMEP).
Acknowledgments
The authors thank Denise Krol, Hanneke Gelderblom, Allard Mosk and Ingmar Swart for fruitful
discussions. The authors also thank Paul Jurrius, Cees de Kok and Dante Killian for discussions
and technical assistance.
|
1803.10821 | 1 | 1803 | 2018-03-28T19:25:10 | Quantitative Measurements of Nanoscale Permittivity and Conductivity Using Tuning-fork-based Microwave Impedance Microscopy | [
"physics.app-ph",
"cond-mat.mes-hall",
"cond-mat.mtrl-sci"
] | We report quantitative measurements of nanoscale permittivity and conductivity using tuning-fork (TF) based microwave impedance microscopy (MIM). The system is operated under the driving amplitude modulation mode, which ensures satisfactory feedback stability on samples with rough surfaces. The demodulated MIM signals on a series of bulk dielectrics are in good agreement with results simulated by finite-element analysis. Using the TF-MIM, we have visualized the evolution of nanoscale conductance on back-gated $MoS_2$ field effect transistors and the results are consistent with the transport data. Our work suggests that quantitative analysis of mesoscopic electrical properties can be achieved by near-field microwave imaging with small distance modulation. | physics.app-ph | physics | Quantitative Measurements of Nanoscale Permittivity and Conductivity Using
Tuning-fork-based Microwave Impedance Microscopy
Xiaoyu Wu1, Zhenqi Hao2, Di Wu1, Lu Zheng1, Zhanzhi Jiang1, Vishal Ganesan1, Yayu Wang2,3,
Keji Lai1
1 Department of Physics, University of Texas at Austin, Austin TX 78712, USA
2 State Key Laboratory of Low Dimensional Quantum Physics, Department of Physics, Tsinghua
University, Bejing 100084, China
3 Collaborative Innovation Center of Quantum Matter, Beijing 100084, China
Abstract
We report quantitative measurements of nanoscale permittivity and conductivity using tuning-fork
(TF) based microwave impedance microscopy (MIM). The system is operated under the driving
amplitude modulation mode, which ensures satisfactory feedback stability on samples with rough
surfaces. The demodulated MIM signals on a series of bulk dielectrics are in good agreement with
results simulated by finite-element analysis. Using the TF-MIM, we have visualized the evolution
of nanoscale conductance on back-gated MoS2 field effect transistors and the results are consistent
with the transport data. Our work suggests that quantitative analysis of mesoscopic electrical
properties can be achieved by near-field microwave imaging with small distance modulation.
1
Introduction
I.
Near-field microwave microscopy is a rapidly evolving technique that can spatially resolve
material properties at a length scale far below the freespace wavelength of microwave radiation1,2.
Early implementations of the microscope utilized a small aperture at a microwave cavity3,4 or a
needle-like probe coupled to a microwave resonator5-8. Because of the lack of tip-sample distance
control, these designs are intrinsically susceptible to tip damage and difficult for applications in
nanoscale quantitative measurements. In the past decade, research in this field has taken advantage
of the standard atomic-force microscopy (AFM) feedback control9-12. The AFM-compatible
probes, either shielded cantilevers with a pyramidal tip9,10 or unshielded cantilevers with an etched
high-aspect-ratio metal tip11,12, are connected to customized electronics9,10 or vector network
analyzers11,12 through an impedance-match section. The systems demonstrate very high sensitivity
(down to sub-aF at 1 – 20 GHz) in impedance detection, thus termed microwave impedance
microscopy (MIM)9,13, and good spatial resolution down to 10 – 100 nm. As a result, the MIM and
other similar tools have found tremendous applications in condensed matter physics14-17, material
science18-20, photovoltaics21-23, device engineering24,25, and biological science26,27, with an ever-
increasing list in the near future.
Along with the scientific exploration, the MIM technique itself has also undergone continuous
improvements over the years. Most MIM experiments to date are based on contact-mode
operation9,10, during which the tip wearing is inevitable. Since the signal level is strongly affected
by the condition of the tip apex, quantification of the contact-mode MIM is very difficult and
extensive calibration process is needed throughout the measurements. The issue of tip degradation
can be partially solved by using tapping-mode MIM28, although the spatial resolution is usually
compromised by the large dithering amplitude (50 – 100 nm) in cantilever-based systems. In this
regard, the recently developed tuning-fork (TF) based MIM29-31 with etched metal tips provides an
elegant solution to the problem. Quartz TFs with small vibration amplitudes (< 10 nm) are widely
used as the feedback elements in scanning probe microscopy32,33, especially under cryogenic
environments. In addition to the self-sensing capability that preserves the tip condition, TF-MIM
with signal modulation also introduces other advantages. For instance, the MIM electronics in
contact mode are only sensitive to the relative electrical contrast between the material of interest
2
and a background region13. In the TF-MIM mode, the distance modulation automatically provides
such a contrast mechanism and the demodulated MIM AC signals carry absolute information of
the sample at every point30. In this paper, we further develop the TF-MIM by using the driving
amplitude modulation (DAM) mode34, which offers satisfactory stability on samples with rough
surfaces. The demodulated MIM AC signals are analyzed by a combination of finite element
analysis (FEA) of the tip-sample admittance and Fourier transformation of the real-time signals.
Quantitative agreement between experimental data and simulated results can be achieved on bulk
dielectrics, standard calibration samples, and field-effect transistor nano-devices. Our work
suggests that the TF-MIM is an excellent tool for quantitative nanoscale imaging of electrical
properties in functional materials.
II.
Experimental Setup and Analytical Methods
Fig. 1 illustrates the TF-MIM setup with DAM feedback control34. Conventional frequency-
modulation (FM) tuning-fork AFMs using stiff cantilevers and small amplitudes prove to be a
powerful tool to achieve atomic resolution in ultrahigh vacuum environments33. However, many
samples in MIM studies may have a surface roughness of 10 to 100 nm. As the change of tip-
sample distance becomes comparable to the vibration amplitude, the atomic force can jump
frequently between attractive and repulsive regimes, leading to non-monotonic frequency shift
throughout the experiment and instability in the feedback control. On the other hand, the power
dissipation of the TF sensor depends monotonically on the tip height and can therefore be used for
robust feedback control for topographic sensing on rough surfaces35.
In our implementation, a sinusoidal voltage Vdrive at a frequency of fTF is generated by the oscillator
to drive the TF. The TF signal is detected by a current-to-voltage (I/V) amplifier and demodulated
by the HF2LI (Zurich Instruments) lock-in amplifier, whose output consists of the phase shift ϕ
and the mechanical oscillation amplitude A. A built-in phase locked loop (PLL) module is
employed to maintain the phase shift to a set point of ϕsp by changing the driving frequency, which
keeps the TF on resonance. The oscillation amplitude is kept at a constant Asp by using a built-in
proportional-integral-derivative (PID) controller. In the DAM mode, this PID output, which
represents the energy dissipation due to atomic-force interaction, is fed into the z-controller of a
3
commercial AFM (Park XE-70) as the feedback signal to maintain a constant average tip-sample
distance. More details of the DAM mode and comparisons with the FM mode can be found in Ref.
34. The distance modulation also leads to the periodic change of MIM signals at the TF frequency.
As shown in Fig. 1, the MIM-Im/Re signals (proportional to the imaginary and real parts of the
tip-sample admittance, respectively) are demodulated at fTF to form the MIM-Im/Re AC outputs
of the system.
Following the same recipe in Ref. 36 and 37, we electrochemically etch a W or Pt/Ir wire with a
diameter of 25 µm, as seen in Fig. 2a. The wire is then glued to one prong of a quartz tuning fork
and soldered to the exposed center conductor of a coaxial cable connected to the MIM circuit. The
sharpness of the tip can be accurately controlled by the etching condition. In this paper, we choose
a relatively blunt tip to enhance the signal strength, whereas sharper tips are preferred for high-
resolution imaging. The tip-sample admittance is computed by commercial FEA software
COMSOL 4.4. As shown in Fig. 2b, the tip diameter d = 300 nm and the half-cone angle = 6
are measured from the scanning electron microscopy (SEM) image in the inset of Fig. 2a. As
discussed below, the only fitting parameter of the tip geometry is the radius of curvature r (in this
case r ~ 600 nm) at the apex, which can be determined by measurements on bulk dielectrics. With
a good TF feedback control, we have confirmed that the tip condition can be preserved over an
extended period of experiments.
At the MIM working frequency near 1 GHz, the metal wire behaves electrically as a lumped
element with effective resistance Rtip = 1 , capacitance Ctip = 0.16 pF, and inductance Ltip = 10
nH connected in series. The impedance (Z) match section13 (inset of Fig. 2c) consists of a flexible
quarter-wave cable (Astro-Boa-Flex III, Astrolab Inc.) of 4.8 cm and a semi-rigid tuning stub (UT-
085C, Micro-Coax Inc.) of 6.0 cm. As shown in Fig. 2c, the measured reflection coefficient S11
can be precisely reproduced by transmission-line analysis13. With a small load of tip-sample
admittance Yt-s, the same modeling can also yield the change of S11 as a function of frequency.
Taking into account of the input microwave power (~ -20 dBm) and the electronic gain (~ 90 dB),
we can plot the conversion factor between the admittance input (in unit of nS) and the MIM output
(in unit of mV) in Fig. 2c, which peaks at the working frequency of 958 MHz. Using this parameter,
Yt-s simulated by the FEA can be directly converted to the MIM signals in a quantitative manner.
4
Fig. 2d shows the simulated Yt-s as a function of the tip-sample distance based on the tip geometry
in Fig. 2b. As the tip oscillates between the contact point and a maximum height of Ap-p, the tip-
sample admittance oscillates accordingly within a range of Y in the approach curve. In this work,
Ap-p is kept at a relatively large value of 10 nm for a better tracking of sample surfaces. In Fig. 2e,
the time dependence of Yt-s is calculated by correlating the tip-sample distance with the simulated
admittance at each moment. In the experiment, the demodulated signals at the fundamental TF
frequency, i.e., the first harmonic peak in the corresponding Fourier spectrum (Fig. 2f), are used
for the MIM AC output. Since the approach curve strongly depends on the local permittivity and
conductivity of the sample, the TF-MIM can quantitatively determine these electrical properties
after a proper calibration.
III. Results and Discussions
Similar to other quantitative microwave microscopy work38,39, our first set of experiment is to
calibrate the instrument with various bulk materials. For each sample, an area of 2 µm2 is scanned
by TF-MIM and the corresponding MIM-Im AC signal is averaged over the area to improve the
signal-to-noise ratio. By comparing the measured signals with the FEA results of fused silica
(relative permittivity r = 3.8) and LaAlO3 (r ~ 25), we can estimate a radius of curvature r ~ 600
nm at the apex. With this fitting parameter determined, the MIM-Im AC signals can be simulated
as a function of permittivity of isotropic dielectrics in Fig. 3. The experimental data on 8 bulk
materials are also plotted in the graph, showing quantitative agreement with the modeling results.
For each anisotropic dielectric, we calculate an effective isotropic permittivity by FEA simulation
such that the MIM AC response matches that using the actual anisotropic permittivity tensor. Note
that for anisotropic samples, the measured value is closer to the permittivity perpendicular to the
sample surface, e.g., along the c-axis for (001) surface. Such a phenomenon can be explained by
the monopole-like tip geometry, which generates quasi-static electric fields mostly in the vertical
direction.
Fig. 4a shows the topographic and MIM-Im AC images of the Al/SiO2/Si sample measured by the
same tip as above. The microwave image is clearly dominated by the electrical response since the
5
insulating surface contamination particle (~ 100 nm in height) and conductive Al dots (~ 20 nm in
height) exhibit opposite contrast with respect to the substrate. In traditional contact-mode MIM,
only the contrast between the Al dots (covered by 4 ~ 5 nm native oxide) and the 100-nm-SiO2/Si
background is meaningful. In TF-based MIM, however, the absolute signals at both regions (~ 40
mV on Al and ~ 8 mV on the substrate) represent local electrical properties and can be readily
simulated by the FEA, as shown in Fig. 4b. An additional advantage of distance modulation at the
kHz range is that the signal does not suffer from the electronic thermal drift in the time scale of
minutes30. As a result, no background removal is needed to post-process the MIM-Im AC raw data
in Fig. 4a. The spatial resolution of ~ 300 nm, as inferred from the line profile across one Al dot
(Fig. 4c), is consistent with the tip diameter. Owing to the robustness of the DAM operation, high-
quality AFM and MIM imaging can be acquired at a fast scan rate (up to 10 µm/s) without obvious
scan instability and tip wear.
Finally, we demonstrate that the TF-based MIM is capable of performing quantitative conductivity
imaging on nano-devices, which is one of the key areas of application in microwave microscopy.
Fig. 5a shows the optical and AFM images of an exfoliated MoS2 FET device on SiO2/Si substrate.
Details of the device structure and analysis of the contact-mode MIM results can be found in Ref.
25. Here the carrier density in the MoS2 flake can be globally tuned by the back-gate voltage VBG.
In Fig. 5b, selected MIM-Im/Re AC images at various VBG's are displayed and substantial local
inhomogeneity is observed in the channel region. To compare the macroscopic transport and
microscopic imaging results, we plot the transfer characteristics (source-drain conductance GDS
versus VBG) in Fig. 5c and MIM-Im/Re AC signals as function of GDS over an area of 1.2 m 1.2
m in Fig. 5d. Using the same FEA process above, we can also simulate the MIM-Im/Re AC
signals as a function of the sheet conductance gsh = h, where is the conductivity and h is the
thickness of MoS2. The response curves in Fig. 5e are similar to that of the contact-mode MIM9,
except that the signals are now absolute values rather than relative contrast over the insulating
background. In particular, the MIM-Im AC signals increase monotonically as gsh increases and
saturates at both the insulating (gsh < 10-9 S·sq) and conductive (gsh > 10-5 S·sq) limits. The MIM-
Re AC signals, on the other hand, peak at an intermediate gsh ~ 10-7 S·sq. Comparing the data in
Fig. 5d and simulation in Fig. 5e, the local sheet conductance within the dashed square in the MIM
images can be quantitatively extracted. The results in Fig. 5f nicely track the transport behavior in
6
Fig. 5c, with deviations due to the strong inhomogeneity in the sample and contact resistance in
the device. The error bar indicates the uncertainty of conductivity measurement which, as shown
in the inset of Fig. 5f, can be larger close to the insulating and conductive limits as a result of the
saturated MIM AC signal. We emphasize that the spatial variation of gsh carries rich information
on the material properties and device performance. The ability to quantitatively map out the
conductance distribution is therefore highly desirable for fundamental and applied research.
IV. Conclusions
In summary, we have demonstrated quantitative measurements using tuning fork-based
microwave impedance microscopy operated in the driving amplitude modulation mode. The
demodulated MIM AC signal can be simulated by a combination of FEA and Fourier
transformation. Excellent agreement is achieved between the modeling and the experiment data
on both bulk dielectrics and working nano-devices. Our work provides the pathway to perform
quantitative near-field microwave imaging, where absolute signal levels can be readily interpreted
as the local permittivity and conductivity.
Acknowledgements
This research is funded by the U.S. Department of Energy (DOE), Office of Science, Basic Energy
Sciences, under the Award No. DE-SC0010308. The authors thank E.Y. Ma and Y.T. Cui for
helpful discussions.
Reference:
1. B. T. Rosner and D. W. van der Weide, Rev. Sci. Instrum. 73, 2505 (2002).
2. S. M. Anlage, V. V. Talanov, and A. R. Schwartz, in Scanning Probe Microscopy: Electrical
and Electromechanical Phenomena at the Nanoscale, edited by S. V. Kalinin and A.
Gruverman (Springer, New York, 2006), pp. 207–245.
3. E. A. Ash and G. Nichols, Nature 237, 510 (1972).
4. M. Golosovsky and D. Davidov, Appl. Phys. Lett. 68, 1579 (1996).
5. C. P. Vlahacos, R. C. Black, S. M. Anlage, and F. C. Wellstood, Appl. Phys. Lett. 69, 3272
(1996).
7
6. T. Wei, X. D. Xiang, W. G. Wallace-Freedman, and P. G. Schultz, Appl. Phys. Lett. 68, 3506
(1996).
7. C. Gao, T. Wei, F. Duewer, Y. L. Lu, and X. D. Xiang, Appl. Phys. Lett. 71, 1872 (1997).
8. M. Tabib-Azar, D.-P. Su, A. Pohar, S. R. LeClair, and G. Ponchak, Rev. Sci. Instrum. 70, 1725
(1999).
9. K. Lai, W. Kundhikanjana, M. Kelly, and Z. X. Shen, Appl. Nanosci. 1, 13 (2011).
10. Y. Yang, K. Lai, Q. Tang, W. Kundhikanjana, M. A. Kelly, K. Zhang, Z.-X. Shen, and X. Li,
J. Micromech. Microeng. 22, 115040 (2012).
11. H. P. Huber, M. Moertelmaier, T. M. Wallis, C. J. Chiang, M. Hochleitner, A. Imtiaz, Y. J.
Oh, K. Schilcher, M. Dieudonne, J. Smoliner, P. Hinterdorfer, S. J. Rosner, H. Tanbakuchi, P.
Kabos, and F. Kienberger, Rev. Sci. Instrum. 81, 113701 (2010).
12. S. Wu and J.-J. Yu, Appl. Phys. Lett. 97, 202902 (2010).
13. K. Lai, W. Kundhikanjana, M. Kelly, and Z. X. Shen, Rev. Sci. Instrum. 79, 063703 (2008).
14. K. Lai, M. Nakamura, W. Kundhikanjana, M. Kawasaki, Y. Tokura, M. A. Kelly, and Z.-X.
Shen, Science 329, 190 (2010).
15. K. Lai, W. Kundhikanjana, M. A. Kelly, Z.-X. Shen, J. Shabani, and M. Shayegan, Phys. Rev.
Lett. 107, 176809 (2011).
16. E. Y. Ma, Y.-T. Cui, K. Ueda, S. Tang, K. Chen, N. Tamura, P. M. Wu, J. Fujioka, Y. Tokura,
and Z.-X. Shen, Science 350, 538 (2015).
17. X. Wu, U. Petralanda, L. Zheng, Y. Ren, R. Hu, S.-W. Cheong, S. Artyukhin, and K. Lai, Sci.
Adv. 3, e1602371 (2017).
18. K. Lai, H. Peng, W. Kundhikanjana, D. T. Schoen, C. Xie, S. Meister, Y. Cui, M. A. Kelly,
and Z.-X. Shen, Nano Lett. 9, 1265 (2009).
19. J.-S. Kim, Y. Liu, W. Zhu, S. Kim, D. Wu, L. Tao, A. Dodabalapur, K. Lai, and D. Akinwande,
Sci. Rep. 5, 8989 (2015).
20. E. Seabron, S. MacLaren, X. Xie, S. V. Rotkin, J. A. Rogers, and W. L. Wilson, ACS Nano 10,
360 (2016).
21. M. Tuteja, P. Koirala, V. Palekis, S. MacLaren, C. S. Ferekides, R. W. Collins, and A. A.
Rockett, J. Phys. Chem. C 120, 7020 (2016).
22. S. Berweger, G. A. MacDonald, M. Yang, K. J. Coakley, J. J. Berry, K. Zhu, F. W. DelRio, T.
M. Wallis, and P. Kabos, Nano Lett. 17, 1796 (2017).
8
23. Z. Chu, M. Yang, P. Schulz, D. Wu, X. Ma, E. Seifert, L. Sun, X. Li, K. Zhu, and K. Lai,
Nature Commun. 8, 2230 (2017).
24. Y. Ren, H. Yuan, X. Wu, Z. Chen, Y. Iwasa, Y. Cui, H.Y. Hwang, and K. Lai, Nano Lett. 15,
4730 (2015).
25. D. Wu, X. Li, L. Luan, X. Wu, W. Li, M. N. Yogeesh, R. Ghosh, Z. Chu, D. Akinwande, Q.
Niu, and K. Lai, Proc. Natl. Acad. Sci. 113, 8583 (2016).
26. M. C. Biagi, R. Fabregas, G. Gramse, M. Van Der Hofstadt, A. Juarez, F. Kienberger, L.
Fumagalli, and G. Gomila, ACS Nano 10, 280 (2016).
27. A. Tselev, J. Velmurugan, A. V. Ievlev, S. V. Kalinin, and A. Kolmakov, ACS Nano 10, 3562
(2016).
28. K. Lai, W. Kundhikanjana, H. Peng, Y. Cui, M. A. Kelly, and Z.-X. Shen, Rev. Sci. Instrum.
80, 043707 (2009).
29. E. Y. Ma, M. R. Calvo, J. Wang, B. Lian, M. Muhlbauer, C. Brune, Y.-T. Cui, K. Lai, W.
Kundhikanjana, Y. Yang, M. Baenninger, M. Konig, C. Ames, H. Buhmann, P. Leubner, L.
W. Molenkamp, S.-C. Zhang, D. Goldhaber-Gordon, M. A. Kelly, Z.-X. Shen, Nature
Commun. 6, 7252 (2015).
30. Y.-T. Cui, E. Y. Ma, and Z.-X. Shen, Rev. Sci. Instrum. 87, 063711 (2016).
31. Y.-T. Cui, B. Wen, E. Y. Ma, G. Diankov, Z. Han, F. Amet, T. Taniguchi, K. Watanabe, D.
Goldhaber-Gordon, C. R. Dean, and Z.-X. Shen, Phys. Rev. Lett. 117, 186601 (2016).
32. F. J. Giessibl, Appl. Phys. Lett. 73, 3956 (1998).
33. F. J. Giessibl, Rev. Mod. Phys. 75, 949 (2003).
34. M. Jaafar, D. Martínez-Martín, M. Cuenca, J. Melcher, A. Raman and J. Gómez-Herrero,
Beilstein J. Nanotechnol. 3, 336 (2012).
35. J. Rychen, T. Ihn, P. Studerus, A. Herrmann, K. Ensslin, H. J. Hug, P. J. A. van Schendel and
H. J. Güntherodt, Appl. Surf. Sci. 157 (4), 290-294 (2000).
36. E. Inger, W. Erik, C. Dan, O. Hakan and O. Eva, Meas. Sci. Tech. 10, 11 (1999).
37. I. H. Musselman, P. A. Peterson, and P. E. Russel, Precis. Eng. 12 (1), 3-6 (1990).
38. K. Lai, W. Kundhikanjana, M. Kelly, and Z. X. Shen, Appl. Phys. Lett. 93, 123105 (2008).
39. Z. Wei, E. Y. Ma, Y.-T. Cui, S. Johnston, Y. Yang, K. Agarwal, M. A. Kelly, Z.-X. Shen, and
X. Chen, Rev. Sci. Instrum. 87, 094701 (2016).
9
FIG. 1. Schematic of the TF-based AFM configured for the DAM mode and the microwave electronics
(detailed in the text).
10
FIG. 2. (a) SEM image of a typical etched W tip. The inset shows a zoom-in view near the tip apex. (b)
Quasi-static potential distribution around the tip and a bulk dielectric sample simulated by the FEA software.
(c) Measured S11 (red squares) of a TF-based sensor and a fit to the transmission line analysis (black dashed
line). The blue curve is the simulated conversion factor between the tip-sample admittance and the MIM
output. The inset shows the equivalent circuit of the impedance-match network. (d) Tip-sample admittance
Yt-s on a dielectric sample (εr = 25) as a function of tip-sample distance (sketched in the inset) modeled by
FEA. The tip oscillates between the contact point and a maximum height of Ap-p, resulting in a change of
Y in the tip-sample admittance. (e) Time dependence of Yt-s, assuming a simple harmonic oscillation of
the tip. (f) Fourier spectrum of Yt-s. The amplitude of the first harmonic peak corresponds to the MIM AC
signal demodulated by the lock-in amplifier.
11
FIG. 3. Demodulated tip-sample admittance and the corresponding MIM-Im AC signals as a function of
the relative permittivity. Materials and the permittivity values are listed next to the symbols. Bars in the
vertical axis indicate experimental uncertainties. Bars in the horizontal axis, on the other hand, identify the
range between permittivity values at a-axis (εa) and c-axis (εc) for anisotropic materials. Data points for
anisotropic materials represent the effective isotropic permittivity (see the text for details). The solid line is
the simulated curve using FEA.
12
FIG. 4. (a) AFM and MIM-Im AC images of a patterned Al dot sample measured by the TF-based MIM.
The scale bars are 2 µm. The inset shows the sample structure. An insulating surface particle (inside the red
circle) shows higher topographic and lower MIM signals than the substrate. (b) Simulated approach curves
on the Al dot (solid blue line) and the substrate (dashed red line). (c) Line profile of the MIM-Im AC signal
across an Al dot marked in (a), showing the electrical contrast between the two regions and a spatial
resolution of ~ 300 nm.
13
FIG. 5. (a) Optical and AFM images of a back-gated MoS2 field effect transistor. (b) Selected MIM-Im/Re
AC images at different back gate voltages. All scale bars are 2 µm. (c) Transfer characteristics of the device
at a source-drain voltage VDS = 0.1 V. (d) Averaged MIM AC signals inside the dashed square in (b) as a
function of source-drain conductance GDS. (e) Simulated MIM AC signals as a function of the sheet
conductance gsh. (f) Local sheet conductance versus VBG calculated by comparing (d) and (e). The inset
shows the same data with y-axis in log scale.
14
|
1908.02364 | 1 | 1908 | 2019-07-15T16:54:10 | Epitaxial Growth and Characterization of AlInN Based Core-Shell Nanowire Light Emitting Diodes Operating in the Ultraviolet Spectrum | [
"physics.app-ph",
"physics.optics"
] | We report on the demonstration of the first axial AlInN ultraviolet core-shell nanowire light-emitting diodes with highly stable emission in the UV wavelength range. During the epitaxial growth of AlInN layer, an AlInN shell is spontaneously formed, resulted in the reduced nonradiative recombination on nanowire surface. The AlInN nanowires exhibit high internal quantum efficiency of ~ 52% at room temperature for emission at 295nm. The peak emission wavelength can be varied from 290 nm to 355 nm by changing the growth condition. Moreover, significantly strong transverse magnetic (TM) polarized emission is recorded which is ~ 4 times stronger compared to the transverse electric (TE) polarized light at 295 nm. This study provides alternative approach for the fabrication of new type of high-performance ultraviolet light-emitters. | physics.app-ph | physics | Epitaxial Growth and Characterization of AlInN Based Core-
Shell Nanowire Light Emitting Diodes Operating in the
Ultraviolet Spectrum
Ravi Teja Velpula 1, Moab Rajan Philip1, Barsha Jain1, Hoang Duy Nguyen2,¥, Renjie
Wang3, and Hieu Pham Trung Nguyen1,*
1Department of Electrical and Computer Engineering, New Jersey Institute of Technology,
323 Dr Martin Luther King Jr Boulevard, Newark, New Jersey, 07102
2Institute of Chemical Technology, Vietnam Academy of Science and Technology, 1 Mac
Dinh Chi Street, District 1, Ho Chi Minh City 700000 Vietnam
3Department of Engineering Physics, McMaster University, 1280 Main Street West,
Hamilton, Ontario, L8S 4L7 Canada
E-mail: *[email protected]; ¥[email protected]
Abstract: We report on the demonstration of the first axial AlInN ultraviolet core-shell
nanowire light-emitting diodes with highly stable emission in the UV wavelength range.
During the epitaxial growth of AlInN layer, an AlInN shell is spontaneously formed, resulted
in the reduced nonradiative recombination on nanowire surface. The AlInN nanowires
exhibit high internal quantum efficiency of ~ 52% at room temperature for emission at
295nm. The peak emission wavelength can be varied from 290 nm to 355 nm by changing the
growth condition. Moreover, significantly strong transverse magnetic (TM) polarized
emission is recorded which is ~ 4 times stronger compared to the transverse electric (TE)
1
polarized light at 295 nm. This study provides alternative approach for the fabrication of new
type of high-performance ultraviolet light-emitters.
Introduction
Ultraviolet (UV) light-sources have been in great attention due to their wide range of applications.
Such UV light-emitters have primarily used in several key applications include remote detection of
biological and chemical compound1, phototherapy2, water/air/surface purification and
disinfection3,4, cancer detection5 and fluorescence sensing or Raman spectroscopy6. Currently,
major UV light-emitting diode (LED) customers are users of UV-A (315-400 nm) and UV-B (280-
315 nm) LEDs, representing over 90% of the overall UV LED market7. Among these applications,
UV curing is the most dynamic and most important market, due to significant advantages offered
over traditional technologies, comprising lower cost of ownership, system miniaturization,
etc.7 Several efforts have been made to develop high efficiency deep UV LEDs using AlGaN
material. However, the performance of AlGaN based deep UV LEDs has been fundamentally
limited by the large dislocation density and the extremely inefficient p-type doing, resulted to low
efficiency and low output power7-10. Moreover, the device light extraction efficiency is further
limited by UV light polarization particularly in the spectral range of ~290-355 nm, when the emitted
light polarization switches from transverse electric (TE) polarization to transverse magnetic (TM)
polarization due to prohibitively high Al-composition and nanowire geometry.11,12 As such,
polarization state switches from TE to TM reducing light extraction efficiency.13 Due to these
barriers, the external quantum efficiency (EQE) of UV LEDs with emission wavelength above 300
nm can reach up to nearly 10%. However, their performance deteriorates drastically with
decreasing wavelengths 14. For instance, the EQE of deep UV LEDs with emission wavelength
below 250 nm decreases dramatically from less than 1% to ~ 0.04%, and the extremely low output
power which is just a few tens of nW for emission 210 nm8. Such power is extremely low for
practical application.
2
Until recently, fundamental and applied research approaches for light-emitters, have
essentially focused on the use of InGaN and AlGaN alloys in the active region for near UV 15,16 and
UV photonic devices, respectively, while the approach of using different III-nitride UV materials
is relatively unexplored. Identifying and developing the potential of alternative UV materials will
be critical to make further progress in development of deep UV emitters. In this regard, AlxIn1-xN
alloy has not been widely studied even though it holds great potential application in UV and visible
light-emitting devices. For example, AlInN can be grown perfectly lattice-matched to GaN for an
indium content closes to 17 -- 18% 17,18, the large refractive index contrast and high bandgap with
respect to GaN make these a highly promising candidate for UV light-emitters compared to AlGaN
alloy. Recent studies shown that AlInN offers a large optical gain for deep UV LEDs19. Free of
defects and quantum-confine Stark effect were achieved for m-plan GaN/AlInN multi-quantum
wells in core-shell nanowires for UV emitters20. Using k.p perturbation theory, Fu et al. reported
that AlInN compounds can be grown on both GaN and AlN templates, while AlGaN is detrimental
to be grown on GaN templates21. AlInN offers wider windows of optimal alloy composition for UV
emission compared to AlGaN, especially to deeper UV emission21. AlInN has many advantages
and is a great of interest that can replace AlGaN or InGaN in several photonic and electronic
devices. For instance, lattice matched GaN/AlInN superlattices have been chosen for the near-
infrared based on intersubband transitions 22, high reflectivity short wavelength distributed Bragg
reflectors (DBRs) 23, high quality factor microcavities for vertical cavity surface emitting laser
structures 24, and the realization of high performance high electron mobility transistors (HEMTs)
25.
Albeit holding tremendous advantages, AlInN semiconductor research is highly limited due
to the immature epitaxial growth of high quality AlInN. Molecular beam epitaxy (MBE) growth of
group-III nitrides under metal-rich conditions usually provides smooth surface morphologies at low
growth temperatures. However, under nitrogen-rich growth at these low temperatures results in
3
rough surfaces 26. The main growth issue for AlInN by MBE is composition inhomogeneity which
is commonly presented in AlInN layer 27. It is suggested that phase separation is attributed to
random compositional fluctuations during the early stages of growth, possibly associated with
misfit-strain relaxation 27,28. The difficulties in epitaxial growth of AlInN have resulted from
extremely large differences in optimal growth temperatures for InN (~ 450oC) and AlN (~ 800oC)
29. Additionally, inefficient p-type doping in AlInN also strongly affects the electrical properties of
the related devices. Such difficulties result in the low crystalline quality and low device
performances. The MBE growth under nitrogen rich condition offers an effective approach to
eliminate the composition inhomogeneity in the AlInN, reported by Speck et al. 30,31. By decreasing
Al flux and growth of AlInN under N-rich conditions, homogenous AlInN layers with high In
content could be achieved 30,31. Therefore, the nitrogen rich grown nanowires seem to be the best
option offering homogenous AlInN structures with nearly-free of dislocation at high In content.
However, to our best knowledge, axial nanowire-based AlInN semiconductor grown by MBE has
not been reported even though nanowire structures offers several advantages. Nanowire structures
offer significant attributes, for instance, significantly improved light output power due to drastically
reduced dislocations and polarization fields 32,33. As intensively reported recently, III-nitride
nanowire based LEDs with exceptional performance have been successfully achieved on Si
substrates 32,34,35. More recently, it has been shown that the formation energy for substitutional
doping in the near surface region of nanowires can be significantly reduced, together with the
nearly-free of dislocation can lead to the enhanced surface doping and conductivity in nanowire
LEDs for high-efficiency emission 36.
In this context, we have performed a detailed investigation of the epitaxial growth, structural
and optical properties of catalyst-free AlxIn1-xN/GaN nanowires grown on Si (111) substrate by
plasma-assisted molecular beam epitaxy (PAMBE). We have also further demonstrated the first
axial AlInN core-shell nanowire UV LED heterostructures operating in the UV-A and UV-B bands.
4
An AlInN shell is spontaneously formed during the growth of AlInN epi-layer which can lead to
drastically reduced nonradiative surface recombination. By controlling the Al compositions in the
AlInN active region, the emission wavelength can be varied from 290 nm to 355 nm. The AlInN
UV nanowires with emission wavelength at 295nm exhibit high internal quantum efficiency (IQE)
of ~52% at room temperature. Moreover, the UV LED device exhibits strong UV light emission
with highly stable peak emission at 295nm. The polarized optical properties of AlInN nanowire
LEDs were also investigated. It is suggested that the UV light from AlInN nanowire LEDs is mainly
TM polarized with emission about 4 times stronger than that of TE light.
Results
Structural characterizations
In this study, AlInN nanowire light-emitters were grown on n-Si (111) substrates by a Veeco Gen
II MBE system equipped with a radio-frequency plasma-assisted nitrogen source. Illustrated in
Figure 1(a), GaN nanowire template was first grown on Si substrate to facilitate the formation of
AlInN segment. Subsequently, detailed study of epitaxial growth of AlInN nanowires on GaN
nanowire templates was performed to acquire the optimal growth condition for AlInN nanowire
LEDs. Structural properties of AlInN nanowires were characterized by scanning transmission
electron microscopy (STEM). Figure 1(b) confirms the presence of GaN and AlInN segments. The
wire diameter increase from GaN segment to AlInN portion and remains constant at the top of
nanowire which is similar to our reported studies on InGaN, and AlGaN nanowires35,37,38.
Moreover, it is also suggested that a core-shell AlInN/GaN structure is spontaneously formed
during the epitaxial growth of AlInN layer. In order to reveal the compositional variation along the
nanowire radial direction, energy dispersive X-ray spectrometry (EDXS) analysis was performed
along the GaN and AlInN regions which are indicated as lines 1-2 and 3-4 in Figure 1(b),
respectively. Shown in Figure 1(c), corresponding to line-scan 1-2, the Ga signal exhibits a
5
maximum in the nanowire center and drops near the sidewalls. In contrast, the Al signal shows
clear peaks near the sidewalls and a dip in the core region of the nanowire. The presence of In is
also recorded with detected In signal even though In signal is significantly lower compared to Ga
and Al signals. Therefore, it is suggested that a unique GaN/AlInN radial core-shell heterostructure
were grown. At the top portion of nanowire corresponding to line-scan 3-4, an AlInN shell around
the AlInN core is also confirmed by the EDXS line scan, which is illustrated in Figure 1(d). The In
signal is well confined in the core region of the nanowire. The Al signal is again maximum at the
sidewalls and significantly reduced at the core section of the wire. The thickness of the shell is
about 13.6 nm at the nanowire top and reduces gradually to about 8.4 nm at the nanowire bottom.
The formation of such core-shell nanowire structures can be well explained by the diffusion-
controlled growth mechanism of III-nitride nanowires under nitrogen-rich conditions which was
carefully studied in previous studies39,40. Moreover, the presence of the shell layer significantly
improves the optical properties of the underlying GaN nanowire templates as well as the AlInN
core.
We have further demonstrated the AlInN/GaN UV nanowire LEDs on Si substrate utilizing
the optimal growth conditions of AlInN nanowire on GaN templates. The device structure is
schematically illustrated in Figure 2(a) which includes a ~ 200 nm GaN:Si segment, 100 nm AlxIn1-
xN:Si/40 nm i-AlyIn1-yN/100 nm AlxIn1-xN:Mg quantum well, and ~10 nm GaN:Mg. The Al and In
compositions in the active region can be varied by adjusting the Al/In flux ratios and/or the growth
temperatures to control the emission wavelengths of these AlInN UV nanowire LEDs. Illustrated
in Figure 2(b), the nanowires are vertically aligned to the substrate and exhibit nearly uniform
heights, with diameter at the top nanowire in the range of ~ 90 nm. Such nanowire properties are
suitable for device fabrication.
Optical Characterizations
6
Photoluminescence (PL) spectra of AlInN nanowire on GaN templates were measured using a 266
nm diode-pumped solid-state laser as the excitation source. The PL emission was spectrally
resolved by a high-resolution spectrometer and detected by a photomultiplier tube. Figure 3(a)
shows photoluminescence spectra of different AlInN/GaN nanowire structures which were grown
at different growth conditions. It is clearly shown that the peak emissions vary from 290 nm to 355
nm by varying the Al composition in the AlInN layers. In this study, the Al/In BEP ratio was kept
constant while the substrate temperature was increased from 670 oC to 720 oC. The nitrogen flow
rate was kept at 2.5 sccm. The peak emission is shifted to shorter wavelength when the substrate
temperature is increased which is attributed to the increased In adatom desorption at higher growth
temperature, resulted in the reduced In composition in the AlInN segment. Shown in Figure 3(a),
the peak emission at ~ 365nm is related to the emission from GaN nanowire templates. We also
estimated the Al composition in AlInN layer using the room-temperature PL peak energy EPeak via
the following equation:
EPeak(x) ≈ Eg(x) = xEg(AlN) + (1-x)Eg(InN) − bx(1 − x), where x is the Al composition, Eg is the
bandgap energy. In our calculation, Eg(AlN) and Eg(InN) are considered as 6.2 eV 41and 0.7 eV 41
respectively and b is taken as the bowing parameter. b is chosen to be 3.4 eV 42. Shown in Figure
3(b), the Al content is estimated to be in the range of 64 %-76 %, corresponding to emission
wavelength from 290 nm -- 355 nm. At growth temperature of 710 oC, AlInN nanowires with
emission wavelength at 295 nm were recorded with strong emission intensity and spectral linewidth
of ~ 28 nm. The optical properties of those AlInN/GaN nanowires were further characterized at
different temperatures varying from 20 K to 300 K using liquid Helium to estimate their IQE.
Presented in Figure 3(c), the AlInN/GaN nanowire exhibits relatively high IQE which is estimated
of ~ 52% at room temperature attributed to the strong carrier confinement provided by the AlInN
shell and nearly intrinsic AlInN core. The IQE is calculated by comparing the PL intensity at room
temperature and 20 K, assuming the IQE at 20 K is near-unity43.
7
Device Performance
Such vertically aligned AlInN nanowire LEDs are fully compatible with the conventional
fabrication process for large area nanowire LED devices. The device fabrication is described in the
method section. The UV nanowire LED devices with an areal size of 500×500 µm2 were chosen
for characterization. The AlInN LEDs have excellent current-voltage characteristic with low
resistance measured at room temperature, shown in Figure 4(a). The leakage current was found to
be very small which is about 1 µA at -8 V. Turn on voltage of these UV nanowire LEDs is ~ 5 V
which is significantly lower compared to current thin-film AlGaN LEDs at similar wavelength
range44,45 and is also better/comparable to that of currently reported AlGaN UV nanowire
LEDs43,46,47. Figure 4(b) presents electroluminescence (EL) spectra of the AlInN nanowire LEDs
under various injection currents from 5 mA to 100 mA. No obvious shift in the peak wavelength
was observed attributed to the negligible quantum-confined Stark effect (QCSE) in the LED
structures, further confirming the high crystalline quality of such AlInN nanowire heterostructures.
The light emission polarization properties of the AlInN UV nanowire LEDs were also characterized
at room temperature. Transverse-magnetic (TM) and transverse-electric (TE) are defined as the
electric field parallel (E//c) and perpendicular (EꞱc) to c-axis, respectively. The measurement was
performed at injection current of 10 A/cm2. Illustrated in Figure 5(a), the UV light emission is
predominantly TM polarized which is about > 4 times stronger than that of TE polarized emission.
This observation agrees well with the simulation results in which the TM polarized emission is
more than two orders of magnitude strong than TE polarized light, shown in Figure 5(b). Similar
trend of polarization for LEDs using AlGaN at the same UV wavelength regime are also reported
by others46,48. This result plays important role in the design of surface emitting UV LEDs using
AlInN compounds to achieve high light extraction efficiency.
Discussions
8
In addition to study the performance of AlInN UV nanowire LEDs, we have performed detailed
simulation comparing the characteristics of AlInN nanowire LEDs with and without the integration
of electron blocking layer. It is clearly shown that, in both LED device structures, electron leakage
does not exist or is negligible with quite similar electron current density distribution, shown in
Figure 6(a). However, the EBL has strong impact on hole injection efficiency. The EBL-free LED
has better hole injection efficiency compared to the other. It is observed in the band diagram of the
LED with EBL that there is band bending in the valence band at the heterointerface of the EBL and
quantum well (See Figure S1(a) in the Supporting Information), resulted in the hole accumulation
at the starting portion of the EBL (See Figure S1(b) in the Supporting Information). This
phenomenon decreases the hole injection efficiency in the quantum well, as well as leads to higher
turn-on voltage of AlInN UV nanowire LEDs with EBL, presented in Figure 6(b). Advantages of
AlInN nanowire structures include the intergration of such nanowire UV LED structures on GaN
templates as well as simple structure without the employment of an EBL for high device
performance. The EBL-free LED structure is particularly important for developing deep UV LEDs
since the Al composition almost reaches maximum for deep UV emission (below 240nm).
Therefore, the optimal EBL structure is limitted which requires higher bandgap energy to
effectively prevent electron overflow. Moreover, the use of EBL will also affect the hole transport,
resulted to the reduced hole injection efficiency to the device active region. Further optimization in
term of device structure, active region thickness and composition will be performed to achieve high
power AlInN deep UV nanowire LEDs.
In summary, we have successfully demonstrated the first AlInN axial nanowire LEDs
operating the UV-A and UV-B bands with relatively high IQE of ~ 52% at room temperature. The
electron overflow was not observed within these nanowire UV emitters. The devices exhibit stable
emission with strong TM polarized light emission. The device performance can be further improved
9
by engineering the device structure, nanowire morphology as well as nanowire diameter and
spacing to enhance the light extraction efficiency of such AlInN UV core-shell nanowire LEDs.
Materials and methods
Molecular beam epitaxial (MBE) growth
Vertically aligned self-organized AlInN/GaN heterostructures and AlInN core-shell nanowire
LEDs were grown on Si(111) substrates by radio frequency plasma-assisted molecular beam
epitaxy. The extremely high purity nitrogen generation system was employed to introduce ultrahigh
quality nitrogen gas to the RIBER RF-nitrogen plasma cell. This system includes a Delux Nitrogen
purifying system with bypass assembly life status indicator, valve control for bypass and purifier
and heating control. The oxide on the substrate surface is desorbed in-situ at 780 oC. First, GaN
nanowire templates are formed under nitrogen-rich conditions without the use of any external
catalyst. The growth conditions of GaN nanowires include a growth temperature of 770 oC, with a
nitrogen flow rate of 1.0 sccm, a forward plasma power of 400 W, and Ga beam equivalent pressure
of 6 × 10-8 Torr. To achieve UV light emission, self-organized AlInN segments are subsequently
grown on top of GaN nanowires. The In composition in the active region can be controlled by
varying the In and Al beam flux and/or the substrate temperature. The growth temperature of AlInN
active regions is varied to enhance the In incorporation which is controlled in between 670 oC to
720 oC. During the epitaxial growth of AlInN segments, the nitrogen flow rate and plasma power
are kept of 2.5 sccm and 400 W, respectively.
Fabrication process
The UV nanowire LED fabrication process includes the following steps. The nanowire LED
samples were first cleaned by HCl then HF to remove native oxide on nanowire surface and oxide
layer on the backside of Si substrates, respectively. Ti(20 nm)/Au(120 nm) metal layers were then
10
deposited on the backside of Si wafers for n-contact. The p-metal contact of Ni(10 nm)/Au(10 nm)
was deposited on the top of nanowire samples by e-beam evaporation. For enhancing efficient
current spreading of this p-contact layer, the top portions of nanowires have to be linked together
which can conducted by tilting the substrate holder with a certain angle during the deposition. Thick
Ni(20 nm)/Au(120 nm) layers were subsequently deposited on top of the device to serve as metal
pad. The fabricated devices with Ti/Au and Ni/Au contacts were annealed at ~ 550 oC for 1 min.
Filling material and Indium Tin Oxide (ITO) were not used in this fabrication to eliminate any
light-absorption in this UV wavelength range which is different from our visible InGaN/(Al)GaN
nanowire LED fabrication. LEDs with chip areas of ~ 500×500 µm2 were fabricated and selected
for characterization.
Transmission electron microscopy (TEM)
JEOL JEM-2100F equipped with near field emission gun with an accelerating voltage of 200 kV
was used to obtain bright-field TEM images. For STEM-and STEM-HAADF imaging same
equipment with a cold field emission emitter operated at 200 kV and with an electron beam
diameter of approximately 0.1 nm was used.
Photoluminescence measurement
A 266 laser (Kimmon Koha) was used as the excitation source for the photoluminescence
measurement of the nanowire heterostructure. The photoluminescence was spectrally resolved by
a high-resolution spectrometer and detected by a photomultiplier tube (PMT).
Electrical, electroluminescence and light polarization characterization
The current-voltage characteristics of AlInN UV core-shell nanowire LEDs were measured
Keithley 2400 digital source meter. Electroluminescence emission of the LED devices were
11
collected by an optical fiber and analyzed using an Ocean Optics spectrometer. The light emission
polarization set up consists of an optical fiber together with a Glan-Taylor polarizer mounted on a
rotating arm. Signal from the lateral surface of AlInN nanowire LEDs was polarization resolved by
polarizer and collected and analyzed by Ocean optics spectrometer. The measurement was
performed at 20mA CW injection current.
Acknowledgment
This work is being supported by NJIT, Instrument Usage Seed Grant from Otto H. York Center at
NJIT, and Vietnam National Foundation for Science and Technology Development (NAFOSTED)
under grant number 103.03-2017.312. Part of the work was performed in the Cornel
NanoFabrication Center.
Conflict of interest
The authors declare that they have no conflict of interest.
Author contributions
H.P.T.N. designed the experiment. H.P.T.N., R.W., and M. R. P. contributed to the MBE growth
and device fabrication. H.D.N., R.T.V., and M.R.P. performed the EL, PL and power
measurements. H.D.N. performed TEM studies. R.T.V. and B. J. contributed to the device
simulation. H.P.T.N. and R.T.V wrote the manuscript with contributions from other co-authors.
References
Lytvyn, P. M. et al. Polarization effects in graded AlGaN nanolayers revealed by current-
sensing and Kelvin probe microscopy. ACS Applied Materials & Interfaces 10, 6755-
6763 (2018).
Smith, P. K. C. Laser (and LED) therapy is phototherapy. Photomedicine and Laser
Surgery 23, 78-80 (2005).
Banas, M. A. et al. Final LDRD report: ultraviolet water purification systems for rural
environments and mobile applications. (Sandia National Laboratories, 2005).
12
1
2
3
Würtele, M. et al. Application of GaN-based ultraviolet-C light emitting diodes -- UV
LEDs -- for water disinfection. Water research 45, 1481-1489 (2011).
Alimova, A. N. et al. Hybrid phosphorescence and fluorescence native spectroscopy for
breast cancer detection. Journal of biomedical optics 12, 014004 (2007).
Peng, H. et al. Ultraviolet light-emitting diodes operating in the 340 nm wavelength range
and application to time-resolved fluorescence spectroscopy. Applied Physics Letters 85,
1436-1438 (2004).
Muramoto, Y., Kimura, M. & Nouda, S. Development and future of ultraviolet light-
emitting diodes: UV-LED will replace the UV lamp. Semiconductor Science and
Technology 29, 084004 (2014).
Kneissl, M. et al. Advances in group III-nitride-based deep UV light-emitting diode
technology. Semiconductor Science and Technology 26, 014036 (2011).
Kneissl, M., Seong, T.-Y., Han, J. & Amano, H. The emergence and prospects of deep-
ultraviolet light-emitting diode technologies. Nature Photonics 13, 233 (2019).
Khan, A., Balakrishnan, K. & Katona, T. Ultraviolet light-emitting diodes based on group
three nitrides. Nature photonics 2, 77 (2008).
Kent, T. F. et al. Deep ultraviolet emitting polarization induced nanowire light emitting
diodes with AlxGa(1-x)N active regions. Nanotechnology 25, 455201 (2014).
Zhao, S., Djavid, M. & Mi, Z. Surface Emitting, High efficiency near-vacuum ultraviolet
light source with aluminum nitride nanowires monolithically grown on silicon. Nano
Letters 15, 7006-7009 (2015).
Sarwar, A. T., May, B. J., Chisholm, M. F., Duscher, G. J. & Myers, R. C. Ultrathin GaN
quantum disk nanowire LEDs with sub-250 nm electroluminescence. Nanoscale 8, 8024-
8032 (2016).
Monroy, E., Omnès, F. & Calle, F. Wide-bandgap semiconductor ultraviolet
photodetectors. Semiconductor Science and Technology 18, R33 (2003).
Chiu, C. et al. Improved output power of InGaN-based ultraviolet LEDs using a heavily
Si-doped GaN insertion layer technique. IEEE Journal of Quantum Electronics 48, 175-
181 (2012).
Lin, L. et al. InGaN/GaN ultraviolet LED with a graphene/AZO transparent current
spreading layer. Opt. Mater. Express 8, 1818-1826 (2018).
Carlin, J.-F. & Ilegems, M. High-quality AlInN for high index contrast Bragg mirrors
lattice matched to GaN. Applied Physics Letters 83, 668-670 (2003).
Lorenz, K. et al. Anomalous ion channeling in AlInN GaN bilayers: Determination of the
strain state. Physical Review Letters 97, 085501 (2006).
13
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Tan, C.-K., Sun, W., Borovac, D. & Tansu, N. Large optical gain AlInN-Delta-GaN
quantum well for deep ultraviolet emitters. Scientific Reports 6, 22983 (2016).
Durand, C. et al. M-plane GaN/InAlN multiple quantum wells in core -- shell wire
structure for UV emission. ACS Photonics 1, 38-46 (2014).
Fu, D. et al. Exploring optimal UV emission windows for AlGaN and AlInN alloys
grown on different templates. physica status solidi (b) 248, 2816-2820 (2011).
Nicolay, S. et al. Midinfrared intersubband absorption in lattice-matched AlInN∕GaN
multiple quantum wells. Applied Physics Letters 87, 111106 (2005).
Dorsaz, J., Carlin, J.-F., Gradecak, S. & Ilegems, M. Progress in AlInN -- GaN Bragg
reflectors: Application to a microcavity light emitting diode. Journal of Applied Physics
97, 084505 (2005).
Feltin, E. et al. Blue lasing at room temperature in an optically pumped lattice-matched
AlInN=GaN VCSEL structure. Electronics Letters 43, 924-926 (2007).
Gonschorek, M., Carlin, J.-F., Feltin, E., Py, M. A. & Grandjean, N. High electron
mobility lattice-matched AlInN∕GaN field-effect transistor heterostructures. Applied
Physics Letters 89, 062106 (2006).
Zywietz, T., Neugebauer, J. & Scheffler, M. Adatom diffusion at GaN (0001) and (0001)
surfaces. Applied Physics Letters 73, 487-489 (1998).
Choi, S., Wu, F., Shivaraman, R., Young, E. C. & Speck, J. S. Observation of columnar
microstructure in lattice-matched InAlN/GaN grown by plasma assisted molecular beam
epitaxy. Applied Physics Letters 100, 232102 (2012).
Zhou, L., Smith, D. J., McCartney, M. R., Katzer, D. S. & Storm, D. F. Observation of
vertical honeycomb structure in InAlN∕GaN heterostructures due to lateral phase
separation. Applied Physics Letters 90, 081917 (2007).
Butté, R. et al. Current status of AlInN layers lattice-matched to GaN for photonics and
electronics. Journal of Physics D: Applied Physics 40, 6328 (2007).
Stephen, W. K. et al. GaN-based high-electron-mobility transistor structures with
homogeneous lattice-matched InAlN barriers grown by plasma-assisted molecular beam
epitaxy. Semiconductor Science and Technology 29, 045011 (2014).
Kyle, E. C. H., Kaun, S. W., Wu, F., Bonef, B. & Speck, J. S. High indium content
homogenous InAlN layers grown by plasma-assisted molecular beam epitaxy. Journal of
Crystal Growth 454, 164-172 (2016).
Guo, W., Zhang, M., Banerjee, A. & Bhattacharya, P. Catalyst-free InGaN/GaN
nanowire light emitting diodes grown on (001) silicon by molecular beam epitaxy. Nano
Letters 10, 3355-3359 (2010).
Zhao, S., Nguyen, H. P. T., Kibria, M. G. & Mi, Z. III-Nitride nanowire optoelectronics.
Progress in Quantum Electronics 44, 14-68 (2015).
14
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Hahn, C. et al. Epitaxial growth of InGaN nanowire arrays for light emitting diodes. Acs
Nano 5, 3970-3976 (2011).
Nguyen, H. P. T. et al. Controlling electron overflow in phosphor-free InGaN/GaN
nanowire white light-emitting diodes. Nano Letters 12, 1317 -- 1323 (2012).
Zhao, S. et al. Tuning the surface charge properties of epitaxial InN nanowires. Nano
Letters 12, 2877-2882 (2012).
Nguyen, H. P. T. et al. p-Type modulation doped InGaN/GaN dot-in-a-wire white-light-
emitting diodes monolithically grown on Si(111). Nano Letters 11, 1919-1924 (2011).
Wang, Q., Nguyen, H. P. T., Cui, K. & Mi, Z. High efficiency ultraviolet emission from
AlxGa(1-x)N core-shell nanowire heterostructures grown on Si (111) by molecular beam
epitaxy. Applied Physics Letters 101, 043115-043114 (2012).
Nguyen, H. P. T. et al. Breaking the carrier injection bottleneck of phosphor-free
nanowire white light-emitting diodes. Nano Letters 13, 5437-5442 (2013).
Nguyen, H. P. T. et al. Engineering the carrier dynamics of InGaN nanowire white light-
emitting diodes by distributed p-AlGaN electron blocking layers. Scientific Reports 5,
7744 (2015).
Wu, J. Q. When group-III nitrides go infrared: New properties and perspectives. Journal
of Applied Physics 106, (2009).
Piprek, J. Nitride semiconductor devices: principles and simulation. (Berlin: Wiley-vch ,
2017).
Zhao, S. et al. Aluminum nitride nanowire light emitting diodes: Breaking the
fundamental bottleneck of deep ultraviolet light sources. Scientific Reports 5, 8332
(2015).
Khan, M. A. et al. 13 mW operation of a 295 -- 310 nm AlGaN UV-B LED with a p-
AlGaN transparent contact layer for real world applications. Journal of Materials
Chemistry C 7, 143-152 (2019).
Priante, D. et al. Highly uniform ultraviolet-A quantum-confined AlGaN nanowire LEDs
on metal/silicon with a TaN interlayer. Opt. Mater. Express 7, 4214-4224 (2017).
Sadaf, S. M. et al. An AlGaN core -- shell tunnel junction nanowire light-emitting diode
operating in the ultraviolet-C band. Nano Letters 17, 1212-1218 (2017).
Growden, T. A. et al. Near-UV electroluminescence in unipolar-doped, bipolar-tunneling
GaN/AlN heterostructures. Light: Science & Applications 7, 17150 (2018).
Kolbe, T. et al. Optical polarization characteristics of ultraviolet (In)(Al)GaN multiple
quantum well light emitting diodes. Applied Physics Letters 97, 171105 (2010).
15
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
FIGURE CAPTIONS
Figure 1
(a) Schematic structure of AlInN nanowire on GaN template. (b) TEM image of
AlInN/GaN nanowire in which the presence of core-shell structure is clearly shown.
EDXS line scan profile showing the quantitative variation of Ga, In and Al signals
along the line 1-2. (c) and variation of In and Al signal along the line 3-4.
Figure 2
(a) Schematic illustration of AlInN nanowire LED structure on Si. (b) 45o tilted
scanning electron microscopy image of typical AlInN nanowire LED sample showing
uniform nanowires on Si.
Figure 3
(a) Photoluminescence spectra of AlInN/GaN nanowires. The peak emission varies
from 290 nm to 355 nm. (b) Photoluminescence peak wavelength versus estimated Al
composition. (c) Temperature dependent photoluminescence intensity of AlInN/GaN
nanowires.
Figure 4 (a) I-V characteristics of the AlInN UV nanowire LED. The inset shows I-V
characteristic of the AlInN UV LED device in semi-log scale. (b) Electroluminescence
(EL) spectra of the AlInN UV nanowire LEDs under injection current range of 5-100
mA.
Figure 5 (a) TM and TE polarized spectra of the AlInN UV nanowire LED with emission
wavelength at 295 nm measured at 10 A/cm2. (b) The simulation result of TM and TE
polarized lights of AlInN nanowire LEDs at 295 nm.
Figure 6 (a) The simulated electron current density of the AlInN nanowire LEDs with and without
using EBL showing similar trend of electron current distribution. (b) The simulated I-
V characteristics of AlInN nanowire LEDs with and without EBL.
16
Figure 1
17
(a)
(b)
Figure 2
18
Figure 3
(a)
(b)
(c)
19
Figure 4
(a)
(b)
20
(a)
(b)
Figure 5
21
Figure 6
(a)
(b)
22
|
1801.08498 | 1 | 1801 | 2018-01-25T17:37:11 | Simulation studies of CZT(S,Se) single and tandem junction solar cells towards possibilities for higher efficiencies up to 22% | [
"physics.app-ph"
] | We simulated photovoltaic characteristics of single heterojunction solar cell with Cu2ZnSnS4 and Cu2ZnSnSe4 absorber layer numerically using one dimensional solar cell capacitance simulator (SCAPS-1D). n-CdS/ZnO double buffer layer is used for hetrostructure interfaces with the absorber layer. The cell performance is investigated against variation of different material layer properties such as thickness, carrier concentration, defect density etc. The performance is optimized first for the single junction solar cell with Mo as back contact material with work function 5 eV. A double junction CZTS/CZTSe tandem cell structure is realized keeping the same material properties as is used in the single CZTS and CZTSe solar cell simulation and considering the flat band condition at the interface. Tandem cell performance is determined after matching the current condition for top and bottom sub-cells. The CZTS/CZTSe short circuit current density is ~ 20.98 mA/cm2 for current matched 211.33 nm thick CZTS top cell in conjunction with 2000 nm bottom cell. The maximum efficiency obtained under the flat band condition at the contact is ~21.7% with open circuit voltage ~1.324 V. | physics.app-ph | physics | Simulation studies of CZT(S,Se) single and tandem junction solar cells towards possibilities
for higher efficiencies up to 22%
Goutam Kumar Gupta1 and Ambesh Dixit1
1Department of Physics and Center for Solar Energy, Indian institute of technology Jodhpur, Rajasthan,
342037, India
Corresponding author email: [email protected]
Abstract
We simulated photovoltaic characteristics of single heterojunction solar cell with Cu2ZnSnS4 and
Cu2ZnSnSe4 absorber layer numerically using one dimensional solar cell capacitance simulator (SCAPS-
1D). n-CdS/ZnO double buffer layer is used for hetrostructure interfaces with the absorber layer. The cell
performance is investigated against variation of different material layer properties such as thickness,
carrier concentration, defect density etc. The performance is optimized first for the single junction solar
cell with Mo as back contact material with work function 5 eV. A double junction CZTS/CZTSe tandem
cell structure is realized keeping the same material properties as is used in the single CZTS and CZTSe
solar cell simulation and considering the flat band condition at the interface. Tandem cell performance is
determined after matching the current condition for top and bottom sub-cells. The CZTS/CZTSe short
circuit current density is ~ 20.98 mA/cm2 for current matched 211.33 nm thick CZTS top cell in
conjunction with 2000 nm bottom cell. The maximum efficiency obtained under the flat band condition at
the contact is ~21.7% with open circuit voltage ~1.324 V.
Keywords: Compound semiconductor; Cu2ZnSnS4; Cu2ZnSnSe4 kesterites; Tandem Solar Cell
0.00.20.40.6-40-30-20-10040080012001600020406080100EQE (%) CZTS CZTSeWavelength (nm)Current density (mA/cm2) CZTS CZTSeVoltage (V)Voc (mV)615.1491.4Jsc (mA/cm2)30.1641.23FF (%)72.373.45Eff (%)13.4114.88-3-2-1010.00.51.01.52.0-3-2-101EvEcAl:ZnOZnOCdSCZTSe CdSZnOAl:ZnOEcCZTSEnergy (eV)Thickness (m)Ev Introduction:
Kesterite compound semiconductors Cu2ZnSn(SxSe1-x)4 (CZTSSe) received considerable attention in the
last decade due to its potential as a future absorber material for high efficiency thin film solar cell
technology. The properties such as environmentally benign and earth abundant constituent elements,
optimum direct bandgap (1-1.5 eV) and high absorption coefficient make CZTSSe a material of choice as
an alternative to the existing solar photovoltaic thin film technologies, which are expected to scale upto
several GW/year [1]. Considerable developments are reported towards improvement in the photovoltaic
performance of CZTSSe kesterite based solar cells with the maximum efficiency ~ 9.2% for pure sulfide
(CZTS) [2] and 11.6% for pure selenide (CZTSe) [3] based kesterite absorbers. The best kesterite cell
efficiency reported till date is ~ 12.7% for sulfoselenide (CZTSSe) based absorber with double CdS/In2S3
as n type buffer layer and 12.6 % with single CdS buffer layer [4][5]. However, in spite of all these
improvements, the maximum efficiency achieved for CZTSSe kesterite based absorbers is far below
chalcopyrite CuInGaSe2 (CIGS)[6] and CdTe based solar cells[7].
The similar crystal structure and material properties of kesterite provide the possibility of improvement in
photovoltaic conversion efficiency. Al:ZnO/i-ZnO/CdS/CZTS,Se/Mo/SLG kesterite solar cell device
structure is adopted from its chalcopyrite (CIGS) counterpart owing to its similar crystal structure and
electronic properties[8]. The high short circuit current density with low open circuit voltage is always a
trade-off with low energy band gap photovoltaic materials in a single junction configuration. Thus, it
becomes essential to utilize the improved device structures by implementing the tandem structures
(CZTS,Se) kesterite absorber to achieve the enhanced photovoltaic response for such material (CZTS,Se)
material systems.
The tandem or multi-junction solar cell structures utilizes stacking of different band gap p-n junction solar
cells in a specific configuration where the top cell should absorb the high energy portion of the solar
spectrum, equivalent to absorber band gap, and the lower cells should absorb the part of the remaining
spectrum, equivalent to the respective absorber band gaps. This tandem structure concept is well
established with thin film solar cells, based on group III-V nitride based compound semiconductors and
are used for space applications[9].
Recently, CIGS based stacked/tandem solar cell devices are explored, where CuGaSe2 (CGS) absorber
based solar cell is used as the top cell (Eg ~ 1.7 eV) and CIGS absorber cell has been used as the bottom
cell (Eg ~1.14eV). Such CIGS tandem solar cell devices showed improved efficiency up to 7.2% with
short circuit current 10.6 mA/cm2 and an open-circuit voltage of 1.18V [10]. A similar silver and indium
modified tandem solar cell structures e.g. Ag(In0.2Ga0.8)Se2 (AIGS) as top cell (Eg ~ 1.7 eV) and CIGS
based solar cell as the bottom cell has also been reported with maximum efficiency, Jsc, and Voc values ~
8%, 9.1mA/cm2 and 1.3V, respectively [11].
In this work, we have considered tandem structure using CZTS/Se based absorber materials. The one
dimensional electrical solar cell simulation program SCAPS-1D is used for simulating the device
performances [15]. The desired material parameters, required for the present simulation studies are either
borrowed from literature, cited wherever used or assumed with rationale for better understanding of the
device performance under realistic situations. Initially, a single junction solar cell is optimized for both
CZTS and CZTSe absorber layers independently and the effect of absorber layer thickness, carrier
concentration and density of various defects are investigated. The studies are extended further for the
tandem solar cell structures for realizing the enhanced photo conversion efficiencies. The proposed
tandem solar cell structures based on CZTS/Se absorber material have shown photo conversion efficiency
upto ~22%.
Materials' properties and simulation approach:
The considered device structure, shown in Fig 1 (a), includes CZTS and CZTSe as the p- type absorber
layer, CdS as n-type wide band gap buffer layer, ZnO as a window layer, also acting as a passivation layer
and Al doped ZnO (Al:ZnO) as a transparent conducting oxide (TCO) layer. The material properties are
summarized in Table 1, used for simulating the photovoltaic response for the proposed single and tandem
device structures. The optical absorption spectra for the CZTS and CZTSe are taken from literature[12]
and for CdS and ZnO the absorption spectrum file as provided in SCAPS are used and are shown in Fig 1
(b). These absorption data for different layers are used in the present work to mimic the practical
absorbing properties.
Figure 1: (a) Schematic representation of ―Al:ZnO/ZnO/CdS/Absorber layer/Back contact‖ single
300600900120015000.04.0x1078.0x1071.2x1081.6x108Absorption coefficient (m-1)Wavelength (nm) CZTS CZTSe CdS ZnO junction solar cell structure, used for simulating the device performance and (b) the absorption
coefficients of different materials used in the present work.
Table 1: Material parameters, used in the present work for simulating single and tandem junction solar
cell structures
Material properties
CZTSe
CZTS
CdS
i-ZnO
Al:ZnO
Thickness [m]
Bandgap [eV]
2
1
2
1.5
Electron Affinity [eV]
4.46[13]
4.3[13]
0.05
2.42
4.5
0.08
3.37
4.6
0.2
3.37
4.6
Dielectric permittivity
9.1[12]
6.95[12]
9[14]
9[14]
9[16]
Density of states in CB [cm-3]
2.2 x 1018
2.2 x 1018
1.8x1019[1
2.2 x
2.2 x 1018
4]
1018[14]
Density of states in VB [cm-3]
1.8 x 1019
1.8 x 1019
2.4x1018[1
1.8 x
1.8 x 1019
Thermal velocity of electron [cm/s]
Thermal velocity of hole [cm/s]
Electron mobility [cm2/Vs]
Hole mobility [cm2/Vs]
Donor concentration [cm-3]
1 x 107
1 x 107
145
40
0
Acceptor concentration [cm-3]
5 x 1016
Absorption coefficient [cm-1eV1/2]
file[12]
1 x 107
1 x 107
100
35
0
5 x 1016
file[12]
4]
1 x 107
1 x 107
1019[14]
1 x 107
1 x 107
1 x 107
1 x 107
160
150
150
50[14]
25[14]
25[16]
1 x 1017
1 x 1017
1x1020
0
0
SCAPS
SCAPS
Radiative recombination coefficient
1.04 x 10-10
1.04 x 10-10
1.04 x 10-
1.04 x
[cm-3/s]
Effective mass electron
Effective mass hole
0.07
0.2
Hole capture cross section (cm2)
1x 10-15
Electron capture cross section (cm2)
1x 10-15
Minority carrier lifetime
5 ns
Interface recombination speed (cm/s) 103[18]
10
10-10
0.18[16]
0.25[14]
0.275[14] 0.275[16]
0.71[16]
1x 10-15
1x 10-15
10 ns
104[18]
0.7[14]
0.59[14]
0.59[16]
1x 10-13
1x 10-15
1x 10-
15[16]
1x 10-15
1x 10-15
1x 10-15
Defect type at bulk/interface
Donor/Neutral Donor/Neutral
The proposed device structures are simulated using the one dimensional numerical simulator SCAPS[19],
where coupled Poisson and continuity equations for both electrons and holes with the suitable boundary
conditions, defined at the interfaces and contacts are solved numerically[19]. The single junction solar
cell structures with CZTS and CZTSe absorber layer are defined in SCAPS, as shown in Fig.2 (a), in
conjunction with different materials' thin films stacks and incident light towards the front contact.
This structure, Al:ZnO/ZnO/CdS/Absorber layer/Back contact, is considered for the simulation of single
junction solar cells with CZTS and CZTSe as the absorber layer materials. Al:ZnO is providing the front
electrical contact and simultaneously remains invisible/transparent to the incident solar radiation. ZnO
layer serves as the window layer and n-type CdS layer acts as the buffer layer, making a heterostructure
p-n junction with p-type CZTS/Se absorber layer.
The donor defects are introduced at 0.6 eV above valance band, in absorber layer and neutral defects at
absorber layer/CdS hetero-interface with energy 0.6 eV above valance band. The defect energy level are
taken to ascertain mid gap defect. Mo back contact is considered with work function about 5 eV and
surface recombination speed ~ 1 x 105 cm s-1 and 1 x 107 cm s-1 for electrons and holes, respectively at the
back contact. Thermionic emission is considered for the transport properties of majority charge carrier.
The band to band recombination is considered for the bulk materials and also for radiative recombination
with recombination coefficient equal to 1.04 x10-10 cm3s-1 for all the layers in the simulation[20]. Carrier
transport through tunneling mechanism is also considered at each interfaces and the effective mass
considered during the simulation is given in Table 1. Auger recombination is not considered for the
present studies, as it is significant only at higher carrier concentrations. For simplicity of calculation
capture cross section and thermal velocity for electron and holes are kept same 10-15 cm2 and 107 cm/s
respectively. The minority carrier life time at the bulk of absorber layer and the interface recombination
speed at absorber/buffer layer interface are taken 10 ns, 5 ns, and 104 cm s-1 ,103 cm s-1 for CZTS and
CZTSe respectively [18]. This assumption for the base device falls well within the range of mostly
reported kesterite devices. The absorption coefficient as a function of wavelength is shown in Fig 1(b) for
different layer of the devices, used in the simulation. This is more realistic in contrast to considering a
constant value for the simulation, showing more insights in the device performance.
Device analysis for a single junction solar cell with CZTS and CZTSe absorber layers:
The single junction solar cell device structures with CZTS and CZTSe absorber layers, as shown in Fig.
1(a), are first simulated using various materials parameters, as summarized in Table 1 for both CZTS and
CZTSe absorber layers. The estimated current voltage characteristics are shown in Fig. 2(b) for devices
based on these absorbers, with respective energy band diagrams, Fig. 2(a), showing the alignment of
respective bands near interfaces for these solar cell structures. We observed spike like band alignment at
CZTS/CdS interface. The band alignment at the heterointerface is critical for the device performance.
The large cliff in the band structure at heterointerface lowers the open circuit voltage and spike type
heterostructure reduces the short circuit current density (Jsc) of the device. That's why the optimization of
heterostructure interface is necessary for enhanced performance. The photovoltaic characteristics for
single junction device showed large short circuit current density for CZTSe absorber and large open
circuit voltage for CZTS absorber based devices, as shown in Fig. 2(b). The observed difference in short
circuit current density can be well understood by observing the quantum efficiency plot shown in the inset
of Fig. 2(b). The quantum efficiency curve covers large area owing to the smaller band gap of CZTSe
absorber layer, thus allowing lower energy photons to participate in generating photocurrent, thereby
increasing the short circuit current density of the cell. High open circuit voltage for the CZTS absorber
cell is due to its larger band gap as compared to that of CZTSe absorber. The optimal photovoltaic
efficiencies observed are ~ 13.41% and 14.88% for CZTS and CZTSe absorber based photovoltaic
devices, respectively. The respective device parameters are summarized in a table as an inset in Fig 3(b).
(a)
(b)
Figure 2: (a) Band alignments at different interfaces and (b) Current-voltage (I-V) characteristics for
single junction CZTS and CZTSe cells with insets showing external quantum efficiency (EQE) versus
wavelength variation (top left) and device parameters (bottom right table).
Impact of absorber layer thickness on single junction photovoltaic performance:
The influence of absorber layer thickness is investigated for these device structures and the observations
are summarized in Fig. 3. We found that CZTS solar cells exhibit relatively larger open circuit voltage,
Fig. 3(a), as compared to that of CZTSe absorber layer based solar cell. However, there is no significant
effect of thickness on open circuit voltage (Voc) of the device and a small increase with thickness in
noticed, approaching towards the optimal Voc for CZTS. This increase is attributed to the overall increase
-3-2-1010.00.51.01.52.0-3-2-101EvEcAl:ZnOZnOCdSCZTSe CdSZnOAl:ZnOEcCZTSEnergy (eV)Thickness (m)Ev0.00.20.40.6-40-30-20-10040080012001600020406080100EQE (%) CZTS CZTSeWavelength (nm)Jsc (mA/cm2) CZTS CZTSeVoltage (V)Voc (mV)615.1491.4Jsc (mA/cm2)30.1641.23FF (%)72.373.45Eff (%)13.4114.88 in majority carriers in bulk absorber, as detailed in next section. The short circuit current density (Jsc) is
shown in Fig. 3(b) for these cells, increases significantly with increasing the thickness of absorber layers.
The cumulative effect results in the increased device efficiency, Fig. 3(c). The efficiency of these single
junction solar cells increases with increasing thickness and showing saturation after a critical thickness.
The thickness of about 2 m seems sufficient to achieve the maximum photovoltaic response. This result
is in agreement with Beer-Lamberts law , suggesting enhanced absorption in the absorber
layer resulting into the excess photogenerated carriers and thus, improved device efficiency. The variation
in quantum efficiency against wavelength is shown Fig. 3(d) for CZTS absorber solar cell, with inset
showing the same for CZTSe solar cell for different absorber thicknesses. A large area is covered under
the quantum efficiency curve for CZTSe absorber cell as compared to that of CZTS absorber cell because
of its lower band gap. The lower band gap of CZTSe also allows the enhanced absorption at lower
wavelength, resulting in improved quantum efficiency. This is also reflected in the current density plot,
Fig. 3(b), showing relatively larger current density for the CZTSe cell as compared to that of CZTS cell
for same thickness.
(a)
(b)
(c)
(d)
Figure 3: (a) Open circuit voltage; (b) short circuit current density; (c) efficiency, and (d) quantum
efficiency of CZTS and CZTSe absorber solar cells for different thickness of the absorber layer.
0.00.51.01.52.02.53.00.400.450.500.550.600.650.00.51.01.52.02.53.0152025303540450.00.51.01.52.02.53.0681012141640080012001600020406080100Voc (V) CZTS CZTSeThickness (nm)Jsc (mA/cm2)Thickness (nm)Efficiency (%) CZTS CZTSeThickness (nm)40080012001600020406080100CZTSeEQE (%)Wavelength (nm)CZTSEQE (%)Wavelength (nm) 0.25 0.5 0.75 1 1.5 2 2.5 3 Impact of absorber layer acceptor concentration on single junction photovoltaic performance
CZT(S/Se) is an intrinsic p-type photovoltaic absorber material, where carrier concentration relies on the
non-stoichiometry and defects present in the synthesized material. The carrier concentration is varied
from 1 x1012 cm-3 to 1 x1018 cm-3 for the absorber layer in the present work. The open circuit voltage, Voc,
Fig. 4(a), and the short circuit current density, Jsc , Fig. 4(b), increases steadily and then starts decreasing,
with increasing the carrier concentration. The calculated results are in agreement with the fact that
increased carrier concentration should reduce the minority carrier life time, implying more recombination
and thus, reducing the collection of charge carriers at the contacts. Further, increased carrier concentration
decrease the depletion width towards absorber layer, thus affecting the effective separation of photo
generated charge carriers and causing the reduction in the short circuit current density. The open circuit
voltage increases with doping initially and then decreases, because of the compensating nature of donor
type defects considered for the system and thus, optimum carrier concentration should be chosen in such a
way that the large peak power can be achieved. The simulated efficiency versus acceptor density results
(see Fig 4(c)) suggest that the optimal carrier density is up to 5 x 1016 cm-3 for CZTS and 1 x 1017 cm-3 for
CZTSe solar cell to achieve the maximum photovoltaic performance. The calculated maximum
efficiency, Voc and Jsc with these carrier densities are ~13.41%, 615m V and 30.15 mA/cm2, respectively
for CZTS and ~15.6%, 514.29 mV, 40.52 mA/cm2 for CZTSe solar cell, respectively.
(a)
(c)
(b)
Figure 4: (a) Open circuit voltage; (b) short circuit current density; and (c) efficiency as a function of
carrier concentration for CZTS and CZTSe absorber based solar cells.
10121013101410151016101710180.00.10.20.30.40.50.60.7Voc(V) CZTS CZTSeCarrier concentration (cm-3)1012101310141015101610171018510152025303540Jsc(mA/cm2) CZTS CZTSeCarrier concentration (cm-3)1012101410161018051015Efficiency (%) CZTS CZTSeCarrier concentration (cm-3) Impact of defect concentration at absorber and at absorber/CdS interface
CZTS,Se is a defect prone system and consists of acceptor and donor vacancies, antisites and interstitials.
Low formation energy of acceptor defects in the CZTS,Se system makes it an intrinsic p-type system
without additional foreign dopant elements. There is provision to introduce different donor and acceptor
defects in SCAPS, however available characteristic data for all kinds of defects possible in CZTS,Se
system is rare. For simplicity of the simulation, we considered single donor bulk defects in the absorber
layer and neutral defects in absorber/CdS interface. The defects are considered at mid band with
characteristic energy level at 0.6 eV above valance band. The concentration of bulk defects determines
the lifetime of minority carrier in the absorber while defects at the interface determines the recombination
speed (S) at the interface varying the defect density in the absorber and absorber/CdS interface. The
change in efficiency is summarized in Fig 5(a) and 5(b) against absorber and absorber/CdS interface
defect concentrations, respectively. We can see that for moderate defect concentrations, especially in
absorber, the efficiency is not changing significantly. This is probably due to the lower defect-
photogenerated carrier scatterings because of lower defect concentration and thus, not affecting the device
performance. However, after certain defect concentration, in the range of high 1016 cm-3, the efficiency
drops drastically to the very low values, close to zero for both CZTS and CZTSe absorbers. In contrast,
absorber/CdS interface defects seems more prone in degrading the photovoltaic performance and any
increase in interface defect concentration resulted in reduced photovoltaic efficiency, Fig 5(b). Thus,
absorber/buffer interface defects should be minimized to a possible lower level to realize the enhanced
photovoltaic efficiency. Further, we investigated the variation in absorber/CdS interface recombination
(IR) speed against minority carrier life time for different photovoltaic efficiencies and results are
summarized as contour plots in Fig 5(c) and 5(d) for CZTS/CdS and CZTSe/CdS interfaces, respectively.
These measurements suggest that higher minority carrier life time and lower interface recombination (IR)
speeds are very critical to realize the enhanced photovoltaic response for both CZTS and CZTSe absorber
based photovoltaic cells.
(a)
(b)
(c)
(d)
Figure 5: Impact of defect concentration (a) and absorber/CdS interface defect concentration (b) in
efficiency for CZTS and CZTSe solar cells and contour plots showing variation of efficiency with
minority carrier life time and interface recombination speed in CZTS (c); and CZTSe (d).
Impact of buffer layer
CdS is the most commonly used buffer layer in CZTSSe solar cell. The performance of a solar cell
depends on the band alignment at the heterointerface. CdS with wide bandgap of ~2.42 eV allows the
maximum photoabsorption in the absorber layer. The thickness of a buffer layer mainly depends on its
conductivity. Higher is the conductivity of buffer layer, more will be the penetration of depletion region
towards the absorber side and thus the higher device efficiency. It is suggested that a very thin layer of
CdS is required for enhanced solar performance. The thin layer of CdS will also assist in avoiding
absorption loss in CdS buffer layer. Considering the same, thickness of CdS is varied at different carrier
concentration and the variation of carrier concentration as a function of CdS buffer layer thickness is
plotted as contour plots for different efficiency in Fig 6(a, b) for both CZTS and CZTSe absorber based
101410151016101710180481216Efficiency (%) CZTS CZTSeAbsorber defect concentration (cm-3)108101010121014101681012141618Efficiency (%) CZTS CZTSeAbsorber/CdS defect concentration (cm-3)17.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0717.2716.2415.2014.1713.1412.1111.0710-810-710-6101102103104105106CZTSCZTSEfficiency (%)Minority carrier lifetime (s)CZTS/CdS IR speed (cm/s)9.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.729.03110.2611.4916.4215.1913.9612.7210-810-710-6101102103104105106CZTSeCZTSEfficiency (%)Minority carrier lifetime (s)CZTSe/CdS IR speed (cm/s) solar cell devices. These results suggest that the higher CdS carrier concentration is essential for the
optimal device performance. In contrast, the device efficiency is relatively insensitive to its thickness for
CZTS, while for CZTSe at lower CdS carrier concentration about 55nm of thickness is required to
achieve the constant efficiency, Fig 6(b). Further, increase in carrier concentration of CdS buffer layer
does not show any significant improvement on the device performance.
(a)
(b)
Figure 6: Impact of CdS buffer layer thickess and carrier concentration on CZTS (a) and CZTSe (b)
solar cell efficiency
The conduction band offset (CBO) at the absorber/buffer interface plays an important role in governing
the carrier transport through the junction. The offset is primarily determined by the difference in the
values of electron affinity of the absorber and buffer layers. The positive and negative band offsets govern
the spike and cliff like heterostructures, respectively. In present simulation, electron affinity of CdS is
varied to see the effect of band offset at hetero interface in the device performance. Conduction band
alignment at the heterostructure interface is shown in Fig 7(a, b) for CZTS and CZTSe solar cells. As we
go on increasing the electron affinity of buffer layer the band alignment at the absorber/buffer layer
interface changes from spike to cliff.
13.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.4513.0513.3313.6114.7313.8914.1714.450.020.040.060.080.100.120.14101710181019CZTSCdS thickness (m)CdS carrier concentration (cm-3)Efficiency (%)12.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.6012.2312.6313.0213.4213.8114.2014.600.020.040.060.080.100.120.14101710181019CZTSeEfficiency (%)CdS thickness (m)CdS carrier concentration (cm-3) Figure 7: Conduction band offset (CBO) in (a) CZTS (b) CZTSe at different CdS layer electron affinity
(χCdS)
We observed that a cliff in the heterostructure results in a decrease in open circuit voltage, Fig 8; whereas
short circuit current remains relatively unchanged, except at very low CdS electron affinity values.
Further, the efficiency versus CdS electron affinity variation is summarized in Fig 8 (c), showing inverted
U shaped behavior. Thus, a moderate electron affinity range 4.0 eV – 4.4 eV is essential to realize the
maximum efficiency for both CZTS and CZTSe absorber based photovoltaic devices.
(a)
(b)
010101010.00.12.0010.6418 eV CZTSCliffSpikeBarrier heightCBO 0.3 eV 0.5 eVCdS4.0 eV 3.8 eV -0.3 eV -0.1 eV 0.1 eV4.2 eV Energy (eV)4.4 eV 4.6 eV Thickness (m)010101010.00.12.001Barrier height 0.2943 eV Spike 0.66 eV 3.8 eV CliffCdSCBO 0.46 eV4.0 eV 0.06 eV 0.26 eV 4.6 eV 4.4 eV 4.2 eVCZTSe -0.14 eVEnergy (eV)Thickness (m)3.63.84.04.24.44.64.85.00.20.40.60.81.0Voc (V) CZTS CZTSeCdS electron affinity(eV)3.63.84.04.24.44.64.85.0051015202530354045Jsc (mA/cm2) CZTS CZTSeCdS electron affinity(eV)
(c)
Figure 8: Impact of CdS electron affinity (band offset at absorber/CdS interface) on Voc (a); Jsc (b); and
efficiency (c) for CZTS and CZTSe solar cells
These observations suggest that the choice of buffer layer material with CZTSSe absorber material is
crucial and should provide the desired band offset for synthesized photovoltaic devices. We further
carried out studies for different electron affinities of the buffer layer material and the interface
recombination (IR) speed on open circuit voltage, short circuit current and efficiency, to explore the
selection parameters/criteria for a buffer. The results are summarized in Fig 9 as contour plots between
absorber layer/CdS interface recombination (IR) speed and CdS electron affinity with open circuit
voltage, short circuit current and efficiency. It can be seen from the contour plot of Voc Fig 9 (a & b) and
Jsc, Fig 9 (c & d) that cliff in the band structure (χCZTS,Se < χBuffer) affects Voc of the device, whereas Jsc is
nearly insensitive to the same for both CZTS and CZTSe absorber based photovoltaic devices. Voc
decreases with the increase in cliff at the heterointerface. In contrast, Jsc showed decrease with spike
(χCZTS,Se > χBuffer) like band offset at the heterointerface. The interface recombination (IR) speed is
showing a negative impact on the performance and thus, a low interface recombination speed is desired
for enhanced photovoltaic response.
3.63.84.04.24.44.64.85.0051015Efficiency (%) CZTS CZTSeCdS electron affinity(eV) (a)
(c)
(e)
(b)
(d)
(f)
Figure 9: Impact of interface recombination speed and CdS electron affinity on CZTS (a) Voc,(c) Jsc (e)
efficiency and CZTSe (b)Voc (d) Jsc (f) efficiency
Impact of Back contact work function
The back contact material shows severe impact on the solar cell performance. A relatively higher metal
work function is required to realize an ohmic contact to the absorber material. Molybdenum is commonly
used as back contact material for CZTSSe absorber based photovoltaic devices. However, molybdenum is
not the optimum choice as the back contact because of its non-ohmic contact characteristics with kesterite
0.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66180.26420.33050.39680.46300.52930.59550.66184.04.24.44.64.8101102103104105106CZTSCdS electron affinity (eV)Voc (V)CZTS/CdS IR speed (cm/s)0.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45210.18290.22780.27260.31750.36240.40730.45214.04.24.44.64.8101102103104105106CZTSeVoc (V)CdS electron affinity (eV)CZTSe/CdS IR speed (cm/s)29.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.1129.6830.2629.8329.9730.114.04.24.44.64.8101102103104105106CZTSCdS electron affinity (eV) Jsc (mA/cm2)CZTS/CdS IR speed (cm/s)40.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.8940.1740.3541.2540.5340.7141.0740.894.04.24.44.64.8101102103104105106CZTSeCdS electron affinity (eV)Jsc (mA/cm2)CZTSe/CdS IR speed (cm/s)4.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.2196.0887.9569.82511.6915.4313.564.04.24.44.64.8101102103104105106CZTSCdS electron affinity (eV)Efficiency (%)CZTS/CdS IR speed (cm/s)4.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.5505.8697.1888.5069.82511.1412.4613.784.04.24.44.64.8101102103104105106CZTSeCdS electron affinity (eV)Efficiency (%)CZTSe/CdS IR speed (cm/s) CZTSSe absorbers. We systematically investigated the impact of back contact work function by varying it
from 4.4 to 5.5 eV and the results are summarized in Fig 10. Increasing metal work function decreases the
barrier height for majority charge carrier at back contact interface thereby improved performance is
observed. The open circuit voltage, short circuit current and efficiency show the similar trend, increasing
initially with increasing metal work function and above a critical metal work function, these values show
saturation behavior. Thus, a high metal work function is required for the enhanced solar cell performance,
which will allow Ohmic-contact formation at the back electrode. These observations suggest that metal
work function above 5.3 eV and above 5 eV is required for CZTS and CZTSe absorber based solar cell.
(a)
(b)
Figure 10: Impact of metal work function on Voc (inset Jsc) (a) and efficiency (b) for CZTS and CZTSe
solar cell
Design and analysis of CZTS/CZTSe tandem solar cell structure photovoltaic device:
Single junction solar cell performance is limited by the photon absorption for the respective band gap of
absorber layer of the cell. The performance of the solar cell can be improved by stacking single junction
solar cell with different band gap absorbers in such a way that the higher energy photons are absorbed in
the top cell and the lower energy photons are absorbed in respective lower solar cells. In CZTS/CZTSe
tandem cell, as shown in Fig 11, top cell is made of larger band gap CZTS absorber layer (Eg ~ 1.5 eV)
and the bottom cell is made of lower band gap CZTSe bottom cell (Eg ~ 1 eV). The tandem cell structures
can be realized by fabricating the top cell on the transparent conducting surface on the upside layer of the
bottom cell. For example, S. Nishiwaki et al. fabricated the stacked chalcopyrite tandem solar cell by
connecting CuGaSe2 and CuInGaSe2 cells in series and reported the cell efficiency ~ 7.4% with ~ 1.2V
open circuit voltage [10].
The fabrication of tandem cell structures is possible using monolithic integration of one cell on the top of
other cell. The stacking of such two cells forms a reverse biased p-n junction at the interface that hinders
4.504.755.005.255.500.00.20.40.60.81.04.504.755.005.255.50010203040Jsc (mA/cm2) CZTS CZTSeBack contact work function (eV)Voc (V) CZTS CZTSeBack contact work function (eV)4.64.85.05.25.40510152025Efficiency (%) CZTS CZTSeBack contact work function (eV) the current flow across the device. An n+/p+ tunnel junction is essential at the interface to overcome the
reverse bias issues and for proper current flow across the tandem structure. The transparent materials such
as indium tin oxide (ITO), fluorine doped tin oxide (FTO), aluminum doped zinc oxide (AZO),
molybdenum doped indium oxide (IMO) with high mobility can serve well in making tunnel junction
between the cells. Tokio Nakada et al. fabricated chalcopyrite tandem solar cell with Ag(In0.2Ga0.8)Se2
(AIGS) upper cell on top of the transparent IMO and CIGS as the bottom cell. This tandem structure
showed ~ 8% photovoltaic efficiency with 1.3V open circuit voltage [11].
Figure 11: Schematic representation of tandem cell structure with CZTS top cell and CZTSe bottom cell
in conjunction with tandem connections
The simulation of tandem structure is quite cumbersome in SCAPS and needs to be considered carefully
by utilizing the concept of replacing the actual cell by an equivalent fake absorber layer. This can be
simulated by connecting the top and bottom cell in series, which are simulated independently for optimal
performance. However, the problem persists as the bottom cell should be illuminated through the adapted
solar spectrum, the residual solar spectra after its partial absorption across the top cell. A top cell is
considered as a fake CZTSe layer of same thickness as the original bottom cell to overcome this problem.
The fake CZTSe layer is considered p+ layer with its electron affinity to avoid any band discontinuity and
to ensure the optical presence and electrical absence of this fake CZTSe layer, while simulating the top
cell. Additionally, high defect density is considered for fake CZTSe layer to reduce the minority carrier
diffusion length and to ensure that lower energy spectrum (h < 1.5 eV) is not contributing to the
photocurrent of the tandem structure.
A fake CZTS layer is considered which can absorb the substantial fraction of the incident light as in the
real tandem cell to set the bottom cell for the tandems structure. This CZTS layer is taken as n+ layer with
sufficient electron affinity to avoid any band discontinuity for electrically inactiveness of this layer. The
high defect density is taken in the fake CZTS layer for small minority carrier diffusion length and to
ensure high energy solar spectra (h > 1eV) is not contributing to the photocurrent of the device.
The maximum current of a tandem cell is limited by the minimum current produced by the constituent
solar cells (sub cell) and the maximum open circuit voltage is limited by the sum of open circuit voltage
of the individual solar cells. Thus, the individual constituent cells do not limit the current conduction to
get the maximum advantage of the tandem cell. The constituent cells should be made of different bandgap
absorbing materials, which may cover the maximum solar spectrum while current is matched across these
cells. This can be achieved from the current density thickness profile for the tandem structure as shown in
Fig 12(a). The current matched conditions for the tandem structure are obtained at the point of
intersection of top and bottom cell current densities, see Fig 12(a). The matched short circuit current
density values are summarized in Fig 12(b) for top and bottom cell absorber thickness. We observed that
the thickness of the top CZTS cell for matched tandem cell structure varies significantly for different
bottom CZTSe cell thickness, Fig 12(b). This variation suggests that the current matching condition can
be achieved at relatively lower CZTS top cell thickness as compared to that of the bottom CZTSe cell
thickness, Fig 12(b). The tandem cell response improves with increasing the bottom CZTSe cell
thickness, as seen in the enhanced tandem cell current density, Fig. 12(a & c).
(a)
(b)
0.00.20.40.60.81.051015202530 0.25 0.5 1 1.5 2 2.5 3 Top cell JscJsc (mA/cm2)Top cell thickness (nm)0.00.51.01.52.02.53.0150175200225 Jsc matched top cell thicknessTop cell thickness(nm)Bottom cell thickness (nm) (c)
(d)
Figure 12: (a) Variation of current density of CZTS top cell and CZTSe bottom cell at different
thicknesses (b) Thickness of top CZTS cell matched with different thickness of bottom CZTSe cell (c)
Current density and open circuit voltage curve for tandem structure with current matched top and bottom
cell at different bottom cell thickness (d) Tandem structure efficiency at different bottom cell thickness.
The variation in tandem cell current density and open circuit voltage are shown in Fig. 12(c) for different
bottom cell thickness. The increased current density is attributed to the enhanced absorption of solar
energy resulting into the enhanced photogenerated charge carriers, and finally the enhanced open circuit
voltage. The tandem cell efficiency as a function of bottom cell thickness is shown in Fig. 12(d) under
current matching condition. We observed that tandem cell efficiency increases with cell thickness and
saturates at or above 2m thick bottom cell. The current - voltage characteristics and the quantum
efficiency are shown in Fig. 13 (a & b) for the tandem structure at 2m CZTSe thick bottom cell in
conjunction with 211 nm thick matched CZTS top cells. We observed the enhanced absorption of photon
energy at longer wavelength from a lower bandgap bottom cell and at shorter wavelength from a higher
bandgap top cell, resulting in photo current generation over a wide spectral range as seen in the quantum
efficiency curve Fig 13 (b). The advantage of the tandem structure can be seen in terms of improved open
circuit voltage, which is the sum of open circuit voltages of top and bottom cells. However, the maximum
current is limited by the current from CZTS top cell in the tandem cell configuration. This assisted in
achieving the higher open circuit voltage for the complete tandem structure with optimal current density.
The quantum efficiency (QE) curves are shown in Fig 13(b) for the top CZTS and bottom CZTSe cells.
These QE versus wavelength plots indicate that up to 826 nm (h > 1.5 eV) top cell is dominating,
whereas for longer wavelength (h < 1.5 eV) top cell remains invisible and significant absorption takes
place in the bottom cell, and thus improving the overall quantum efficiency for tandem structure in wide
0.00.51.01.52.02.53.01.301.311.321.33 Tandem cell Voc Tandem cell JscBottom cell CZTSe thickness (m)Voc (V)18192021Jsc (mA/cm2)0.00.51.01.52.02.53.0171819202122Tandem cell efficiency(%)Bottom cell CZTSe thickness (m) spectral range. The estimated device parameters are listed in Table 2 for top, bottom and tandem cell
structures. The maximum efficiency about 21.7% is achieved for the optimized tandem CZTS/CZTSe
solar cell structure in the investigated device configuration.
(a)
(b)
Figure 13: (a) Current-voltage characteristics and (b) quantum efficiency for CZTS top cell CZTSe
bottom cell and CZTS/CZTSe tandem cell
Table 2: CZTS and CZTSe absorber based photovoltaic cell performance parameters for single junction
and tandem cell configurations
Photovoltaic cell configuration with absorber
Jsc
Voc
thickness
(mA/cm2)
(mV)
211.33nm CZTS single cell
2 m CZTSe single cell
CZTS Top cell in tandem structure
CZTSe Bottom cell in tandem structure
Tandem cell
20.98
41.73
20.98
20.87
20.98
852.9
496.6
852.9
474.1
FF
(%)
Efficiency
(%)
81.95
14.67
74.61
15.46
81.95
14.67
1324.82
78.2
72.59
7.18
21.7
These studies suggest that CZTSSe based tandem structures may provide a way to realize the proposed
geometries and thus, the enhanced photovoltaic efficiency. The investigated tandem structure showed
21.7% photovoltaic efficiency for CZTS and CZTSe solar cells under current matching conditions with
open circuit voltage ~1.33 V and short circuit current density ~20.98 mA cm-2.
Conclusion
Kesterite CZTS/Se solar cells are investigated using one dimensional SCAPS 1D solar cell simulator. The
possible single junctions and tandem structures are simulated for their optimized photovoltaic
0.00.20.40.60.81.01.21.4-25-20-15-10-502000nm bottom Top cell Bottom cell Tandem cellCurrent density(mA/cm2)Voltage (V)40060080010001200140016000204060801002000nm bottom CZTS Top cell CZTSe Bottom cellEQE(%)Wavelength (nm) performance. The proposed tandem absorber cell structures show the potential to realize the enhanced
efficiency of kesterite based solar cells beyond the reported maximum efficiency ~12.7 % [4]. The
present studies showed the optimal efficiency ~22 % photovoltaic efficiency for tandem CZTS,Se
structures. This present analysis demonstrates the possibility of approaching higher efficiency beyond the
maximum with present set of tandem structure. Further improvement in the efficiency of tandem cells is
possible by introducing wider band gap chalcogenide such as CZGS (Cu2ZnGeS4) in tandem geometries.
Acknowledgement
The authors would like to thank Marc Burgelman and their team for providing SCAPS simulator and
financial support from Department of Science & Technology (DST), Government of India, through grants
DST/INT/Mexico/P-02/2016.
References
[1]
S. Abermann, ―Non-vacuum processed next generation thin film photovoltaics: Towards
marketable efficiency and production of CZTS based solar cells,‖ Sol. Energy, vol. 94, pp. 37–70,
2013.
[2]
H. S. Takuya Kato*, Homare Hiroi, Noriyuki Sakai, Satoshi Muraoka, ―Σ 2 . 4.,‖ Comp. A J.
Comp. Educ., vol. 27th Europ, no. c, pp. 1–3, 2010.
[3]
Y. S. Lee et al., ―Cu2ZnSnSe4 thin-film solar cells by thermal co-evaporation with 11.6%
efficiency and improved minority carrier diffusion length,‖ Adv. Energy Mater., vol. 5, no. 7,
2015.
[4]
J. Kim et al., ―High efficiency Cu2ZnSn(S,Se)4 solar cells by applying a double in 2S3/CdS
Emitter,‖ Adv. Mater., vol. 26, no. 44, pp. 7427–7431, 2014.
[5] W. Wang et al., ―Device characteristics of CZTSSe thin-film solar cells with 12.6% efficiency,‖
Adv. Energy Mater., vol. 4, no. 7, pp. 1–5, 2014.
[6]
P. Jackson et al., ―Properties of Cu(In,Ga)Se2 solar cells with new record efficiencies up to
21.7%,‖ Phys. status solidi – Rapid Res. Lett., vol. 9, no. 1, pp. 28–31, 2015.
[7]
R. M. Geisthardt, M. Topic, and J. R. Sites, ―Status and Potential of CdTe Solar-Cell Efficiency,‖
IEEE J. Photovoltaics, vol. 5, no. 4, pp. 1217–1221, 2015.
[8]
A. Chiril et al., ―Highly efficient Cu ( In , Ga ) Se 2 solar cells grown on flexible polymer films,‖
Nat. Mater., vol. 10, no. 11, pp. 857–861, 2011.
[9]
Y. Sayad, ―Photovoltaic potential of III-nitride based tandem solar cells,‖ J. Sci. Adv. Mater.
Devices, vol. 1, no. 3, pp. 379–381, 2016.
[10] S. Nishiwaki, S. Siebentritt, P. Walk, and M. C. Lux-Steiner, ―A stacked chalcopyrite thin-film
tandem solar cell with 1.2 V open-circuit voltage,‖ Prog. Photovoltaics Res. Appl., vol. 11, no. 4,
pp. 243–248, 2003.
[11] T. Nakada et al., ―Chalcopyrite thin-film tandem solar cells with 1.5 V open-circuit-voltage,‖
Conf. Rec. 2006 IEEE 4th World Conf. Photovolt. Energy Conversion, WCPEC-4, vol. 1, pp. 400–
403, 2007.
[12] S. Adachi, ―Physical Properties: Compiled Experimental Data,‖ Copp. Zinc Tin Sulfide-Based
Thin-Film Sol. Cells, pp. 149–179, 2015.
[13] S. Adachi, ―Earth-Abundant Materials for Solar Cells: Cu2-II-IV-VI4 Semiconductors,‖ Earth-
Abundant Mater. Sol. Cells Cu2-II-IV-VI4 Semicond., pp. 1–462, 2015.
[14] M. Courel, F. A. Pulgar??n-Agudelo, J. A. Andrade-Arvizu, and O. Vigil-Gal??n, ―Open-circuit
voltage enhancement in CdS/Cu2ZnSnSe4-based thin film solar cells: A metal-insulator-
semiconductor (MIS) performance,‖ Sol. Energy Mater. Sol. Cells, vol. 149, pp. 204–212, 2016.
[15] M. Patel and A. Ray, ―Enhancement of output performance of Cu2ZnSnS4 thin film solar cells-
A numerical simulation approach and comparison to experiments,‖ Phys. B Condens. Matter, vol.
407, no. 21, pp. 4391–4397, 2012.
[16] M. Courel, J. A. Andrade-Arvizu, and O. Vigil-Galán, ―Loss mechanisms influence on
Cu2ZnSnS4/CdS-based thin film solar cell performance,‖ Solid. State. Electron., vol. 111, pp.
243–250, 2015.
[17] A. Niemegeers, M. Burgelman, R. Herberholz, U. Rau, and D. Hariskos, ―Model for Electronic
Transport in Cu(In,Ga)Se2 Solar Cells,‖ Prog. PHOTOVOLTAICS Res. Appl. Prog. Photovolt.
Res. Appl. 6, 407±421, vol. 421, no. April, pp. 407–421, 1998.
[18] M. Courel, J. A. Andrade-Arvizu, and O. Vigil-Galán, ―The role of buffer/kesterite interface
recombination and minority carrier lifetime on kesterite thin film solar cells,‖ Mater. Res. Express,
vol. 3, no. 9, p. 95501, 2016.
[19] M. Burgelman and K. Decock, ―Marc Burgelman, Koen Decock, Alex Niemegeers, Johan
Verschraegen, Stefaan Degrave Version: 17 february 2016,‖ Manual, no. February, 2016.
[20] O. K. Simya, A. Mahaboobbatcha, and K. Balachander, ―A comparative study on the performance
of Kesterite based thin film solar cells using SCAPS simulation program,‖ Superlattices
Microstruct., vol. 82, pp. 248–261, 2015.
|
1808.04686 | 1 | 1808 | 2018-08-14T13:46:17 | Photoacoustic Sensing of Trapped Fluids in Nanoporous Thin Films: Device Engineering and Sensing Scheme | [
"physics.app-ph",
"physics.chem-ph"
] | Accessing fluid infiltration in nanogranular coatings is an outstanding challenge, of relevance for applications ranging from nanomedicine to catalysis. A sensing platform, allowing to quantify the amount of fluid infiltrated in a nanogranular ultrathin coating, with thickness in the 10 to 40 nm range, is here proposed and theoretically investigated by multiscale modelling. The scheme relies on impulsive photoacoustic excitation of hypersonic mechanical breathing modes in engineered gas-phase synthesised nanogranular metallic ultathin films and time-resolved acousto-optical read-out of the breathing modes frequency shift upon liquid infiltration. A superior sensitivity, exceeding 26x103 cm^2/g, is predicted upon equivalent areal mass loading of a few ng/mm^2. The capability of the present scheme to discriminate among different infiltration patterns is discussed. The platform is an ideal tool to investigate nano fluidics in granular materials and naturally serves as a distributed nanogetter coating, integrating fluid sensing capabilities. The proposed scheme is readily extendable to other nanoscale and mesoscale porous materials. | physics.app-ph | physics |
Photoacoustic Sensing of Trapped Fluids in
Nanoporous Thin Films: Device Engineering and
Sensing Scheme
Giulio Benetti,†,‡,¶ Marco Gandolfi,†,‡,§ Margriet J Van Bael,¶ Luca Gavioli,†,‡
Claudio Giannetti,†,‡ Claudia Caddeo,(cid:107) and Francesco Banfi∗,†,‡
Cattolica del Sacro Cuore, Via Musei 41, 25121 Brescia, Italy
†Interdisciplinary Laboratories for Advanced Materials Physics (I-LAMP), Universit`a
‡Dipartimento di Matematica e Fisica, Universit`a Cattolica del Sacro Cuore, Via Musei
¶Laboratory of Solid State Physics and Magnetism, Department of Physics and Astronomy,
§Laboratory for Soft Matter and Biophysics, Department of Physics and Astronomy, KU
(cid:107)Istituto Officina dei Materiali (CNR - IOM) Cagliari, Cittadella Universitaria, I-09042
41, 25121 Brescia, Italy
KU Leuven, Celestijnenlaan 200D, B-3001 Leuven, Belgium
Leuven, Celestijnenlaan 200D, B-3001 Leuven, Belgium
Monserrato (Ca), Italy
E-mail: [email protected]
Abstract
coatings
relevance for applications
Accessing fluid infiltration in nanogranu-
is an outstanding challenge,
lar
of
ranging from
nanomedicine to catalysis. A sensing plat-
form, allowing to quantify the amount of fluid
infiltrated in a nanogranular ultrathin coating,
with thickness in the 10 to 40 nm range,
is
here proposed and theoretically investigated
by multiscale modelling. The scheme relies
on impulsive photoacoustic excitation of hyper-
sonic mechanical breathing modes in engineered
gas-phase synthesised nanogranular metallic ul-
tathin films and time-resolved acousto-optical
read-out of the breathing modes frequency shift
upon liquid infiltration. A superior sensitiv-
ity, exceeding 26x103 cm2/g, is predicted upon
equivalent areal mass loading of a few ng/mm2.
The capability of the present scheme to dis-
criminate among different infiltration patterns
is discussed. The platform is an ideal tool to
investigate nanofluidics in granular materials
and naturally serves as a distributed nanogetter
coating, integrating fluid sensing capabilities.
The proposed scheme is readily extendable to
other nanoscale and mesoscale porous materi-
als.
Keywords
Ultrafast opto-mechanics, granular materials,
nano-mechanics, nanoporosity, getter materi-
als, Ag nanoparticles, wettability, mass sensing
1 Introduction
in nanoporous coatings
Nanofluidics
is an
emerging topic at the forefront of nanotechnol-
ogy. The subject, at the cross-road of physics,
material science and engineering,
is of rele-
vance both scientifically, 1,2 fluid dynamics at
the nanoscale differing significantly from its
macroscale counterpart, and from a technolog-
ical stand point, where applications have been
1
proposed throughout disparate fields, ranging
from biotechnology 3 and nanomedicine 4 -- 6 to
gas storage 7,8 and catalysis. 9
enviromental
ellipsometric
In this
context nanoporous films have
been proposed as new platforms to investi-
gate nanofluidics 1 and the related wettabil-
ity issues. 10,11 As for the technological drive,
whatever the application, the exploitation of
nanoporous getter coatings ultimately relies on
their capability to be infiltrated by a fluid. 12
Despite the demonstrated versatility and wide-
spreading of porous thin films, measurements
of their permeability remains an outstanding
issue.
This
issue is primarily tackled by con-
ventional
(EE)
porosimetry 13,14 and by gas adsorption mea-
surements. 15,16 Nevertheless
information re-
trieval from these techniques is not straight-
forward. EE requires surfaces and interfaces of
good optical quality and, eventually, exploita-
tion of an index-matching fluid at the interface
between the granular film and the supporting
substrate (an issue in the frame of porosimetry
measurements, where no other liquids should
be present other than the one under investi-
gation). For the case of EE the number of
free fitting parameters is rather vast 14,17,18 and,
in several instances, the results may be puz-
zling to interpret. 19 For instance, when dealing
with metal nano-objects, as in the present case,
modelling of the experimental data requires
taking into account size-dependent corrections
related to the presence of the surface plasmon
resonance, 18 adding to the complexity of the
information retrieval process. Furthermore, EE
porosimetry is based on recognition of minute,
often spectrally featureless variations of the el-
lipsometric angles Ψ and ∆ upon fluid infiltra-
tion. 20,21 These facts complicate the recognition
of fluid filling levels as opposed, for instance, to
a technique exploiting the variation of a specific
resonance. Similar difficulties are encountered
in interpreting gas adsorption measurements,
where the result may be biased by the spe-
cific model, among the many available ones, 22
chosen to fit the adsorption isotherm.
In this work we propose a novel sensing
platform aimed at investigating fluid infiltra-
2
tion in nanogranular coatings. We engineer,
via atomistic simulations, a gas-phase synthe-
sised nanogranular metallic coating with open-
porosity, specifically tailored for efficient ultra-
fast photoacoustic detection of the filling fluid.
The detection scheme relies on impulsive photo-
acoustic excitation of the device mechanical
breathing modes - in the 50 GHz frequency
range - and time-resolved acousto-optical read-
out of the frequency shift upon fluid infiltration.
From an applicative stand point, the gas-phase
synthesised nanogranular metallic scaffold is
readily exploitable as a distributed nanoget-
ter coating, integrating fluid sensing capabili-
ties and viable for multi-functionality. 23 -- 26 The
production technique is per se competitive ow-
ing to its simplicity, high throughput and flex-
ibility. 27,28
The proposed scheme bears great generality
and can be readily deployed to include other
nanoscale 29 or mesoscale porous systems. 30 -- 32
2 Materials and methods
Molecular Dynamics. All the simulations
have been performed with the LAMMPS pack-
age. 33 The velocity-Verlet algorithm was used
to solve the equations of motion and tempera-
ture was controlled by a Nos´e-Hoover thermo-
stat. The Ag-Ag interactions were described
with the 12-6 Lennard-Jones potential of Heinz
et al., with cutoff at 8 A. Further details are
reported in the SI.
Film's effective properties. The physi-
cal properties of the granular film have been
retrieved by importing the raw xyz Molecular
Dynamics (MD) results into MATLAB and by
dividing the 3D simulation domain in small vox-
els. The average pore sizes have been computed
by using an ImageJ plugin (BoneJ) 35 and the
method proposed by Sainto et al. for the bones'
trabeculae analysis. 36 The effective elastic con-
stants are retrieved by using the MATLAB nu-
merical solver (vpasolve) and the equations pro-
vided in the next chapters. Further details are
reported in SI.
3 Results and discussion
3.1 Device design
consider a gas-phase
We
synthesised Ag
nanoparticles (NPs) ultrathin film deposited
on polydimethylsiloxane (PDMS). It was re-
cently demonstrated that the morphology of
these films may be characterised by intercon-
nected, channel-like pores, 24 thus providing a
potentially wettable device. The choice of Ag,
although not a stringent one, 21 is based on
the availability of an experimentally validated
atomistic model for a virtual film reconstruc-
tion from gas-phase deposition parameters. 24
Moreover, for the Ag case and for film thick-
nesses in the tens of nm range, these films have
been experimentally proven as ultrafast photoa-
coustic transducers 37 with operating acoustic
frequencies spanning the range from tens to
hundreds of GHz. 37 A soft polymeric support,
such as PDMS, for the granular thin films yields
a high acoustic film-substrate mismatch. This
increases the film's breathing mode life time,
thus maximising its quality factor. This fact
will be further appreciated when discussing the
details of the sensing scheme. Additionally, the
choice of a soft polymeric support allows for a
stick-on/stick-off highly bio-compatible device.
The key parameter to tune the porosity of the
film is the Kinetic Energy per atom (KE) of the
NPs during the deposition process. Indeed, the
higher the KE, the higher the NPs deformation
upon landing and, consequently, the film's fill-
ing factor. Assuming an average NPs diameter
of 7 nm 24 1, we implemented four different MD
simulations, tuning the KE of the NPs-forming
atoms across the set of values {0.1, 0.2, 0.3, 0.4}
eV/at, the latter range being quite typical in
Supersonic Cluster Beam Deposition (SCBD).
In brief, each simulation reproduces the landing
of 90 NPs of ∼ 104 atoms on a 20x35nm2 base
1The actual NPs distribution is is peaked at two di-
ameters: 1.5 nm (small NP) and 7 nm (big NP). The
big NPs account for 96% of the total deposited mass.
The small NPs have been shown to be irrelevant in the
reconstruction of the nanogranular thin film scaffold, its
morphology, topography and mechanical properties be-
ing ruled by the big NPs only. We refer the reader to
Ref. 24 for further details on this point.
3
domain, for a total of ∼ 0.9 million atoms. In
gas-phase NP sources, such as a SCBD appara-
tus, the KE of the NPs can be tuned by varying
the source's geometrical parameters 38 and/or
carrier gas type and temperature. 39
The rendering of the virtual films, resulting
from MD simulations, is reported against KE
in Figure 1, left panel - top row. The virtual
films are composed of NPs (spherical agglomer-
ates in Figure 1 left panel - top row). Every NP
is assembled atom-by-atom (the atoms are the
smaller spheres visible upon adopting an high
magnification for figure inspection). The void
scaffolds are reported for increasing KE in Fig-
ure 1, left panel - bottom row, the void portion
being depicted in orange. The void scaffolds
are the complementary of the NPs film scaf-
folds. For each scaffold, the film filling factor,
FF = VN P /V where VN P is the total volume
of all the NPs and V is the overall film vol-
ume, 24 and the average pore size 35,36,40 are re-
ported against KE in Figure 1, right panel, re-
fer to SI for further details. Simulations results
show that, decreasing the KE from 0.4 to 0.1
eV/at, the filling factor decreases from 0.9 to
0.64 whereas the void morphology evolves from
sparse, mostly clogged pores to an open-pore
trabecular-like structure. The average pore size
increases from 2.8 to 4 nm and the average film
thickness h from 25 to 35 nm.
Among the simulated granular films, the best
geometrical features for device engineering are
achieved for a deposition KE of 0.1eV/at. In-
deed, the connected pores morphology, together
with the lower FF =0.64 and the biggest pores
of this film - 4 nm average pore diameter - yield
maximum sensitivity and higher storage capac-
ity while minimising the pinning-related issues,
which impede fluid infiltration inside the film.
Finally, a value of KE = 0.1 eV/at. should pre-
vent in-depth NPs implantation in PDMS, as
is expected for instance for KE=0.5 eV/at. 41
We will henceforth contextualise the discussion
focusing on the Ag granular ultrathin film of
thickness h=35 nm obtained with KE = 0.1
eV/at.
Figure 1: Left panel - top row: nanoporous scaffolds resulting from MD simulations on a 20×35 nm2
cell domain vs the kinetic energy per atom. Each small sphere corresponds to an atom, whereas the
bigger spherical agglomerates are the NPs, partially wrapped upon landing for increasing specific
kinetic energy. Left panel - bottom row: void scaffolds obtained from MD simulations vs the
kinetic energy per atom. The void portion is depicted in orange colour. The renderings are the
complementary of the top row ones. The latter renderings have been cut at a z-quota of 20 nm
for ease of visualisation. Right panel: filling factor (blue triangles, left blue axis) and average pore
dimension (red circles, right red axis) obtained from MD simulations as a function of the kinetic
energy per atom of the NPs at landing. The error bars of the filling factor are due to uncertainty
from MD simulations. The error bars of the average pore size are the standard deviations of the
punctual pore size distribution, see SI for further details.
3.2 Sensing scheme
The sensor working principle relies on ul-
trafast photoacousics, 42 among the emerging
techniques for mechanical nanometrology in
a variety of systems ranging from phononic
crystals, 43 -- 45 ultrathin-films, 46 -- 48 to multilayer
tube 49 and granular materials. 50 The transduc-
tion scheme is as follows:
in the excitation
step (opto-acoustic transduction) an ultrafast
IR laser pump pulse illuminates the device. The
energy absorbed by the granular metallic film
leads to an impulsive lattice temperature in-
crease of the order of a few Kelvins, avoiding
any annealing effect.2. The subsequent thermal
expansion excites the film's breathing modes
at their frequencies fn, n being the mode or-
der. The excited breathing modes decay with
time-constants τn, radiating acoustic waves into
the substrate.
In the detection step the ex-
cited breathing modes modulate the film's di-
electric constants (acousto-optic transduction).
2This value is obtained accounting for the typical
fluences and repetition rates in use in Fiber-Laser 37 or
Ti:sapphire oscillator based set-ups. 51
4
Figure 2: Fourier transform magnitude F (f )
of the photoacoustic signal expected for a dry
(red) and fully infiltrated (blue) granular thin
film sensor. The resonances are ascribed to the
fundamental, n=1, acoustic breathing modes.
Upon water infiltration the acoustic resonance
undergoes a frequency shift ∆f =fwet-fdry to-
gether with a linewidth reduction from Γdry to
Γwet. The two insets represent the dry and fully
infiltrated device respectively, together with a
schematics of the pump and probe technique.
0.1 eV/at.0.2 eV/at.0.3 eV/at.0.4 eV/at.1086420Average pore size (nm)0.40.30.20.1Kinetic energy (eV/at.)1.00.80.60.40.20.0Film filling factorPumpProbeDry sensorPDMSWet sensor1.00.80.60.40.2FT mag (a.u.)4035302520Frequency (GHz) Dry sensor Wet sensorΔf The acoustic oscillations are ultimately encoded
in the relative reflectivity/transmissivity varia-
tions, ∆I/I0, as measured by a time-delayed
probe pulse, the time-delay being with re-
spect to the instant of the pump-pulse arrival
t=0. 37 Here ∆I=I(t)-I0, where I(t) is the re-
flectivity/transmissivity at time-delay t and I0
the reflectivity/transmissivity of the unexcited
sample (static reflectivity/transmissivity). The
contribution of the breathing mode n to the
time-resolved optical trace, once thermal effects
are filtered out, is well mimicked by:
Fn(t) ≡ (∆I/I)n = Ane−t/τn cos(2πfn t+φn) θ(t)
(1)
where θ(t) is the Heaviside step function, τn,
fn, φn and An the breathing mode life-time,
frequency, phase and amplitude contribution to
the optical signal respectively. Calculating the
magnitude of the Fourier transform (FT) of the
time resolved trace given by Eq. 1, under the
assumption fn >> 1/τn, we obtain the follow-
ing resonance line shape:
Fn(f ) =
2(cid:112)1 + [2πτn(f − fn)]2
An τn
(2)
We refer the reader to SI for further details on
this point.
The quantities ruling the resonance, and of
relevance to the present discussion, are fn, τn
and their interplay synthesised by the quality
factor Qn. With this notion in mind, we now in-
spect the device acoustic properties within the
frame of the effective medium approximation
(EMA). The granular NPs thin film, weather
fluid-infiltrated or not, is modelled as an ef-
fective homogeneous and isotropic thin film of
effective density, ρ∗, and effective stiffness ten-
sor, C∗. The longitudinal sound velocity is
11 is the first element
of the film effective stiffness tensor. The acous-
11 ρ∗. The granu-
lar film adheres on a PDMS substrate of acous-
tic impedance ZS < Z. Within this frame, the
relevant breathing mode parameters read:
vN P =(cid:112)C∗
tic impedance reads Z =(cid:112)C∗
11/ρ∗, where C∗
(cid:12)(cid:12)(cid:12)(cid:12)f1ln
(cid:19)(cid:12)(cid:12)(cid:12)(cid:12)−1
(cid:18) Z − ZS
(cid:19)(cid:12)(cid:12)(cid:12)(cid:12)−1
(cid:12)(cid:12)(cid:12)(cid:12)ln
(cid:18) Z − ZS
Z + ZS
Z + ZS
(4)
n
(5)
τn =
Qn = πfnτn = π
We refer the reader to SI for further details on
the model and equations derivation.
A "dry" device is characterised by a resonance
centred at a frequency fdry and with decay-
time τdry (we dropped the mode index n for
brevity). Upon fluid infiltration in the granu-
lar film, both ρ∗ and C∗
11 increase, leading to
a resonance of frequency fwet and lifetime τwet.
For the sake of exemplification we here antic-
ipate results that will be derived further on.
Figure 2 reports the FT modulus of the fun-
damental breathing mode, n=1, expected for
the dry (red resonance) and for the fully water
infiltrated - fully wet - device (blue resonance).
Upon full water filling the resonance shifts by
an amount ∆f =fwet-fdry=1.61 GHz and the
√
linewidth decreases from Γdry= 2.83 GHz to
3/(πτn) is the
Γwet=2.68 GHz, where Γn =
resonance's FWHM. The frequency shift and
the decay time variations thus allow to quantify
the amount of infiltrated fluid. The resonance
frequency is first measured on the bare device
(dry configuration). The device is then loaded
(wet configuration) and its resonance frequency
measured. The resonance frequencies of the dry
and wet device are hence acquired in separate
measurement sessions.3. The minimum resolv-
able shift has thus to be considered as the er-
ror in the estimation of the peaks centers. The
higher the resonance Q-factor, the smaller is the
error in the peak center estimation, a high Q-
factor thus being a desirable feature.
3.3 The Practical Case
In general, the detection strategy is based on
resolving the resonance frequencies between the
wet and dry configurations, linking ∆f to the
fn =
vN P
2h
n = f1n
(3)
3This is at variance with respect to the problem of
separating two peaks from a signal which is the super-
position of them, such as resolving the two diffraction
peaks in a double slit optical diffraction experiment.
5
amount of filling fluid. We here illustrate the
strategy for the paradigmatic case of water ad-
sorption, the idea being alike for other fluids.
We chose water owing to its relevance in biol-
ogy and bio-related applications. We consider
the optimised device, i.e., the one obtained with
KE of 0.1eV/at. The interconnected porous
structure and pores size allow for a homoge-
neous water distribution in the whole accessi-
ble volume, thus justifying an EMA approach
(see SI for further information on this point);
other infiltration scenarios will be addressed
further on. The device is therefore considered
as an effective homogeneous and isotropic film
of effective density, ρ∗, effective shear, G∗, and
bulk, K∗, modulus (the effective stiffness ten-
sor C∗ being completely defined by G∗ and K∗
within the EMA). These quantities depend of
the amount of infiltrated water. The density
ρ∗ is obtained as a weighted mean of the silver
and water densities (ρAg and ρw respectively)
on the corresponding occupied volume in the
scaffold: ρ∗(l) = ρAg FF + ρw(1 − FF ) l. The
relative loading, l = Vw/Vvoid, is the infiltrated
water volume, Vw, normalized against the total
volume available for infiltration, Vvoid. G∗ and
K∗ are retrieved numerically solving Budiansky
equations: 52
K∗ − 1(cid:1) = 1
(cid:0) Ki
G∗ − 1(cid:1) = 1
(cid:0) Gi
(6)
N(cid:88)
N(cid:88)
i=1
ci
3K∗+4G∗
1 + 3K∗
ci
1 + 6(K∗+2G∗)
5(3K∗+4G∗)
i=1
where the index i runs over the N = 3 ma-
terials composing the effective film (i = Ag,
water, voids); Ki and Gi are the constituents
bulk and shear modulus; ci are the constituents
normalised concentrations: cAg = FF , cwater =
(1 − FF ) l and cvoids = (1 − FF ) (1 − l). We
stress that K∗ = K∗(l) and G∗ = G∗(l), are
both functions of the loading l. This depen-
dence has been omitted in Eq. 6 for ease of
visualisation. C∗
11 is a function of l through the
11 = K∗ + 4G∗/3. All the above-
equality C∗
mentioned materials parameters are reported in
SI. We are therefore in the position to calculate
fn, τn and Qn of the first two breathing modes,
6
Figure 3: Frequency of the n=1 (bottom panel)
and n=2 (top panel) acoustic breathing modes
(red squares,
left red axis) and decay times
(blue diamonds, right blue axis) vs water filling
within the homogeneous adsorption scenario.
Water filling is expressed both as relative vol-
umetric loading l (bottom axis) and equivalent
areal mass loading mS (top axis). Insets: qual-
ity factor Q vs relative volumetric loading l.
Cartoon: schematics of the infiltrated device
for l=0, 50% and 100%. Water is depicted in
blue and silver in black.
n=1 and n=2, as a function of the relative load-
ing l (see Equations 3, 4 and 5).
The results are reported in Figure 3. The
frequencies fn(l) are linear with l (red squares,
left vs bottom axis), so as the frequency shifts,
∆fn(l)=fn(l)-fn(0). The decay times τn (blue
diamonds) are mode-independent and, consis-
tently, Q2 = 2Q1 (see insets of Figure 3). The
high value of Q, as compared to the values re-
cently reported on similar systems, 37 Q ≈ 1−5,
stems from the tailored choice of the substrate
material. A soft substrate maximises the acous-
tic impedance mismatch, the device approach-
ing the ideal free-standing case.
We now focus on the device sensitivity issue.
Making the necessary changes from Ref. 53, we
31.531.030.530.029.5100806040200215210205200195201918QLoading (%)63.062.061.060.059.021521020520019504.59.013.417.922.4Areal mass loading (ng/mm2)403836Q100500Loading (%)Decay Time (ps)Frequency (GHz) Loading (%)n=2n=1100500 Table 1: Comparison of operation frequencies
and mass sensors sensitivities. NOT stands
for nanogranular optoacoustic transducer (this
work), QCM for quartz crystal microbalance
and FPW for flexural plate wave devices. The
values for typical QCM and FPW are taken
from ref. 54
Device type
NOT
QCM
FPW
f0 (GHz)
3 × 101
6.0 × 10−3
2.6 × 10−3
Sms
26 × 103
14
951
dl
1
vN P (l=0)
introduce Sl ≡ dfn(l)
1
fn(l) as a sound definition
to quantify the device's sensitivity to liquid in-
filtration. Provided the linearity of f (l) (see
Figure 3) and that ∆fn(l = 1) (cid:28) fn(l), the sen-
sitivity reads Sl ≈ ∆f (l=1)
1
f (0) = 0.05. Further-
more, the sensitivity Sl is thickness- and mode-
independent: from Eq. 3 Sl = vN P (l=1)−vN P (l=0)
where no dependence on h and n appears. On
the contrary Q is mode-dependent, as shown in
the insets of Figure 3. As previously mentioned,
a high Q is a desirable feature to minimise the
error in frequency shift read-out, thus privileg-
ing higher modes for sensing purposes.
For the sake of comparison against typical fig-
ures of merit, 44,53 we introduce the areal mass
loading mS = cwater h ρw = (1 − FF ) l h ρw.
The value mS quantifies the mass of infiltrated
water normalized on the device's unit area.
The standard sensitivity to mass-loading defi-
nition reads 53 Sms ≡ df (mS )
f (mS ). Holding the
same approximations discussed to evaluate Sl,
the mass sensitivity reads Sms ≈ ∂f
1
f (0) =
(1−F F )ρwh = 26 × 103 cm2/g. The latter fig-
Sl
ure can be further increased, since it scales as
1/h (Sl is h-independent). A minimum value of
h in excess of 14 nm is a realistic figure, grant-
ing a film with a fully developed granularity. 24
For the sake of comparison, we note that the
proposed device sensitivity outperforms, by
three orders of magnitude, that of commercially
available quartz crystal microbalances (QCM)
and, by an order of magnitude, that of flexural
plate wave (FPW) devices, see Table 1.
∂l
∂m
∂l
1
dmS
l
The frequency vs loading curves are well
7
within the detectability range of current ul-
trafast photo-acoustic technology.
In Figure
3 we calculated the device response discretis-
ing the liquid loading in steps of 5%, result-
ing in frequencies separated by ∼ 0.1 GHz for
the n=1 case (0.2 GHz for n=2). Neverthe-
less, the frequency resolution which may be
achieved with state-of-the art photo-acoustic
technology is way higher. For instance, sub-
harmonic resonant optical excitation of acoustic
modes in thin membranes allows resolving fre-
quency shifts with megahertz resolution. 55 The
minimum detectable infiltrated liquid variation
. Taking df ∼ 1 MHz, Sl ∼0.05
reads dl = df
and f (0) = fn=1(0) ∼29.5 GHz for mode n=1
f (0)
we find dl ∼7×10−4. This value can be fur-
ther decreased exploiting higher modes, in fact
dl scales as
nf0(0). This fact may be
readably appreciated comparing the slopes of
the frequency vs loading curves for the n = 1
and n = 2 reported in Figure 3.
fn(0) ∼ 1
1
Sl
1
As opposed to Inter Digital Transducer (IDT)
technology, within the present sensing scheme
acoustic wave generation, detection and sensing
of the infiltrated fluid take place in the same ac-
tive region, identified by the probe beam spot
size. This fact favours miniaturization and al-
lows working with minute quantities of total in-
filtrated fluid. For instance, assuming a typical
probe beam diameter of 10 µm a fully infiltrated
active area is tantamount to ∼1 femtoliter of
infiltrated fluid.
3.4 Discriminating among differ-
ent infiltration patterns
In the previous discussion we assumed a ho-
mogeneous infiltration scheme. We now argue
that the proposed sensing platform may be ex-
ploited to gain insight into the fluid infiltra-
tion pattern, a yet unsolved issue in nanoporous
materials and forming the object of extensive
research. 13,56,57 Let's analyse the layered infil-
tration pattern which comprises two scenarios.
The "water layer on top" (L-TOP) scenario
consists of a water infiltrated layer sitting on
top of an empty one (see inset at bottom-left
corner of Figure 4). This situation might be ex-
pected for high filling factors, such as the one
predicted for KE of 0.4eV/at.
In this situa-
tion some inner pores might be clogged and not
accessible, as suggested by Bisio et al. 14 The
opposite scenario, "water layer on bottom" (L-
BOT), consists in a water infiltrated layer sit-
ting on the polymeric substrate and covered by
an empty layer on top (see inset in top-left cor-
ner of Figure 4). This scenario might arise, for
instance, when water starts evaporating from
a fully infiltrated scaffold. In this context one
may for instance follow the evaporation process
as it takes place, tracking in time the loading
curve evolution. We pinpoint that, for the case
of a fully infiltrated device, the L-TOP and L-
BOT scenario coincide.
The stratified scenario is conveniently mod-
elled as two effective media in series de-
posited on a semi-infinite substrate, the effec-
tive media mimicking the fully wet (fully wa-
ter infiltrated-Ag scaffold: Ag and water filled
voids) and dry layers (non-infiltrated Ag scaf-
fold: Ag and empty voids). For the layered-
case the breathing mode frequency and decay
time read fn=Re{ωn}/2π and τn=1/Im{ωn}
respectively, where the complex-valued ωn is
the nth root of the following equation: 37
−2iω ht
−2iω( ht
vt
+
hb
vb
)−rt,b rb,se
−2iω
hb
vb −1 = 0.
rt,b e
vt +rb,se
(7)
The subscripts t, b and s stand for top layer,
bottom layer and substrate respectively; rm,n =
(Zm − Zn)/(Zm + Zn) is the acoustic reflection
amplitude between the m and n-indexed layers
and Zn, vn and hn the n-indexed layer acous-
tic impedance, longitudinal sound velocity and
thickness respectively. C∗
11 for the fully dry
and wet layers (corresponding to the homoge-
neous wetting case of l=0 and l=100% respec-
tively) are calculated from Equation 6, whereas
ρ∗ equals ρAg and ρAg FF + ρw(1 − FF ) for the
dry and wet layers respectively. These quan-
tities allow calculating Zb, Zt and vb, vt. The
total film thickness is h = ht + hb = 35 nm.
The relative water loading reads l=ht/h for the
L-TOP case (l=hb/h for L-BOT) . For sake of
comparison, we analyse the acoustic response
keeping the same scaffold as in the homoge-
neous wetting case, i.e. the one obtained for KE
8
= 0.1 eV/at. The breathing mode frequency,
f1, and decay time, τ1, for the n=1 mode, ob-
tained upon numerical solution of Equation 7,
are reported in Figure 4. As opposed to the
homogeneous wetting case, both f1 and τ1 are
non linear vs l.
The L-TOP and L-BOT are characterised by
the same f1 vs loading curve (continous and
dashed red lines respectively in Figure 4). This
is due to the fact that, with regards to f , the
boundary condition at the nanoporous film-
PDMS interface is substantially stress free.
With respect to f the device behaves as a free-
standing layered membrane, resulting in a sym-
metric situation between the L-TOP and L-
BOT scenarios.
Conversely, the τ1 vs loading curves (continous
and dashed blue lines in Figure 4) show op-
posite trends in the L-TOP and L-BOT sce-
narios.
In particular, whatever the quantity
of adsorbed water, τ1(l) is smallest for the L-
TOP scenario, is biggest for the L-BOT sce-
nario and sits between the two for the homo-
geneous filling case. The physical explanation
has to do with the acoustic impedance jumps
across the device thickness. The smoother are
the impedance changes across the device, the
highest is the acoustic wave transmission to the
PDMS substrate, hence the lower is the breath-
ing mode damping time.
In the L-TOP sce-
nario the acoustic impedance decreases across
the sample from the device free-surface all the
way into the PDMS layer: Zt > Zb > Zs.
This scenario maximises acoustic transmission
from the device to the substrate, hence lead-
ing to the lowest decay times. The L-BOT sce-
nario is characterised by the greatest acoustic
impedance jumps across the sample: Zt < Zb
and Zb > Zs, hence leading to the longest de-
cay times. The homogeneous filling case lays
in between: ZHomo > Zs where ZHomo, the
nanoporous film acoustic impedance for the ho-
mogeneous water infiltration case, ranges be-
tween the fully wet and dry device, correspond-
ing to l=0 and l=100% respectively. The cases
for n=2 and n=3 are discussed in SI.
The f (l) curves thus allow discriminating be-
tween homogeneous and layered infiltration sce-
narios. As for the latter, the L-TOP and L-
Figure 4: Frequency of the n=1 acoustic breathing modes (left axis, red) and decay times (right axis,
blue) vs water filling within the layered adsorption scenarios L-TOP (full lines), L-BOT (dashed
lines) and, for sake of comparison, for the homogeneous wetting case (markers). Water filling is
expressed both as relative volumetric loading l (bottom axis) and equivalent areal mass loading
mS (top axis). Cartoon: schematics of the infiltrated device for the L-TOP (bottom cartoon) and
L-BOT (top cartoon) scenarios. Water is depicted in blue and silver in black.
BOT cases may be differentiated on the basis
of the τ (l) (or Q(l)) curves. This proves the
potential of the present strategy in uncovering
the fluid infiltration pattern in nanoporous ma-
terials.
4 Conclusions
We designed a novel sensing platform allow-
ing to quantify the amount of fluid infiltrated
in a nanoporous coating. The platform is
based on a gas-phase synthesised nanogranu-
lar metallic coating with open-porosity, specifi-
cally engineered via molecular-dynamics for ef-
ficient ultrafast photoacoustic detection of the
filling fluid.
For the paradigmatic case of
water filling we predict a sensitivity exceed-
ing 26×103 cm2/g, upon equivalent areal mass
few ng/mm2, outperforming cur-
loading of
rent microbalance-based technology by three
orders of magnitude. The predictions are ro-
bust, the theoretical frame having been recently
benchmarked against experiments, 24 and ultra-
fatst photoacoustic read-out of mechanical reso-
nances in metallic nanoporous coatings demon-
strated. 37
The nanogranular metallic scaffold is readily
exploitable as a distributed nanogetter coating,
integrating fluid sensing capabilities and serv-
ing as a stick-on/stick-off highly bio-compatible
device. The film may be deposited on virtu-
ally any surface while varying the metal com-
position, 27,28 allowing to foresee integration of
the fluid sensing capabilities with a variety of
applications. 23,24,26 The present strategy, rely-
ing on the variation of a specific acoustic res-
onance upon fluid infiltration, provides data
that are intrinsically simpler to interpret as
compared to existing techniques such as EE
and gas adsorption-based porosimetry.
For
this reason it might serve as a valid alterna-
tive/complement to current technologies.
Finally, we showed that the proposed sens-
ing scheme allows discriminating among differ-
9
230220210200190180170160Decay Time (ps)100806040200Loading (%)31.531.030.530.029.529.028.528.0Frequency (GHz) 04.517.922.4Areal mass loading (ng/mm2) 9.0 13.4/ Homo/L-TOP/L-BOTn=1 ent filling patterns, providing a means to in-
vestigate pinning-related issues of relevance for
nanoporous membrane wettability.
The present scheme is general and eas-
ily implementable.
It may be expanded to
other emerging gas-phase synthesized gran-
ular materials 29 and extended to materials
with mesoscale porosity. 30 -- 32 In perspective, the
present scheme could be expanded to include
granular multi-layers synthesized with differ-
ent metals. This would allow expanding the
range of exploitable acoustic parameters so as
to fine-tune the loading curve when choosing
other substrate materials, for instance by engi-
neering an acoustic-impedance graded device.
Furthermore, this strategy would allow to ex-
ploit a Ti granular film as an adhesion layer to
increase the overall device sticking-factor when
working with rigid substrates. The sensing ap-
paratus may be implemented taking advantage
of readily available compact, table-top sources
relying on superior sampling speed and telecom
technology - such as the ASOPS technique 46,58
- and EUV coherent sources granting superior
photoacoustic sensitivity. 59,60
5 Supporting Information
Device design details (MD details, Filling fac-
tor, Volume and thickness, pore size), Sensing
scheme (Derivation of Eq.2 and its applicabil-
ity, Details of the model: derivation of Eq.s 3
and 4, A possible mechanism to convey water
to the device), The Practical Case (Materials
constant, Layered cases for n≤3).
6 Acknowledgements
G.B. acknowledges financial support from the
Research Foundation Flanders (FWO). C.C.
and F.B. acknowledge financial support from
the MIUR Futuro in ricerca 2013 Grant in the
frame of the ULTRANANO Project (project
number: RBFR13NEA4).
F.B. and C.G.
acknowledge support
from Universit`a Cat-
tolica del Sacro Cuore through D.2.2 and
D.3.1 grants. F.B. acknowledge financial sup-
port from Fondazione E.U.L.O. C.C. acknowl-
edges financial support from Programma Op-
erativo Nazionale 2007-2013 "Ricerca e com-
petitivit`a" financed by EU through project
PON04a2 00490 "Ricerca Applicata a Reti
di comunicazione M2M e modem integrati
innovativi dedicati a servizi avanzati per le
Smart Cities - M2M Netergit", and compu-
tational support from PRACE for awarding
access to Marconi hosted at CINECA, Italy,
through project UNWRAP (call 14), and IS-
CRA through project UNWRAPIT.
References
(1) Bocquet, L.; Charlaix, E. Nanofluidics,
From Bulk to Interfaces. Chem. Soc. Rev.
2010, 39, 1073 -- 1095.
(2) Schoch, R. B.; Han, J.; Renaud, P. Trans-
port Phenomena in Nanofluidics. Rev.
Mod. Phys. 2008, 80, 839 -- 883.
(3) Zhang, H.; Tian, Y.; Jiang, L. Fundamen-
tal Studies and Practical Applications of
Bio-Inspired Smart Solid-State Nanopores
and Nanochannels. Nano Today 2016, 11,
61 -- 81.
(4) van den Berg, A.; Wessling, M. Nanoflu-
idics: Silicon for the Perfect Membrane.
Nature 2007, 445, 726.
(5) Adiga, S. P.; Jin, C.; Curtiss, L. A.;
Monteiro-Riviere, N. A.; Narayan, R. J.
Nanoporous Membranes for Medical and
Biological Applications. Wiley Interdiscip.
Rev. Nanomed. Nanobiotechnol. 2009, 1,
568 -- 581.
(6) Kholmanov, I. N.; Stoller, M. D.; Edge-
worth, J.; Lee, W. H.; Li, H.; Lee, J.;
Barnhart, C.; Potts, J. R.; Piner, R.;
Akinwande, D.; Others, Nanostructured
Hybrid Transparent Conductive Films
With Antibacterial Properties. ACS Nano
2012, 6, 5157 -- 5163.
(7) Kumar, K. V.; Preuss, K.; Titirici, M.-M.;
Rodriguez-Reinoso, F. Nanoporous Mate-
rials for the Onboard Storage of Natural
10
Gas. Chem. Rev. 2017, 117, 1796 -- 1825,
PMID: 28094515.
Theory. J. Am. Chem. Soc. 1952, 74,
1367 -- 1371.
(8) Morris, R. E.; Wheatley, P. S. Gas Storage
In Nanoporous Materials. Angew. Chem.
Int. Ed. 2008, 47, 4966 -- 4981.
(16) Foo, K.; Hameed, B. Insights Into the
Modeling of Adsorption Isotherm Sys-
tems. Chem. Eng. J. 2010, 156, 2 -- 10.
(9) Bau, J. A.; Takanabe, K. Ultrathin
Microporous SiO2 Membranes Photode-
posited on Hydrogen Evolving Catalysts
Enabling Overall Water Splitting. ACS
Catal. 2017, 7, 7931 -- 7940.
(10) Ceratti, D. R.; Faustini, M.; Sinturel, C.;
Vayer, M.; Dahirel, V.;
Jardat, M.;
Grosso, D. Critical Effect of Pore Charac-
teristics on Capillary Infiltration in Meso-
porous Films. Nanoscale 2015, 7, 5371 --
5382.
(11) Chen, C.; Xu, X.; Li, Y.; Jans, H.;
Neutens, P.; Kerman, S.; Vereecke, G.;
Holsteyns, F.; Maes, G.; Lagae, L.; Stak-
enborg, T.; van Dorpe, P. Full Wetting
of Plasmonic Nanopores Through Two-
Component Droplets. Chem. Sci. 2015, 6,
6564 -- 6571.
(12) Mercuri, M.; Pierpauli, K. A.; Berli, C. L.;
Bellino, M. G. An Open Pit Nanofluidic
Tool: Localized Chemistry Assisted by
Mesoporous Thin Film Infiltration. ACS
Appl. Mater. Interfaces 2017, 9, 16679 --
16684, PMID: 28463480.
(13) Boissiere, C.; Grosso, D.; Lepoutre, S.;
Nicole, L.; Bruneau, A. B.; Sanchez, C.
Porosity and Mechanical Properties of
Mesoporous Thin Films Assessed by En-
vironmental Ellipsometric Porosimetry.
Langmuir 2005, 21, 12362 -- 12371.
(17) Toccafondi, C.; Uttiya, S.; Caval-
leri, O.; Gemme, G.; Barborini, E.; Bi-
sio, F.; Canepa, M. Optical Properties
of Nanogranular and Highly Porous TiO2
Thin Films. J. Phys. D: Appl. Phys. 2014,
47, 485301.
(18) Cavaliere, E.; Benetti, G.; Van Bael, M. J.;
Winckelmans, N.; Bals, S.; Gavioli, L.
Exploring the Optical and Morphological
Properties of Ag and Ag/TiO2 Nanocom-
posites Grown by Supersonic Cluster
Beam Deposition. Nanomat. 2017, 7, 442.
(19) Jungk, G. Possibilities and Limitations of
Ellipsometry. Thin Solid Films 1993, 234,
428 -- 431.
(20) Bisio, F.; Palombo, M.; Prato, M.; Cav-
alleri, O.; Barborini, E.; Vinati, S.;
Franchi, M.; Mattera, L.; Canepa, M.
Optical Properties of Cluster-Assembled
Nanoporous Gold Films. Phys. Rev. B
2009, 80, 205428.
(21) Bisio, F.; Prato, M.; Barborini, E.;
Canepa, M. Interaction of Alkanethiols
With Nanoporous Cluster-Assembled Au
Films. Langmuir 2011, 27, 8371 -- 8376.
(22) Quirk, J. P. Significance Of Surface Areas
Calculated From Water Vapor Sorption
Isotherms By Use Of The B.E.T. Equa-
tion. Soil Sci. 1955, 80, 423 -- 430.
(14) Bisio, F.; Prato, M.; Cavalleri, O.; Bar-
borini, E.; Mattera, L.; Canepa, M. Inter-
action of Liquids with Nanoporous Clus-
ter Assembled Au Films. J. Phys. Chem.
C 2010, 114, 17591 -- 17596.
(23) Bettini, L. G.; Bellacicca, A.; Piseri, P.;
Milani, P. Supersonic Cluster Beam Print-
ing of Carbon Microsupercapacitors on
Paper. Flexible Printed Electron. 2017, 2,
025002.
(15) Mooney, R. W.; Keenan, A. G.;
Wood, L. A. Adsorption of Water
Vapor by Montmorillonite.
I. Heat of
Desorption and Application of BET
(24) Benetti, G.; Caddeo, C.; Melis, C.; Fer-
rini, G.; Giannetti, C.; Winckelmans, N.;
Bals, S.; Van Bael, M. J.; Cavaliere, E.;
11
Gavioli, L.; Banfi, F. Bottom-Up Me-
chanical Nanometrology of Granular Ag
Nanoparticles Thin Films. J. Phys. Chem.
C 2017, 121, 22434 -- 22441.
(25) Galvanetto, N.; Perissinotto, A.; Pe-
droni, A.; Torre, V. Fodis: Software for
Protein Unfolding Analysis. Biophys. J.
2018, 114, 1264 -- 1266.
(26) Borghi, F.; Podest´a, A.; Piazzoni, C.; Mi-
lani, P. Growth Mechanism of Cluster-
Assembled Surfaces: From Submonolayer
to Thin-Film Regime. Phys. Rev. Appl
2018, 9, 044016.
(27) Benetti, G.; Cavaliere, E.; Canteri, A.;
Landini, G.; Rossolini, G. M.; Pallec-
chi, L.; Chiodi, M.; Van Bael, M. J.;
Winckelmans, N.; Bals, S.; Gavioli, L. Di-
rect Synthesis of Antimicrobial Coatings
Based on Tailored Bi-Elemental Nanopar-
ticles. APL Mat. 2017, 5, 036105.
(28) Corbelli, G.; Ghisleri, C.; Marelli, M.; Mi-
lani, P.; Ravagnan, L. Highly Deformable
Nanostructured Elastomeric Electrodes
With
Improving Conductivity Upon
Cyclical Stretching. Adv. Mat. 2011, 23,
4504 -- 4508.
(29) Nasiri, N.; Elmøe, T. D.; Liu, Y.;
Qin, Q. H.; Tricoli, A. Self-Assembly
Dynamics and Accumulation Mechanisms
of Ultra-Fine Nanoparticles. Nanoscale
2015, 7, 9859 -- 9867.
(30) Voti, R. L.; Leahu, G.; Larciprete, M.;
Sibilia, C.; Bertolotti, M.; Nefedov, I.;
Anoshkin,
I. Photoacoustic Character-
ization of Randomly Oriented Silver
Nanowire Films.
Int. J. Thermophys.
2015, 36, 1342 -- 1348.
(31) de Castro, L. D. C.; Larocca, N. M.; Pes-
san, L. A. Towards the Development of
Superhydrophilic SiO2-Based Nanoporous
Coatings: Concentration and Particle Size
Effect. Thin Solid Films 2018, 651, 138 --
144.
(32) Lamastra, F.; Grilli, M.; Leahu, G.; Be-
lardini, A.; Voti, R. L.; Sibilia, C.; Salva-
tori, D.; Cacciotti, I.; Nanni, F. Diatom
Frustules Decorated With Zinc Oxide
Nanoparticles for Enhanced Optical Prop-
erties. Nanotechnology 2017, 28, 375704.
(33) Plimpton, S. Fast Parallel Algorithms
for Short-Range Molecular Dynamics. J.
Comput. Phys. 1995, 117, 1 -- 19.
(34) Heinz, H.; Vaia, R. A.; Farmer, B. L.;
Naik, R. R. Accurate Simulation of Sur-
faces and Interfaces of Face-Centered Cu-
bic Metals Using 12-6 and 9-6 Lennard-
Jones Potentials. J. Phys. Chem. C 2008,
112, 17281 -- 17290.
I.;
Jackson,
Cordeli`eres,
(35) Doube, M.; K(cid:32)losowski, M. M.; Arganda-
F.
P.;
Carreras,
Dougherty, R. P.;
J. S.;
Schmid, B.; Hutchinson, J. R.; She-
felbine, S. J. BoneJ: Free and extensible
bone image analysis in ImageJ. Bone
2010, 47, 1076 -- 1079.
(36) Saito, T.; Toriwaki, J.-I. New Algorithms
for Euclidean Distance Transformation of
an N-Dimensional Digitized Picture With
Applications. Pattern Recognit. 1994, 27,
1551 -- 1565.
(37) Peli, S.; Cavaliere, E.; Benetti, G.; Gan-
dolfi, M.; Chiodi, M.; Cancellieri, C.;
Giannetti, C.; Ferrini, G.; Gavioli, L.;
Banfi, F. Mechanical Properties of Ag
Nanoparticle Thin Films Synthesized by
Supersonic Cluster Beam Deposition. J.
Phys. Chem. C 2016, 120, 4673 -- 4681.
(38) Vahedi Tafreshi, H.; Piseri, P.; Bar-
borini, E.; Benedek, G.; Milani, P. Sim-
ulation on the Effect of Brownian Motion
on Nanoparticle Trajectories in a Pulsed
Microplasma Cluster Source. J. Nanopart.
Res. 2002, 4, 511 -- 524.
(39) Mazza, T.; Devetta, M.; Milani, P.; Bon-
giorno, G.; Coreno, M.; Piseri, P. Access-
ing the Fractal Dimension of Free Clusters
in Supersonic Beams. New J. Phys. 2011,
13, 023009.
12
for
(40) Hildebrand, T.; Ruegsegger, P. A New
the Model-Independent
Method
in Three-
Assessment
Dimensional Images. J. Microsc. 1997,
185, 67 -- 75.
of Thickness
(41) Ravagnan, L.; Divitini, G.; Rebasti, S.;
Marelli, M.; Piseri, P.; Milani, P. Poly
(Methyl Methacrylate) -- Palladium Clus-
ters Nanocomposite Formation by Super-
sonic Cluster Beam Deposition: a Method
for Microstructured Metallization of Poly-
mer Surfaces. J. Phys. D: Appl. Phys.
2009, 42, 082002.
(42) Matsuda, O.;
Larciprete, M. C.;
Voti, R. L.; Wright, O. B. Funda-
mentals of Picosecond Laser Ultrasonics.
Ultrasonics 2015, 56, 3 -- 20.
(43) Travagliati, M.; Nardi, D.; Giannetti, C.;
Gusev, V.; Pingue, P.; Piazza, V.;
Ferrini, G.; Banfi, F. Interface Nano-
Confined Acoustic Waves in Polymeric
Surface Phononic Crystals. Appl. Phys.
Lett. 2015, 106, 021906.
(44) Nardi, D.; Zagato, E.; Ferrini, G.; Gian-
netti, C.; Banfi, F. Design of a Surface
Acoustic Wave Mass Sensor in the 100
GHz Range. Appl. Phys. Lett. 2012, 100,
253106.
(45) Mante, P.-A.; Chen, C.-C.; Wen, Y.-C.;
Chen, H.-Y.; Yang, S.-C.; Huang, Y.-R.;
Ju Chen, I.; Chen, Y.-W.; Gusev, V.;
Chen, M.-J.; Kuo, J.-L.; Sheu, J.-K.;
Sun, C.-K. Probing Hydrophilic Interface
of Solid/Liquid-Water by Nanoultrason-
ics. Sci. Rep. 2014, 4, 6249.
(46) Nardi, D.;
Travagliati, M.; Mur-
nane, M. M.; Kapteyn, H. C.; Ferrini, G.;
Impulsively
Giannetti, C.; Banfi, F.
Excited Surface Phononic Crystals:
a
Route Toward Novel Sensing Schemes.
IEEE Sens. J. 2015, 15, 5142 -- 5150.
(47) Hoogeboom-Pot, K. M.; Turgut, E.;
Hernandez-Charpak, J. N.; Shaw, J. M.;
Kapteyn, H. C.; Murnane, M. M.;
13
Nardi, D. Nondestructive Measurement of
the Evolution of Layer-Specific Mechani-
cal Properties in Sub-10 nm Bilayer Films.
Nano Lett. 2016, 16, 4773 -- 4778.
(48) Grossmann, M.; Schubert, M.; He, C.;
Brick, D.; Scheer, E.; Hettich, M.; Gu-
sev, V.; Dekorsy, T. Characterization of
Thin-Film Adhesion and Phonon Life-
times in Al/Si Membranes by Picosec-
ond Ultrasonics. New J. Phys. 2017, 19,
053019.
(49) Brick, D.; Engemaier, V.; Guo, Y.;
Grossmann, M.; Li, G.; Grimm, D.;
Schmidt, O. G.; Schubert, M.; Gu-
sev, V. E.; Hettich, M.; Dekorsy, T. Inter-
face Adhesion and Structural Character-
ization of Rolled-up GaAs/In 0.2 Ga 0.8
As Multilayer Tubes by Coherent Phonon
Spectroscopy. Sci. Rep. 2017, 7, 5385.
(50) Avice, J.; Boscher, C.; Vaudel, G.; Bro-
tons, G.; Juv, V.; Edely, M.; Mthivier, C.;
Gusev, V. E.; Belleville, P.; Piombini, H.;
Ruello, P. Controlling the Nanocontact
Nature and the Mechanical Properties of
a Silica Nanoparticle Assembly. J. Phys.
Chem. C 2017, 121, 23769 -- 23776.
(51) Caddeo, C.; Melis, C.; Ronchi, A.; Gi-
annetti, C.; Ferrini, G.; Rurali, R.;
Colombo, L.; Banfi, F. Thermal Boundary
Resistance From Transient Nanocalorime-
try: A Multiscale Modeling Approach.
Phys. Rev. B 2017, 95, 085306.
(52) Budiansky, B. On the Elastic Moduli of
Some Heterogeneous Materials. J. Mech.
Phys. Solids 1965, 13, 223 -- 227.
(53) Vellekoop, M. J. Acoustic Wave Sensors
and Their Technology. Ultrasonics 1998,
36, 7 -- 14.
(54) Cheeke, J.; Wang, Z. Acoustic Wave Gas
Sensors. Sens. Actuators, B 1999, 59,
146 -- 153.
(55) Bruchhausen, A.; Gebs, R.; Hudert, F.;
Issenmann, D.; Klatt, G.; Bartels, A.;
Schecker, O.; Waitz, R.; Erbe, A.;
Scheer, E.; Huntzinger, J.-R.; Mlayah, A.;
Dekorsy, T. Subharmonic Resonant Op-
tical Excitation of Confined Acoustic
Modes in a Free-Standing Semiconduc-
tor Membrane at GHz Frequencies with a
High-Repetition-Rate Femtosecond Laser.
Phys. Rev. Lett. 2011, 106, 077401.
(56) Huang, P.; Chang, W.-S.; Ansell, M. P.;
John, C. Y. M.; Shea, A. Porosity es-
timation of Phyllostachys edulis (Moso
bamboo) by computed tomography and
backscattered electron imaging. Wood Sci.
Technol. 2017, 51, 11 -- 27.
(57) Zhang, P.; Li, J.; Lu, S.; Xue, H.;
Zhang, J.; Yang, J. A Precise Porosity
Measurement Method for Oil-Bearing Mi-
cro/Nano Porous Shales Using Low-Field
Nuclear Magnetic Resonance (LF-NMR).
J. Nanosci. Nanotechnol. 2017, 17, 6827 --
6835.
(58) Schubert, M.; Grossmann, M.; Ristow, O.;
Hettich, M.; Bruchhausen, A.; Bar-
retto, E. C.; Scheer, E.; Gusev, V.; Deko-
rsy, T. Spatial-Temporally Resolved High-
Frequency Surface Acoustic Waves on Sil-
icon Investigated by Femtosecond Spec-
troscopy. Appl. Phys. Lett. 2012, 101,
013108.
(59) Nardi, D.; Hoogeboom-Pot, K. M.;
Hernandez-Charpak, J. N.; Tripp, M.;
King, S. W.; Anderson, E. H.; Mur-
nane, M. M.; Kapteyn, H. C. Probing Lim-
its of Acoustic Nanometrology Using Co-
herent Extreme Ultraviolet Light. Metrol-
ogy, Inspection, and Process Control for
Microlithography XXVII. 2013; p 86810N.
(60) Nardi, D.; Travagliati, M.; Siemens, M. E.;
Li, Q.; Murnane, M. M.; Kapteyn, H. C.;
Ferrini, G.; Parmigiani, F.; Banfi, F. Prob-
ing Thermomechanics at the Nanoscale:
Impulsively Excited Pseudosurface Acous-
tic Waves in Hypersonic Phononic Crys-
tals. Nano Lett. 2011, 11, 4126 -- 4133.
14
|
1909.06268 | 1 | 1909 | 2019-09-13T14:48:32 | Asymptotically exact strain-gradient models for nonlinear slender elastic structures: a systematic derivation method | [
"physics.app-ph"
] | We propose a general method for deriving one-dimensional models for nonlinear structures. It captures the contribution to the strain energy arising not only from the macroscopic elastic strain as in classical structural models, but also from the strain gradient. As an illustration, we derive one-dimensional strain-gradient models for a hyper-elastic cylinder that necks, an axisymmetric membrane that produces bulges, and a two-dimensional block of elastic material subject to bending and stretching. The method offers three key advantages. First, it is nonlinear and accounts for large deformations of the cross-section, which makes it well suited for the analysis of localization in slender structures. Second, it does not require any a priori assumption on the form of the elastic solution in the cross-section, i.e., it is Ansatz-free. Thirdly, it produces one-dimensional models that are asymptotically exact when the macroscopic strain varies on a much larger length scale than the cross-section diameter. | physics.app-ph | physics | Asymptotically exact strain-gradient models
for nonlinear slender elastic structures:
a systematic derivation method
Claire Lestringanta, Basile Audolyb
aMechanics & Materials, Department of Mechanical and Process Engineering, ETH Zurich, 8092 Zurich, Switzerland
bLaboratoire de m´ecanique des solides, CNRS, Institut Polytechnique de Paris, Palaiseau, France
Abstract
We propose a general method for deriving one-dimensional models for nonlinear structures.
It captures
the contribution to the strain energy arising not only from the macroscopic elastic strain as in classical
structural models, but also from the strain gradient. As an illustration, we derive one-dimensional strain-
gradient models for a hyper-elastic cylinder that necks, an axisymmetric membrane that produces bulges,
and a two-dimensional block of elastic material subject to bending and stretching. The method offers three
key advantages. First, it is nonlinear and accounts for large deformations of the cross-section, which makes
it well suited for the analysis of localization in slender structures. Second, it does not require any a priori
assumption on the form of the elastic solution in the cross-section, i.e., it is Ansatz-free. Thirdly, it produces
one-dimensional models that are asymptotically exact when the macroscopic strain varies on a much larger
length scale than the cross-section diameter.
Keywords: A. Localization B. elastic material, B. finite strain, C. asymptotic analysis, C. energy methods
1. Introduction
There exists a variety of models for slender structures, going much beyond the traditional models for the
stretching of bars and the bending of beams. The applicability of classical models being limited to materials
having linear, homogeneous and isotropic elastic properties, a number of extensions have been considered
to account for different elastic behaviors such as hyperelastic materials (Cimeti`ere et al., 1988) or more
specifically nematic elastomers (Agostiniani et al., 2016), for inhomogeneous elastic properties in the cross-
section, for the presence of natural curvature or twist (Freddi et al., 2016) or more generally for the existence
of inhomogeneous pre-stress in the cross-section (Lestringant and Audoly, 2017). As the classical rod models
are inapplicable if the cross-section itself is a slender 2d domain, specific models have been derived, e.g.,
to address inextensible ribbons (Sadowsky, 1930; Wunderlich, 1962), as well as thin walled beams having
a flat (Freddi et al., 2004) or curved (Hamdouni and Millet, 2006) cross-section. The classical models are
inapplicable as well in the presence of a large contrast of elastic moduli within the cross-sections, as happens
for sandwiched beams:
in this case, the presence of shear is often accounted for using the Timoshenko
beam model. Specific models are also required to account for physical effects such as the interaction with a
magnetic field (Geymonat et al., 2018) or surface tension arising in soft beams immersed in a fluid (Xuan
and Biggins, 2017).
One can easily get lost in view of not only the multiplicity of these models but also their justification
(or lack thereof). Rigorous justifications based on asymptotic expansions have made use of restrictive
assumptions: the work in this direction was initiated in the context of linear elasticity (Bermudez and
Viano, 1984; Sanchez-Hubert and Sanchez Palencia, 1999), and extended to finite elasticity under specific
assumptions regarding material symmetries (Cimeti`ere et al., 1988). The different models, such as Navier-
Bernoulli beams, Timoshenko beams, Vlasov beams, inextensible ribbons, etc., are justified by different
arguments each, and a unified justification method is lacking. There are many phenomena in slender
Preprint submitted to J. Mech. Phys. Solids
September 16, 2019
9
1
0
2
p
e
S
3
1
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
1
v
8
6
2
6
0
.
9
0
9
1
:
v
i
X
r
a
structures for which no asymptotically justified 1d model is available, such as the ovalization of tubes
subjected to bending (Calladine, 1983) or pinching (Mahadevan et al., 2007) and the propagative instabilities
in shallow panels (Kyriakides and Chang, 1991).
In some work, one-dimensional (1d) models have been proposed based on kinematic hypotheses. This is
the case, for instance, for the analysis localization in hyperelastic cylinders (Coleman and Newman, 1988)
and tape springs (Picault et al., 2016). Even when these kinematic hypotheses turn out to be valid, their
domain of validity is typically limited and dependent, in a hidden way, on the simplifying assumptions of
the model. For instance, the most common assumptions used to derive the classical theory of beams is that
cross-sections remain planar and perpendicular to the center line, and that the shear in the plane of the
cross-sections is zero. These assumptions are incorrect unless specific material symmetries are applicable,
which is ill-appreciated. Moreover, they cannot be used to derive higher-order models, as discussed by
Audoly and Hutchinson (2016).
This paper proposes a systematic and rigorous dimension reduction method for obtaining 1d models for
nonlinear slender elastic structures, which works under broad assumptions. The main features of the method
are as follows. The reduction method can start from a variety of models, such as a hyperelastic model for
cylinder (e.g., for the stretching of bars), a nonlinear model for a thin membrane (e.g., for the analysis of
bulges in axisymmetric balloons) or a shell model (e.g., for the tape spring problem). It can handle arbitrary
elastic constitutive laws (including nonlinear and anisotropic ones), arbitrary pre-stress distributions in the
cross-section, and inhomogeneous material properties in the cross-section. The mechanical and geometrical
properties of the structure are assumed to be invariant in the longitudinal direction; the extension to slowly
variable properties is straightforward, as discussed in §6. Nonlinearity of both the elastic and geometric
types are permitted, and large spatial variations in the deformed shape of the cross-sections are accounted
for. No a priori kinematic hypothesis is made, the microscopic displacement being found by solving the
equations of elasticity. Our reduction method is built on a two-scale expansion, assuming slow variations
in the longitudinal directions. As such, it is asymptotically exact. Its justification is based on a formal
expansion, not on a rigorous proof. We hope that our formal argument can be turned into a rigorous one in
the future.
An important asset of the method is that it captures the gradient effect, i.e., the dependence of the strain
energy on the gradient of strain and not just on the strain. This makes it possible to derive higher-order
reduced models offering the following advantages: (i ) they feature faster convergence towards the solution
of the full (non-reduced) problem and (ii ) they are well-suited to the analysis of localization in slender
structures. Localization is ubiquitous in slender structures, from neck formation in polymer bars under
traction (G'Sell et al., 1983), to beading in cylinders made up of soft gels (Matsuo and Tanaka, 1992; Mora
et al., 2010), to bulges produced by the inflation of cylindrical party balloons (Kyriakides and Chang, 1990),
and to kinks in bent tape springs (Seffen and Pellegrino, 1999). Classical reduced models depending on strain
only cannot resolve the sharp interfaces that result from localization, and are mathematically ill-posed. By
contrast, higher-order models capturing the dependence on the gradient of strain allows the interfaces to be
resolved and are well-posed in the context of localization. In prior work, asymptotic 1d strain-gradient models
have been obtained as refinements over the standard theory for linearly elastic beams (Trabucho and Viano,
1996; Buannic and Cartaud, 2000), inextensible ribbons (Sadowsky, 1930; Wunderlich, 1962) and thin-walled
beams (Freddi et al., 2004). The possibility of using 1d models to analyze localization in slender structures
easily and accurately has emerged recently in the context of necking in bars and bulging in balloons (Audoly
and Hutchinson, 2016; Lestringant and Audoly, 2018). Several other localization phenomena could be better
understood if 1d models were available.
Our method can be described in general terms as follows. First, we introduce the so-called canonical
form, which is a unified and abstract formulation into which the various structural models for slender
structures can be cast. The canonical form serves as a starting point for the reduction process. The set of
degrees of freedom are split between (master) macroscopic degrees of freedom which are retained in the 1d
model, and (slave) microscopic degrees of freedom which are relaxed; the choice of which degrees of freedom
are retained as master is left to the user. Next, the asymptotic expansion is carried out: the equations
of elasticity are expanded about a configuration having finite and inhomogeneous pre-strain, whereby each
cross-section is in a state parameterized by the local value of the macroscopic degrees of freedom. This
2
expansion is implemented as a series of steps (i.e., a mere recipe, albeit a slightly technical one at places)
which ultimately yields a 1d elastic potential governing the reduced model. Classical structural model,
such as the Euler-Bernoulli beam model, are recovered at the dominant order while corrections depending
on the strain gradient are obtained at the subdominant order. In this first paper, the general method is
presented and illustrated on simple examples for which the 1d strain-gradient model is already known from
the literature; the method will be applied to original problems in future work.
In section 2, we give a general account of the reduction method: the series of steps needed to carry
out the reduction are listed. In Sections 3 to 5, three examples of applications are worked out, by order
of increasing complexity: we establish the 1d models for the bulging of inflated membranes, for a linearly
elastic block in 2d, and for an axisymmetric hyperelastic cylinder. In Section 6, we conclude and make a
few general remarks about the method. Appendix A presents a detailed proof of the reduction method.
Appendix B provides the detailed calculations for the analysis of an axisymmetric cylinder.
In mathematical formula, we use bold face symbols for vectors and tensors. Their components are
denoted using plain typeface with a subscript, as in h = (h1, h2). Functions of a cross-sectional coordinate
are denoted by surrounding their generic values using curly braces, with their dummy argument in subscript,
as in f = {f (T )}T . We denote by S the longitudinal coordinate. In our notation, the primes will be reserved
for the derivation with respect to the longitudinal coordinate S,
f(cid:48) =
df
dS
.
2. Main results
We present the dimension reduction method in a generic and abstract form that will be applied to
specific structures in the forthcoming sections. We limit attention to a practical description of the method:
the method is justified in full details separately in Appendix A.
2.1. Starting point: full model in canonical form
The elastic model used as a starting point for the dimension reduction will be referred to as the full
model . One can use a variety of full models, such as an axisymmetric membrane model (for the analysis of
bulges in balloons), a hyperelastic cylinder (for the analysis of necking) or a shell model (for the analysis
of a tape spring). We start by casting the full models into a standardized form, called the canonical form,
which exposes their common properties and hides their specificities. The conversion of particular structural
models into the canonical form is not discussed here, and will be demonstrated based on examples in the
following sections.
We assume that the structure is invariant along its longitudinal direction in the reference configuration,
i.e., it is a block in two dimensions or a prismatic solid in three dimensions. The reference configuration
does not need to be stress-free: naturally curved or twisted elastic rods for instance can be handled. The
extension to structures whose geometric or mechanical properties are not invariant but slowly varying in the
longitudinal direction is straightforward and will be discussed at the end of the paper. We denote by S a
Lagrangian coordinate along the long dimension of the structure. The range of variation of S is denoted as
0 ≤ S ≤ L, where L typically denotes the natural length of the structure. The parameter S is used to label
the cross-sections, see figure 1(a).
Let E(S) be the strain map over a particular cross-section S in the current configuration, as sketched in
figure 1(b). By strain map, we mean that E(S) is the restriction to a particular cross-section of the set of
strain measures relevant to the particular structural model. If we are dealing with a hyperelastic cylinder, for
instance, E(S) collects all the strain components {ESS(·,·), ESX (·,·), EXX (·,·), ESY (·,·), EY Y (·,·), EXY (·,·)},
each taking the cross-section coordinates (X, Y ) as arguments.
Next, we introduce two mathematical objects in each cross-section: a vector of macroscopic strain
h(S) = (h1(S), h2(S), . . .) made up of the strain measures that will survive in the 1d model, and a set of
microscopic degrees of freedom y(S) that will be ultimately be eliminated. Their exact definitions vary, but
typically h(S) is the (apparent) 1d strain, as calculated from the center line passing through the centers
3
Figure 1: Dimension reduction for an abstract slender structure. Left column (a,a'): reference configuration highlighting a
particular cross-section with coordinate S. Right-hand side column (b,b'): deformed configuration. Top row (a,b): full model
used as a staring point, including a microscopic displacement y(S) and a microscopic strain E(S). Bottom row (a',b'): equivalent
1d model obtained by dimension reduction, in which the details at the scale of cross-section are effectively hidden.
of all the cross-section, while y(S) parameterizes the deformation of the cross-section relative to the center
line. Typically, h(S) is a vector of low dimension, while y(S) is a (collection of) functions defined over the
cross-sections, i.e., an infinite-dimensional vector. For the axisymmetric hyperelastic cylinder, for instance,
h is made up of a single entry, the axial stretch, while y(S) is the cross-sectional map of displacement.
Together, the macroscopic strain h(S) and microscopic degrees of freedom y(S) determine the current
configuration of the structure (up to a rigid-body motion) hence the microscopic strain E. Therefore, each
particular structural model prescribes a method for calculating the cross-sectional strain map E(S) in terms
of h, y and their longitudinal derivatives,
(cid:48)
E(S) = E(h(S), h
(S); y(S), y(cid:48)(S), y(cid:48)(cid:48)(S)).
(2.1)
In the example of the cylinder, the longitudinal strain ESS depends on the longitudinal gradient of the
displacement, hence the dependence of E on y(cid:48).
Since y(S) is a function defined on the cross-sections, the function E in the right-hand side of (2.1) is
a functional when the domain of the cross-section is continuous. Since E(S) is a map of strain over the
cross-section, any dependence of the strain on the transverse gradients of displacement is hidden in the
definition of E above. By contrast, we make sure that the dependence on longitudinal gradients takes place
explicitly through the supplied argument y(cid:48) (and possibly y(cid:48)(cid:48)). The additional dependence on y(cid:48)(cid:48) will allow
us to handle the bending of plates or shells without change. It is easy to take into account an additional
dependence of E on higher-order gradients of h(S) or y(S); this does not affect any of the results.
In terms of the strain map E(S), the structural model defines a density of strain energy W (E) per unit
length dS. The strain energy of the structure therefore writes
(cid:90) L
Φ[h, y] =
(cid:48)
W (E(h(S), h
(S); y(S), y(cid:48)(S), y(cid:48)(cid:48)(S))) dS,
(2.2)
where the square brackets emphasize the functional dependence on the arguments.
0
Some structural models are conveniently expressed by imposing kinematic constraints q(y) = 0 on
the microscopic displacement, where q(y) = (q1(y), q2(y), . . .). For structures whose cross-section involve
infinitely many degrees of freedom, y(S) is (a set of) functions defined in the cross-sections, i.e., q is a
functional. We focus attention on kinematic constraints that are linear and independent of S. These
assumptions can be relaxed easily. For structures that are free of kinematic constraints, we set q as the
empty vector, q(y) = (), implying that any term such as q(y) · x = 0 must be discarded in the following.
We deal with dimension reduction by addressing the following relaxation problem: the macroscopic strain
h(S) is prescribed and we seek the microscopic variables y(S) making the strain energy Φ[h, y] stationary,
subject to the kinematic constraint
∀S q(y(S)) = 0.
(2.3)
Our goal is to calculate the relaxed strain energy Φ(cid:63) in terms of the macroscopic strain h(S). It can be
4
(a)(b)(a')(b') obtained by inserting the optimal microscopic displacement y(S) into Φ, as in
Φ(cid:63)[h] =
min
y:(∀S)q(y(S))=0
Φ[h, y].
(2.4)
This paper derives an expansion of Φ(cid:63)[h] in successive derivatives of h(S) using an asymptotic method.
Note that the relaxed energy Φ(cid:63)[h] is 1d: it no longer makes any reference to the cross-sectional degrees of
freedom. Once Φ(cid:63)[h] has been obtained, the equilibrium equations for the 1d model can be derived using
standard variational techniques.
2.2. Analysis of homogeneous solutions
The first step in our analysis is to characterize homogeneous solutions under finite strain. To do so,
we focus attention on the case where both the macroscopic strain h = (h1, h2, . . .) and the microscopic
displacement y are independent 1 of S. The strain for homogeneous solutions E(h, y) is obtained by setting
(cid:48)
h
= 0, y(cid:48) = 0 and y(cid:48)(cid:48) = 0 in (2.1) as
E(h, y) = E(h, 0; y, 0, 0).
(2.5)
For a given value of the macroscopic strain h = (h1, h2, . . .), we seek the microscopic displacement y = yh =
y(h1,h2,...) such that the cross-sections are in equilibrium. To do so, we seek the value(s) of y that make
stationary the strain energy per unit length W ( E(h, y)), among those satisfying the kinematic constraint
q(y). This yields the variational problem
∂y (h, yh) · y
+ f h · q(y) = 0
(2.6)
(cid:40) ∀y − dW
dE ( E(h, yh)) ·(cid:16) ∂ E
q(yh) = 0,
(cid:17)
where the unknown f h is a Lagrange multiplier enforcing the constraint on the second line (f h can be inter-
preted as the macroscopic load that is required for the homogeneous solution to be globally in equilibrium,
such as a transverse external load in the case of a rod subject to a combination of uniform tension and
bending). For structures whose cross-sections define a continuous domain in the plane, W (E) is a functional
taking on scalar values, and dW
structures possessing discrete cross-sectional degrees of freedom, dW
dE (E) · δE denotes the Gateaux derivative at E in the direction δE. For
dE (E) the gradient of the function W (E).
Equation (2.6) warrants stationarity with respect to the microscopic displacement, but not with respect
to the macroscopic strain. For a solution of these equations to represent an actual equilibrium, one would
need to set up macroscopic forces conjugate to the macroscopic strain, labeled F h in figure 2.
If the
structure is an elastic cylinder, for instance, equation (2.6) imposes the contraction of cross-sections by
Poisson's effect; to maintain the global equilibrium, a macroscopic tensile load, not discussed here, would
be required. Macroscopic load do not enter into the dimension reduction process: they can be introduced
directly in the 1d model, after the dimension reduction.
Equation (2.6) is a non-linear elasticity problem defined on the cross-section: the longitudinal variable
has been removed. This problem can be solved, most often analytically (see the examples in the following
sections) or in some cases numerically. By solving equation (2.6) for yh and f h for any value of the
macroscopic strain h, one obtains a catalog of homogeneous solutions, which is at the heart of the dimension
reduction method. It is derived without any approximation: the catalog is made up of nonlinear solutions.
In terms of the catalog of microscopic displacement yh, we can define the homogeneous strain Eh, the
homogeneous strain energy density Whom(h), the homogeneous pre-stress Σh, and the homogeneous tangent
stiffness, as follows,
Eh = E(h, yh)
Whom(h) = W (Eh)
dE (Eh)
dE2 (Eh).
Σh = dW
Kh = d2W
(2.7)
1Here, we are assuming that the splitting of the microscopic strain E(S) into master (h(S)) and slave (y(S)) degrees of
freedom has been set up in such a way that homogeneous solutions correspond to constant h and constant y. Any reasonable
choice of h(S) and y(S) satisfies this property.
5
Figure 2: A homogeneous solution with uniform macroscopic strain h: microscopic displacement yh, microscopic strain Eh
and microscopic stress Σh. Note that the we are not interested at this stage in calculating the external loading F h that
maintains equilibrium with respect to the macroscopic variables.
In our notation, the homogeneous quantities are either subscripted with the letters 'hom', as in Whom, or
simply by the vector of macroscopic strain h = (h1, h2, . . .) on which they depend.
2.3. Reduced models without gradient effect
Most 1d models used for slender structures depend on macroscopic strain variables, but not on their
gradients. The Euler-Bernoulli rod model, for instance, depends on curvature and twist and not on their
gradients. These standard structural models are governed by the strain energy
(cid:90) L
Φ(cid:63)[h] ≈
Whom(h(S)) dS
(reduction without strain gradient),
0
energy (cid:82) L
If we start from an
and can therefore be derived directly from the catalog of homogeneous solutions.
elastic block, for example, and choose the axial strain and curvature as macroscopic variables, the strain
0 Whom(h(S)) dS defines a classical beam model (see §4.7). Note that the 1d model associated
with the energy functional Φ(cid:63)[h] above might suffer from poorer convexity properties than the original 3d
model; this happens typically when a string model is derived (i.e., when h(S) is set up to include just an
axial strain variable, so that there is no bending energy in the resulting 1d model), and in this case an
additional relaxation step is needed to remove the unphysical part of the constitutive law predicting axial
compression (Acerbi et al., 1991).
So far, our method carries out dimension reduction without strain gradient. It does so without using any
kinematic assumption and works under very general conditions: no material symmetry has been assumed,
and it can handle inhomogeneous cross-sections and nonlinear elastic materials.
2.4. Microscopic correction, energy expansion
We return to the main focus of our work, which is on capturing strain-gradient effects. Given a distribu-
tion of macroscopic strain h(S) with 0 ≤ S ≤ L, we aim at calculating the optimal microscopic displacement
y(S) and, thus, the relaxed energy Φ(cid:63) appearing in equation (2.4). We do so by assuming slow variations in
the longitudinal direction: a proper stretched variable is introduced in the detailed proof of Appendix A,
but it will suffice here to assume that the successive longitudinal derivatives of quantities such as h(S) scale
like h = O(1), h
(cid:48)
= O(γ2), etc., where γ (cid:28) 1 is a slenderness parameter.
= O(γ), h
(cid:48)(cid:48)
We seek the microscopic displacement y(S) that achieves the optimum in equation (2.4) in the form:
y(S) = yh(S) + z(S).
(2.8)
In words, we use the leading order microscopic displacement yh(S) obtained by looking up our catalog of
homogeneous solutions h (cid:55)→ yh as a first approximation; this look-up is performed with the parameter h
set to the local prescribed value of the macroscopic strain, h = h(S). We refine this approximation by a
(cid:48)
correction z(S) proportional to the gradient term h
, which we calculate next.
To reflect the change of unknown from y(S) to z(S) in (2.8), let us first define the function eh that
yields the strain as in (2.1):
†
eh(h
‡
, h
†
; z, z†, z‡) = E(h, h
; yh + z, h
† · ∇yh + z†, h
6
‡ · ∇yh + h
† · ∇2yh · h
†
+ z‡).
(2.9)
The variables bearing a dag (†) or a double dag (‡) are those that will be set later to the local value of
†
, for instance, is a dummy variable that will be set later to
the first or second gradients. The quantity h
(S). Besides, the ∇ in equation (2.9) stands for gradients with respect to the macroscopic strain h,
†
h
(cid:48)
= h
Anticipating on the fact that we will need to expand the strain in (2.9), we define the structure coefficients
eij
klm(h) as the gradients of eh, evaluated at a homogeneous solution: for any set of integers (i, j, k, l, m),
∇kyh =
dkyh
dhk .
(2.10)
eij
klm(h) =
†
[∂h
‡
]i [∂h
]j [∂z]k [∂z†]l [∂z‡]m
∂(i+j+k+l+m)eh
(0, 0; 0, 0, 0).
(2.11)
‡
and h
Note that the upper set of indices correspond to gradients with respect to the gradients of the macroscopic
†
strain parameters h
while the lower set of indices correspond to gradients with respect to the
microscopic variable z and its gradients z† and z‡. The quantities eij
klm(h) are either tensors, or operators
(if the cross-sectional degrees of freedom are continuous and at least one integer among k, l, m are non-
†
zero): they will always appear contracted i times with h
, k times with z, etc. Each one of
these contractions will be denoted by a dot, representing either the standard contraction of tensors or the
application of the operator.
‡
, j times with h
The Taylor expansion of the strain (2.9) near a homogeneous solution can be written in terms of the
structure coefficients as
†
eh(h
‡
, h
; z, z†, z‡) = eh(0, 0; 0, 0, 0) + e10
000(h) · h
†
+ e00
100(h) · z +
1
2
(cid:16)
† · e10
2 h
100(h) · z + ···(cid:17)
+ ···
Structure coefficients will be calculated explicitly in the second part of the paper, when explicit structures
are considered.
In terms of the structure coefficients, we further introduce the following operators,
Ah · h
= Σh · (e10
000(h) · h
†
†
)
= Σh · (e01
h · h
000(h) · h
‡
‡
C(0)
)
h · z† = Σh · (e00
010(h) · z†)
C(1)
† · B(0)
h · h
) · Kh · (e10
000(h) · h
†
†
1
2 (e10
2 h
h · z = (e10
† · B(1)
000(h) · h
) · Kh · (e00
†
h
2 z · B(2)
h · z = 1
100(h) · z) · Kh · (e00
1
2 (e00
= 1
) + 1
000(h) · h
†
100(h) · z) + Σh · (h
100(h) · z) + 1
2 Σh · (h
† · e10
2 Σh · (z · e00
† · e20
000(h) · h
) − h
†
100(h) · z) − h
† · ∇C(1)
200(h) · z).
† · ∇C(0)
h · z
h · h
†
(2.12)
They depend on the (local) macroscopic strain h. They operate on the cross-sectional degrees of freedom z
†
and z† (but not z‡) and on the local values of the derivatives h
of the macroscopic strain.
‡
and h
As shown in Appendix A, the expansion of the energy Φ[h, yh + z] in powers of the successive gradients
of macroscopic strain can be expressed in terms of these operators as
(cid:90) L
0
(cid:90) L
(cid:18) 1
(cid:90) L
0
0
2
Φ[h, yh + z] =
Whom(h(S)) dS +
Ah(S) · h
(cid:48)
(S) dS ···
+ [C(0)
h · h
(cid:48)
+ C(1)
h · z]L
S=0 +
(cid:48) · B(0)
h
h · h
(cid:48)
+ h
(cid:48) · B(1)
h · z +
(cid:19)
1
2
z · B(2)
(cid:48)3,h
h · z
(cid:48)(cid:48)h
dS ···
(cid:48)(cid:48)(cid:48)).
(cid:48)2,h
S
+ O(h
(2.13)
In the boundary term in square brackets, both the arguments h in subscript of the operators and the
(cid:48)
operands h
and z must be evaluated at S = 0 and S = L, respectively. Likewise in the integrand on the
(cid:48)
second line, the quantities h, h
and z must be evaluated at the current point S.
The form of the strain gradient model above is similar to that derived in different contexts, see for
example in Bardenhagen and Triantafyllidis (1994); our main contribution is a method for calculating the
coefficients Ah(S), B(1)
h , etc. explicitly.
7
2.5. Optimal correction
The last step in the reduction process is to determine the correction z(S) such that the microscopic
displacement (2.8) satisfies the optimality condition (2.4). All derivatives of the unknown z(S) can be
eliminated from equation (2.13), thanks to an integration by parts, as shown in Appendix A.5. The benefit
is that the relaxation of the unknown z leads to a local problem in the cross-sections: as established in
Appendix A, the optimal correction z(S) S is
z (cid:55)→(cid:16) 1
where the dominant contribution zopt = O(h
(cid:48)
2 h
(S) · B(0)
h(S) · h
(cid:48)
(cid:48)
(S) + h
(S) · B(1)
h(S) · z + 1
(cid:48)) is the one that minimizes the local elastic potential
2 z · B(2)
, subject to the constraint q(z) = 0. The
h(S) · z
correction zopt(S) is therefore the solution to the following variational problem,
z(S) = zopt(S) + O(h
(cid:48)2),
(cid:17)
h(S) · z + zopt(S) · B(2)
h(S) · z − f opt(S) · q(z) = 0
(2.14)
(cid:40) ∀z h
(cid:48)
(S) · B(1)
q (zopt(S)) = 0,
where f opt(S) is a Lagrange multiplier, to be determined as part of the solution process.
(cid:48)
This variational problem is linear with respect to the local value of the strain gradient h
(cid:48)
that its solution zopt(S) is proportional to h
(S), i.e., there exists a catalog of corrections Zh(S)
opt
(S). This implies
such that
zopt(S) = Zh(S)
opt
· h
(cid:48)
(S).
(2.15)
The catalog Zh
opt is found by solving (2.14). It can be determined once for all in terms of the geometric and
mechanical properties of a reference cross-section and in terms of the macroscopic strain h, as we show in
the examples.
(cid:48)
Equation (2.14) is a problem of linear elasticity in the cross-section. The first term h
h(S) · z
can be interpreted as a pre-stress arising from the presence a gradient (an interpretation of this pre-stress
term will be obtained based on the analysis of specific structures, see §5.4 in particular). The second term
zopt(S) · B(2)
h(S) in equation (2.12)
has two contributions: a tangent elastic stiffness Kh(S), and a geometric stiffness arising from the pre-stress
Σh(S) associated with the local state of stress.
h(S) · z is an elastic stiffness term which, in view of the definition of B(2)
(S) · B(1)
2.6. Relaxed energy
The relaxed energy Φ(cid:63)[h] is finally obtained by inserting the optimal displacement y(S) = yh(S) +
zopt(S) + ··· into the energy expansion in equation (2.13). The result is
Φ(cid:63)[h] =
Whom(h(S)) dS +
Ah(S) · h
(cid:48)
(S) dS + [Ch(S) · h
(cid:48)
(S)]L
0 +
1
2
(cid:48)
h
(S) · Bh(S) · h
(cid:48)
(S) dS + . . .
(2.16)
Here, the operator Ah has been introduced in equation (2.12) and the additional elastic moduli Bh and Ch
are defined by
(cid:90) L
0
(cid:90) L
0
(cid:90) L
0
h −(cid:16)
(cid:17)T · B(2)
h · Zh
opt,
Bh = B(0)
Ch = C(0)
Zh
h · Zh
h + C(1)
opt.
opt
(2.17)
The energy functional in equation (2.16) and the explicit expression for the strain-gradient modulus Bh
are the main results of this paper.
In equation (2.16), the leading order term in the expansion depends Whom, and defines structural models
without the gradient effect, see § 2.3. The second term depending on Ah yields an energy contribution that
(cid:48)
: it is zero in most cases due to symmetry reasons, as shown in the
is linear with respect to the gradient h
8
Figure 3: An axisymmetric membrane:
(a) reference and (b) current configurations.
forthcoming examples. The terms depending on Ch is a boundary term arising from a gradient effect, while
the last term is the bulk strain-gradient term.
For further reference, we note that the strain gradient term is available in alternative form as
(cid:48)
h
(S) · Bh(S) · h
(cid:48)
(S) dS =
1
2
(cid:48)
h
(S) · B(0)
h(S) · h
(cid:48)
(S) dS − 1
2
zopt · B(2)
h · zopt dS.
(cid:90) L
0
1
2
(cid:90) L
0
(cid:90) L
0
2.7. A necessary stability condition at the microscopic scale
A necessary condition for the microscopic correction derived in section 2.5 to be stable (and, hence, for
h appearing in the microscopic
the relaxed energy Φ(cid:63) to be meaningful) is that the stiffness operator B(2)
problem in equation (2.14) is non-negative,
(∀z such that q(z) = 0)
z · B(2)
h · z ≥ 0.
(2.18)
Note that this condition does not warrant that the matrix Bh of strain-gradient moduli is non-negative, see
equation (2.17) (a matrix Bh having negative eigenvalues is indeed obtained for the elastic block, see §4.7).
However, equation (2.18) does warrant
1
2
(cid:48)
h
(S) · Bh(S) · h
(cid:48)
(S) ≤ 1
2
(cid:48)
h
(S) · B(0)
h · h
(cid:48)
(S),
which, as discussed in section 6, shows that our 1d model relaxes the elastic energy better than strain
gradient models derived from the ad hoc kinematic assumption z(S) = 0: this benefit is a consequence of
the fact that our 1d model is asymptotically exact.
3. Application to an axisymmetric membrane
Upon inflation, axisymmetric rubber membranes feature localized deformations in the form of propa-
gating bulges (Kyriakides and Chang, 1991). Standard dimension reduction without gradient terms yields
a non-convex elastic potential Whom and thus fails at describing the details of localization. Localized so-
lutions can be analyzed using the full membrane model (Fu et al., 2008; Pearce and Fu, 2010), but are
more easily and very accurately described based on a 1d strain-gradient model, as recently shown by the
authors, starting from the theory of axisymmetric elastic membranes and using a typical constitutive law for
rubber (Lestringant and Audoly, 2018). This 1d model is rederived here as a first illustration of the general
reduction method presented in section 2.
3.1. Full axisymmetric membrane model
The reference configuration is chosen as the natural, cylindrical configuration of the membrane, and the
natural radius of the circular membrane is denoted by ρ. In the current configuration, the membrane is
deformed under the action of an inflating pressure p, and a pulling force F equally distributed over the
terminal cross-sections, see figure 3. Natural boundary conditions are used, i.e., there is no restraint on the
terminal cross-sections.
9
(a)(b) (cid:16) √
An axisymmetric configuration of the membrane is parameterized by two functions Z(S) and R(S),
such that the cross-section with arc-length coordinate S in the reference configuration is transformed into
a circle perpendicular to the axis of the shell, with axial coordinate Z(S) and radius R(S), see figure 3(b).
We consider a standard set of strain measures from the theory of finite-strain axisymmetric elasticity,
E =
ρ , usually denoted as (E1, E2) = (λS, λΘ),
are the membrane stretches in the (principal) longitudinal and circumferential directions, respectively. The
additional 'strain' E3 has been included for convenience, as it allows us to write the potential energy of the
pulling force F as −F [Z(S)]L
ρ Z(cid:48) (cid:17)
Z(cid:48)2 + R(cid:48)2 and E2 = R
0 = −F(cid:82) L
Z(cid:48)2 + R(cid:48)2 R
: E1 =
√
0 E3 dS.
The sum of the membrane strain energy, and the potential energy of the loads p and F is captured by
an effective potential W (E) per unit length dS,
W (E) = W (E1, E2) − p π ρ2 E2
2 E3 − F E3,
where W (E1, E2) = W (λS, λΘ) is the strain energy of the hyperelastic membrane model (we use bars
generally for quantities relating to the full model). Upon integration with respect to S, the second term
yields (−p) times the volume enclosed by the membrane, which is the potential energy of the pressure
force. Note that we have chosen to include the potential energy of the loads p and F into the potential
0 W dS which normally captures the strain energy only; in line with this, the loading parameters p
Φ =(cid:82) L
and F are considered constant.
3.2. Macroscopic and microscopic variables
A natural choice of macroscopic strain parameter is the apparent axial stretch Z(cid:48)(S): this is the stretch
of a virtual bar obtained by collapsing all the circular cross-sections to a point located at their center.
However, this choice has the drawback that, for typical constitutive laws for rubber, there can be several
homogeneous solutions corresponding to a given value of the apparent stretch. To work around this difficulty,
it is preferable to define instead the macroscopic strain parameter as the hoop stretch h1(S) = E2(S) = R(S)
ρ .
As we will see, it is possible to reconstruct the apparent axial stretch Z(cid:48)(S) in terms of this h1(S). We thus
apply the general formalism using a single macroscopic strain and a single microscopic degree of freedom,
defined as
, y(S) = (y1(S)) = (Z(cid:48)(S)).
h(S) = (h1(S)) =
using R(S) = ρ h1(S) and Z(S) = Z(0) +(cid:82) S
With this choice of macroscopic and microscopic variables, it is possible to reconstruct the configuration
0 y1(S) dS, where Z(0) is an unimportant rigid-body translation.
As we do not need any constraint for this particular structural model, we set q(y) = () and drop all the
terms containing q(y) in the general formalism.
The strain vector E for the axisymmetric membrane given in section 3.1 can be cast in the canonical
(cid:18) R(S)
(cid:19)
ρ
form from equation (2.1) by choosing the strain function as
†
E(h, h
; y, y†, y‡) =
ρ2 h
†2
1 + y2
1 h1
y1
(cid:17)
,
(cid:16) (cid:113)
10
where the arguments are vectors whose length matches that of the macroscopic strain h and microscopic
†
variable y proposed above, i.e., h = (h1), h
†
1), y = (y1), y† = (y
= (h
†
1) and y‡ = (y
‡
1).
3.3. Homogeneous solutions
Homogeneous solutions are first analyzed, by setting to zero the derivative terms in the definition of the
strain, see equation (2.5). This yields the homogeneous strain as
E(h, y) =(cid:0) y1 h1
(cid:1) .
y1
The generalized stress in the homogeneous solution is given by the gradient of the potential E(S),
( E(h, y)) =(cid:0) ΣS(y1, h1) ΣΘ(y1, h1) − 2 p π ρ2 h1 y1 −F − p π ρ2 h2
(cid:1) ,
1
dW
dE
where ΣS(λS, λΘ) = ∂W
(λS, λΘ) are the components of the Piola-Kirchhoff
∂λS
stress in the longitudinal and circumferential directions, respectively, as predicted by the elastic strain
potential W (λS, λΘ) characterizing the elastic properties of the membrane.
(λS, λΘ) and ΣΘ(λS, λΘ) = ∂W
∂λΘ
Next, we proceed to write and solve the principle of virtual work (2.6) for homogeneous solutions. Noting
∂y (h, y) =(cid:0) 1
that ∂ E
1 (cid:1), it writes
0
− ΣS(y(h1)
1
, h1) + F + p π ρ2 h2
1 = 0.
(3.1)
This is an equation for the longitudinal stretch y(h1)
in a homogeneous solution, in terms of the hoop stretch
h1 = R
ρ . The load parameters F and ρ are considered fixed, i.e., the dependence on F and ρ will always
be silent. Equation (3.1) expresses the equilibrium of a homogeneous solution in the longitudinal direction.
For typical constitutive laws, equation (3.1) cannot be solved explicitly for y(h1)
in terms of h1, and will be
viewed as an implicit equation.
1
1
In terms of the homogeneous solution y(h1)
, we obtain the properties of homogeneous solution from
equation (2.7) as
=
y(h1)
E(h1)
1
Whom(h1) = W (y(h1)
Σ(h1)
=
1
y(h1)
1
h1
, h1) − ΣS(y(h1)
, h1) y(h1)
, h1) ∗ −ΣS(y(h1)
1
1
, h1)
(cid:17)
(3.2)
1
(cid:16)
(cid:16)
K SS(y(h1)
ΣS(y(h1)
1
1
∗
0
(cid:17)
1
.
, h1) ∗
∗
∗
0
∗
0
K(h1)
=
Here, the star symbol (∗) denote quantities that play no role and do not need to be calculated. The quantity
K SS(λS, λΘ) = ∂2W
(∂λS )2 (λS, λΘ) is the tangent elastic modulus, as calculated from the membrane model.
3.4. Change of microscopic variable
According to the general method, we introduce a correction z(S) = (z1(S)) to the microscopic variable
+ z1(S). In terms of the new unknown, the strain function
by y(S) = yh(S) + z(S), i.e., y1(S) = y(h1(S))
reads, see equation (2.9),
1
(cid:16) (cid:113)
†
eh(h
‡
, h
; z, z†, z‡) =
ρ2 h
†2
1 + (y(h1)
1 + z1)2 h1
y(h1)
1 + z1
(cid:17)
,
†
‡
= (h
1), h
‡
1), z = (z1), z† = (z
= (h
where again the arguments are vectors whose dimension is imposed by the macroscopic strain and microscopic
‡
†
variable as h
1). Note that the strain function eh = e(h1)
depends on the macroscopic strain h1 which appears in subscript, and that we have made use of the catalog
of homogeneous solutions y(h1)
The structure coefficients eij
in the right-hand side.
klm(h) are the successive partial derivatives of the right-hand side above, see
; z, z†, z‡) with its
†
equation (2.11). These partial derivatives are most easily found by identifying eh(h
†
1) and z‡ = (z
‡
, h
1
(cid:18)
Taylor expansion
. The result is
†2
1 + z1 + ρ2 h
y(h1)
1
2 y(h1)
1
h1
000(h) = 0 e20
e10
100(h) =(cid:0) 1
000(h) =
e00
y(h1)
1 + z1
(cid:16) ρ2
1 (cid:1) e00
y(h1)
1
0
0
(cid:19)
(cid:17)
0
000(h) = 0 e10
e01
010(h) = 0.
200(h) = 0 e00
100(h) = 0
(3.3)
These are the only structure coefficients that are required in the following. Recall that the dimension of
klm, which are of dimensions 3× 1×···× 1
both h and y is one for an axisymmetric membrane: the tensors eij
according to the general rule, where the one's are repeated i + j + k + l + m times, have been identified with
vectors of dimension 3.
11
When these expressions are combined with those for homogeneous quantities obtained in (3.2), one can
calculate the first batch of operators from equation (2.12) as
= Σh · (e10
Ah · h
†
= Σh · (e01
h · h
‡
C(0)
h · z† = Σh · (e00
C(1)
h · h
† · B(0)
†
1
2 h
h · z = 0 · Kh · (∗) + Σh · 0 − h
† · B(1)
h
2 z · B(2)
h · z = 1
1
2 Σh ·(cid:16) ρ2
1 (cid:1) · Kh ·(cid:0) 1
000(h) · h
) = Σh · 0 = 0
†
) = Σh · 0 = 0
000(h) · h
‡
010(h) · z†) = Σh · 0 = 0
†
1)2
(h
2 0 · Kh · 0 + 1
(cid:0) 1
y(h1)
1
† · 0 · z = 0
1 (cid:1) + 1
= 1
0
0
0
2
(cid:17) − h
† · 0 · h
†
0
= 1
2 ΣS(y(h1)
1
, h1) ρ2
y(h1 )
1
†
1)2
(h
2 Σh · 0 = 1
2 K SS(y(h1)
1
, h1) (z1)2.
3.5. Local optimization problem
The local optimization problem (2.14) is particularly simple, because it has no source term (B(1)
h = 0)
and no constraint term (q(y) = ()). In view of the operators just derived, it reads
∀z1
z1 K SS(y(h1)
1
, h1) zopt
1 = 0.
We rule out the possibility of a material instability in the membrane model, i.e., K SS(y(h1)
, h1) > 0 (note that
with this assumption of material stability at the 'microscopic' level, the stability condition from section 2.7 is
automatically satisfied). The variational problem above can then be solved for zopt =(cid:0)zopt
opt =(cid:0) 0 (cid:1).
The correction to the microscopic variable arising from the gradient effect is zero for this particular structure.
To comply with the general form of equation (2.15), we set accordingly Zh
(cid:1) as zopt( S) = 0.
1
1
3.6. Regularized model
In view of equation (2.17), we obtain the operators entering into the strain-gradient model as Ah = 0,
Bh = (B(h1)
11 ) where B(h1)
11 = ρ2 ΣS (y(h1 )
y(h1)
1
1
,h1)
and Ch = 0.
(cid:0)λhom
We switch to the more standard notation λΘ = h1 = R
(cid:1) = F + p π ρ2 λ2
ρ for the hoop stretch and λS = y1 for the
apparent axial stretch, and recapitulate the main results for the axisymmetric membrane as follows. We
must first solve the implicit equation (3.1) for the apparent axial stretch y(h1)
(λΘ), which reads
Θ and yields the homogeneous equilibria of the balloon. In this equation,
ΣS
(λΘ), λΘ
ΣΘ(λS, λΘ) = ∂W
In terms of this catalog of
∂λΘ
homogeneous solutions, we can calculate Whom(λΘ) by (3.2). The balloon is governed by the strain-gradient
bar model, see equation (2.16),
(λS, λΘ) is the hoop stress in the homogeneous solution.
= λhom
S
S
1
Φ(cid:63)[λΘ] ≈
Whom(λΘ(S)) +
B(λΘ(S)) λ(cid:48)2
Θ(S)
1
2
where the strain-gradient modulus reads
0
ρ2 ΣS
B(λΘ) =
(cid:0)λhom
S
λhom
S
(λΘ), λΘ
(cid:1)
,
dS,
(3.4a)
(3.4b)
and where ΣS(λS, λΘ) = ∂W
∂λS
(λS, λΘ) is the longitudinal stress in the homogeneous solution.
3.7. Comments
We have recovered the model established by Lestringant and Audoly (2018). Typical solutions predicted
by the 1d model are compared to those of the full axisymmetric model in figure 4: the 1d models appears
to be highly accurate, even in the regime where the bulges are fully localized.
The axisymmetric membrane model, which we used as a starting point was already a 1d model: it does
not make use of any transverse variable, and has discrete degrees of freedom (Z, R) in each cross-section. The
12
(cid:90) L
(cid:20)
(cid:21)
Figure 4: Solutions for a propagating bulge in an axisymmetric membrane with initial aspect ratio L/ρ = 30: comparison of
the predictions of the full axisymmetric membrane model (§3.1) and of the reduced model in equation (3.4), from Lestringant
and Audoly (2018). The material model for rubber proposed by Ogden (1972) is used, with the same set of material parameters
as used in the previous experimental work of Kyriakides and Chang (1991), see also §2 in Lestringant and Audoly (2018).
reduction method led us to another 1d model and it therefore is improper to speak of dimension reduction
in this case. The reduction method is still useful, as reduced model is simpler and, more importantly, much
more standard: it is the well-known diffuse-interface model introduced by van der Walls in the context of
liquid-vapor phase transition, as discussed by by Lestringant and Audoly (2018).
Even when bulges are fully formed, the typical length of the interface between the bulged and unbulged
regions never gets much less than ∼ ρ, i.e., remains always much larger than than the membrane's thickness
t (assuming the membrane is thin in a first place ρ (cid:29) t). This warrants that the assumptions underlying
the membrane model remain valid. To address the case of thick membranes, i.e., when the ratio t/ρ is not
small, one could apply our reduction method to a theory of thick membranes, or to a finite-strain model for
a hyperelastic cylinder in 3d.
4. Application to a linearly elastic block
Our next example is a homogeneous block of linearly elastic material in 2d, having length L and thickness
a, as sketched in figure 5. We account for both stretching and bending of the block. In the first step of
the dimension reduction, we will recover the classical beam model. Its energy is convex, implying that this
particular structure does not tend to localize. The strain-gradient model obtained at the next step is still of
interest as its solutions generally 2 converge faster towards those of the full (2d) elasticity model than those
of the classical beam model.
There is a large amount of work on higher-order asymptotic expansions for prismatic solids in the specific
context of linear elasticity with the aim to derive linear higher-order beam theories, see for instance the work
of Trabucho and Viano (1996). The forthcoming analysis shows that these results can be easily recovered
with our method. It also reveals that the assumption of linear elasticity brings in severe, somewhat hidden
limitations.
The elastic block is our first example where a cross-section possesses infinitely many degrees of freedom.
4.1. Full model: a linearly elastic block in 2d
We consider an elastic block in reference configuration. The axial and transverse coordinates in reference
configuration are denoted as S and T , respectively, and are used as Lagrangian coordinates. Their domains
are 0 ≤ S ≤ L and −a/2 ≤ T ≤ a/2.
2It is known, however, that boundary conditions can prevent strain-gradient models from converging faster. This happens
when the imposed boundary conditions are incompatible with the kinematics y(S) = yhom(h(S)) + zopt + · · · of the strain-
gradient model at the microscopic scale. We do not address this question in this paper, and limit attention to natural boundary
conditions.
13
axisym. membranereduced model202.55.0406080100202.55.0406080202.50.00.00.05.040 Figure 5: A block of a linearly elastic material in (a) reference and (b) current configuration. The 1d model makes use of the
center line (brown curve), defined as the curve passing through the centers of mass (brown dots) of the cross-section. The
macroscopic strain are the apparent stretch h1(S) = U(cid:48)(S) and the apparent curvature h2(S) = V (cid:48)(cid:48)(S) of the center line.
We introduce the displacement (u, v) in a Cartesian frame (ex, ey) aligned with axes of the undeformed
block: a point with position (S, T ) in reference configuration gets mapped to x(S, T ) = (S + u(S, T ), T +
v(S, T )) in the current configuration, see figure 5(b). The linear strain is presented in vector form as
E =(cid:0) ∂u
∂S
(cid:0) ∂u
(cid:1) (cid:1)
∂v
∂T
1
2
∂T + ∂v
∂S
(4.1)
where E1, E2 and E3 are respectively the SS, T T and ST components of the 2-d strain tensor from linear
elastic theory. We use a linear isotropic and uniform constitutive in 2d (Hookean elasticity), corresponding
to an elastic potential per unit length dS
(cid:90) +a/2
−a+2
W (E) =
1
2
(2 µ (E2
1 + E2
2 + 2 E2
3 ) + λ (E1 + E2)2) dT,
(4.2)
(cid:82) +a/2
where (cid:104)f(cid:105)(S) = 1
a
where the elastic constants µ and λ are known as the Lam´e parameters.
4.2. Macroscopic and microscopic variables
We choose to define the center line as the curve passing through the centers of mass of the cross-sections.
The components of the center line displacement are therefore
U (S) = (cid:104)u(cid:105)(S) V (S) = (cid:104)v(cid:105)(S),
(4.3)
−a/2 f (S, T ) dT denotes the cross-section average of a function f (S, T ).
The deformed center line is parametrized as (S + U (S)) ex + V (S) ey. In the theory of linear elasticity,
it is associated with an apparent longitudinal strain U(cid:48)(S), deflection angle V (cid:48)(S), and curvature V (cid:48)(cid:48)(S),
where by 'apparent' we emphasize the fact that the center line is non-material. In our reduction of the elastic
block to a 1d model, we use as macroscopic strain measures these apparent axial strain and curvature,
h(S) = (U(cid:48)(S), V (cid:48)(cid:48)(S)).
Let x(S, T ) the final position of the point initially at position S ex + T ey if the cross-section S were to
undergo a rigid body motion following the center line, namely the combination of a rigid-body translation
(U (S) ex +V (S) ey) and a rigid-body rotation with angle V (cid:48)(S). Since V (cid:48)(S) is infinitesimal, the unit normal
to the center line writes −V (cid:48)(S) ex + ey and so x(S, T ) = [(S + U (S)) ex + V (S) ey] + T [−V (cid:48)(S) ex + ey].
We choose to define the microscopic displacement y(S, T ) as the difference between the actual position
x(S, T ) = (S + u(S, T )) ex + (T + v(S, T )) ey and x(S, T ):
y(S, T ) = x(S, T ) − x(S, T ) = (u(S, T ) − U (S) + V (cid:48)(S) T ) ex − (v(S, T ) − V (S)) ey.
The Cartesian components are found as y1(S, T ) = u(S, T )− U (S) + V (cid:48)(S) T and y2(S, T ) = v(S, T )− V (S).
This definition of y warrants y(S, T ) = 0 automatically whenever the block is moved rigidly, since x(S, T ) =
x(S, T ) in this case.
14
(a)(b) In our general presentation of the method in section 2, y(S) (with a single argument) was defined as the
collection of the microscopic degrees of freedom on a given cross-section S. To comply with this convention,
we define y(S) = (y1(S), y2(S)) as a pair of functions defined on the cross-section taking the transverse
coordinate T as an argument,
y1(S) = {(u(S, T ) − U (S) + V (cid:48)(S) T )}T
y2(S) = {v(S, T ) − V (S)}T .
We recall that {g(T )}T is a notation for the function g that maps T to g(T ), the index T appearing in
subscript after a curly brace being a dummy variable.
In view of equation (4.3), the microscopic displacement must satisfy the condition (cid:104)u(cid:105)(S) = U (S) and
(cid:104)v(cid:105)(S) = V (S). Upon elimination of (u, v) in favor of (y1, y2), this yields (cid:104)y1(S)(cid:105) = (cid:104)y2(S)(cid:105) = 0 for all S.
We handle these constraints by setting
(cid:32)(cid:90) a/2
−a/2
q(y) =
1
a
(cid:33)
(cid:90) a/2
−a/2
y1(T ) dT,
y2(T ) dT
in the general formalism of section 2.
The displacement in the Cartesian basis is u(S, T ) = U (S) + [y1(S)](T ) − V (cid:48)(S) T and v(S, T ) =
[y2(S)](T ) + V (S), and therefore the strain in equation (4.1) can be expressed as
E(S) =(cid:0) h1(S) + [y(cid:48)
1(S)](T ) − h2(S) T ∂T [y2(S)](T )
1
2 (∂T [y1(S)](T ) + [y(cid:48)
2(S)](T )) (cid:1) .
2(S) = {y(cid:48)
(cid:111)
(cid:17)
†
2(T ))
T
In the above expression, y(cid:48)
Since primes are reserved for derivatives with respect to the longitudinal variable S, we use the symbol ∂T
for transverse derivatives.
2(S, T )}T .
For consistency with the discrete case, we define the strain function E(. . .) as an operator that takes as
arguments the pair of functions y = y(S) = (y1(S), y2(S)), and their derivatives y† = y(cid:48)(S), as well as the
pair of scalars h(S) = (h1(S), h2(S)), and returns the strain map in the cross-section,
1(S) denotes the function {y(cid:48)
1(S, T )}T , and similarly y(cid:48)
†
E(h, h
, y, y†, y‡) =
†
1(T ) − h2 T}T
{∂T y2(T )}T
2 (∂T y1(T ) + y
.
(4.4)
(cid:16) {h1 + y
(cid:110) 1
We use the same ordering conventions for the strain components as in equation (4.1), i.e. the longitudinal,
transverse and shear strain appear in this order. We continue to use the same notation as earlier whereby
†
†
†
variables bearing a dagger, such as h
= (h
1, h
2) are dummy variables that are intended to hold the local
(cid:48)
value of the derivative, here h
(S).
4.3. Homogeneous solutions
In the homogeneous case, the arguments of E corresponding to axial gradients in (4.4) (thus, bearing
a single or a double dagger) are all set to zero, see equation (2.5). Doing so, we are left with the map of
homogeneous strain,
The first variation of the strain energy (4.2) is calculated as
E(h, y) =(cid:0) {h1 − h2 T}T
(cid:90) +a/2
( Eh) · δE =
{∂T y2(T )}T
(cid:8) 1
2 ∂T y1(T )(cid:9)
(cid:0) ∗ λ E1 + (2 µ + λ) E2
4 µ E3
(cid:0) ∗ λ (h1 − h2 T ) + (2 µ + λ) ∂T y2
−a+2
T
(cid:1) .
(cid:1) · δE(T ) dT
(cid:1) · δE(T ) dT,
2 µ ∂T y1
δW =
=
dW
dE
(cid:90) +a/2
−a+2
where again stars denote values that play no role in the following.
15
By setting the variation of strain as δE = ∂ E(h,y)
∂y
tion (2.6), we obtain the principle of virtual work as
∀(y1, y2)
−(2 µ ∂T y1)
∂T y1 + f h
(cid:90) +a/2
(cid:18)
(cid:19)
dT +
1 y1
1
2
−a/2
· y = (cid:0) {0}T
(cid:90) +a/2
−a/2
where f h = (f h
1 , f h
2 ) is a Lagrange multiplier.
The solution satisfying the constraint q(y) = 0 is found as
(cid:110)−ν h1 T + ν h2
(cid:16) T 2
2 − a2
24
(cid:17)(cid:111)
T
1 = {0}T
yh
yh
2 =
f h = 0
(cid:8) 1
2 ∂T y1
(cid:9)
T
(cid:1) as in equa-
{∂T y2}T
(−(λ (h1−h2 T )+(2 µ+λ) ∂T y2) ∂T y2+f h
2 y2) dT = 0
where we have defined the 2-d Poisson's ratio ν = λ
Y = 4 µ (λ+µ)
(note that these expressions of ν and Y are valid for 2d elasticity but not for 3d elasticity).
We can then calculate the quantities characterizing the homogeneous solutions from equation (3.2) as
2 µ+λ
2 µ+λ . We also define the Young's modulus in 2d as
Eh = (cid:0) {h1 − h2 T}T
Σh · δE = (cid:82) +a/2
∗
δE · Kh · δE = (cid:82) +a/2
2 (Y a h2
−a+2
−a+2 δE(T ) ·
Whom(h) = 1
{−ν (h1 − h2 T )}T
1 + Y I h2
2)
(cid:0) Y (h1 − h2 T ) 0
{0}T
(cid:1)
0 (cid:1) · δE(T ) dT
· δE(T ) dT,
0
0
4 µ
∗
0
2 µ + λ
∗
0
−a/2 T 2 dT = a3
12 is the geometric moment of inertia of the cross-section.
where I =(cid:82) +a/2
(4.5)
Seeking the microscopic displacement as y(S) = yh(S) + z(S) from equation (2.8), we can calculate the
4.4. Change of microscopic variable
(cid:16) {h1 + z
strain in terms of the new microscopic variable z as
†
eh(h
‡
, h
; z, z†, z‡) =
†
1(T ) − h2 T}T
{−ν (h1 − h2 T ) + ∂T z2}T
†
1 T + ν h
†
†
2), z = ({z1(T )}T ,{z2(T )}T ) and z† = ({z
1, h
= (h
†
2
†
where h = (h1, h2), h
‡
elastic block, the structure coefficients do not depend on h
2
(cid:110) 1
The structure coefficients introduced in equation (2.11) are then obtained as
(cid:16)
(cid:110)− ν
(cid:16)
e20
000(h) = 0 e01
000(h) · h
†
e10
000(h) = 0 h
z · e00
=
† · e10
†
†
1 T − h
0 0
h
2
100(h) · z = 0 e00
2
†
1}T
Next, the operators introduced in (2.12) are calculated as follows,
200(h) · z = 0 e00
010(h) · z† =
24
. . .
(cid:17)(cid:111)
(cid:16)−ν h
(cid:16) T 2
(cid:17)
(cid:17)
†
2 − a2
+ ∂T z1 + z
2
†
†
2(T )}T ). For the linear
1(T )}T ,{z
(cid:17)(cid:17)(cid:111)
(cid:17)
(cid:8) 1
100(h) · z =(cid:0) 0 {∂T z2}T
(cid:111)
(cid:16) {z
(cid:17)
(cid:16) T 2
2 − a2
(cid:110) 1
or z‡.
2 ∂T z1
(cid:9)
(cid:1)
0
24
T
T
T
.
.
†
2
2 z
T
Ah = 0
C(0)
h = 0
h · z = (cid:82) +a/2
−a+2 µ ν2 (cid:16)
(cid:16)
(cid:82) +a/2
−a+2 Y (h1 − h2 T ) z1 dT
2 µ ν2 (cid:16) a3
2(cid:17)
†
1 T − h
h
(cid:16)
(cid:16)
h · z = −ν µ(cid:82) +a/2
†
+ a5
720 h
(cid:82) +a/2
2
†
†
1 T − h
h
2
2
= 1
2
= 1
−a+2
12 h
†
2
2
†
1
h · z = 1
2
2
C(1)
† · B(0)
h · h
†
1
2 h
† · B(1)
h
2 z · B(2)
1
(cid:17)(cid:17)2
(cid:17)(cid:17)
∂T z1 dT − Y (cid:82) +a/2
dT
T 2 − a2
12
T 2 − a2
12
−a+2(2 µ + λ) (∂T z2)2 + µ (∂T z1)2 dT.
16
(4.6)
†
†
1 − h
−a+2(h
2 T ) z1 dT
1
2
, zopt
(cid:90) +a/2
variational problem (2.14),
∀(z1(T ), z2(T ))
4.5. Local optimization problem
The correction zopt = (cid:0)zopt
(cid:18)
(cid:20)
1 − ν
(cid:90) +a/2
(cid:1) to the cross-sectional displacement is found by writing down the
(cid:18)
(cid:2)(2 µ + λ) ∂T zopt
(cid:21)
(cid:1) dT = 0.
(cid:11) = 0.
(cid:11) =(cid:10)zopt
We proceed to solve this variational problem together with the incremental constraint(cid:10)zopt
(cid:19)(cid:19)
As there is no source term in factor of z2, the transverse solution is easily found as zopt
2 = 0.
The remaining terms in the variational problem above concern the axial correction, and can be rearranged
(cid:19)(cid:19)
(cid:3) dT − 1
(cid:90) +a/2
∂T z1 − Y (h(cid:48)
(cid:0)f opt
(T ) = 0 and f opt
(cid:90) +a/2
1 T + h(cid:48)
(cid:32)(cid:32)
z1 + f opt
1 − h(cid:48)
2 ∂T z2
∂T z[1]
2 T ) z1
+ h(cid:48)
(cid:33)
(cid:33)
T 2
2
−h(cid:48)
(cid:18)
(cid:18)
dT . . .
−a+2
−a+2
a2
24
−a/2
z2
as
+
µ
a
2
2
2
1
2
1
2
T 2
2
−h(cid:48)
+ h(cid:48)
1 T + h(cid:48)
2
(cid:11) = 0 can be worked out as
2
a2
24
∂T z1 dT = 0.
f opt
1
a
∀z1(T )
+ Y h(cid:48)
− Y h(cid:48)
z1− µ
∂T zopt
1 − ν
2 T
1
−a+2
The solution to this equation satisfying the constraint(cid:10)zopt
(cid:17) − h(cid:48)
f opt
1
(T ) = h(cid:48)
1 Y
T 2 − a2
= −a h(cid:48)
(cid:16)
1
zopt
1
1 ν
2
12
The detailed expression of zopt
(cid:90) a/2
(cid:18) 1
To sum up, the displacement correction zopt =(cid:0)zopt
(cid:0)∂T zopt
a3 ν2
12
(cid:1)2
dT =
−a/2
2 +
h(cid:48)
1
1
30
, zopt
2
1
functions as zopt = Zh
opt · h
(cid:48)
, where
(cid:16)
(cid:32) ν
2
Zh
opt =
T 2 − a2
12
0
(cid:17) − 1
24 ((6 + 5 ν) a2 T − 4 (2 + ν) T 3).
2
(cid:19)
a5 h(cid:48)
2.
11 ν
180
7 ν2
240
+
+
(cid:1) can be written in terms of a fixed basis of
2
24 ((6 + 5 ν) a2 T − 4 (2 + ν) T 3)
0
(cid:33)
.
1 will not be used, other than to evaluate the following integral,
The entries in the top-left (respectively top-right) slot is a longitudinal displacement along ex in response
to a gradient of axial strain h1 = U(cid:48) (respectively, to a gradient of curvature h2 = V (cid:48)(cid:48)).
The necessary stability condition from section 2.7 requires 2 µ + λ ≥ 0 and µ ≥ 0, which are standard
condition of material stability in 2d, as discussed for example in Barenblatt and Joseph (1997).
4.6. Regularized model
Two last operators are defined in equation(2.17). They can now be evaluated as
(cid:18) a3
12
2(cid:19)
† · Bh · h
†
h
= µ ν2
2
†
h
1
+
a5
720
†
h
2
− µ
(∂T z1)2 dT
and
Ch · h
†
=
(cid:90) +a/2
−a+2
(cid:32)
Y (h1 − h2 T )
− 1
24 ((6 + 5 ν) a2 T − 4 (2 + ν) T 3)
· h
†
dT = Y a5 12 + 11 ν
720
†
h2 h
2.
+
11 ν
180
+
ν2
36
2
†
h
2
= −Y a5 6 + 5 ν
360
2
†
h
2
(4.7)
(cid:90) +a/2
−a+2
= −µ a5
(cid:16)
T 2 − a2
12
ν
2
17
(cid:18) 1
(cid:17)
30
(cid:19)
(cid:33)
In addition, recall that Ah = 0.
Using equation (2.16), we obtain the energy function of the 1d model as
(cid:90) L
0
Φ(cid:63)[h] =
(cid:90) L
0
h(cid:48)2
2 dS,
(4.8)
1
2
(Y a h2
1 + Y I h2
2) dS + Y a5 12 + 11 ν
720
[h2 h(cid:48)
0 − 1
2]L
2
Y a5 6 + 5 ν
360
where h1 = U(cid:48) denotes the axial stretch and h2 = V (cid:48)(cid:48) denotes the curvature.
4.7. Comments
2 z · B(2)
As the operator 1
We have recovered the classical Euler-Bernoulli rod model at the dominant order, with a stretching
modulus Y a and a bending modulus Y I. Indeed, the first step of the reduction method yields classical
structural model (i.e., without the gradient effect), see the expression of Whom in equation (4.5).
h · z in equation (4.6) is non-negative, the necessary condition of stability with
respect to the microscopic degrees of freedom is satisfied, see §2.7. However, the coefficient of the gradient
term h(cid:48)2
2 in (4.8) is negative, like the second-gradient modulus Bh acting on the macroscopic degrees of
freedom, see (4.7). As a result, Φ(cid:63)[h] can be decreased without bound by means of small-scale oscillations.
This behavior can likely be regularized by pushing the expansion to a higher order. In its present form, the
functional (4.8) should not be used to set up a minimization or to analyze stability; it is still useful, as its
stationary points provide a more accurate approximation of the 3d solution than that of the Euler-Bernoulli
model.
To connect with the existing literature, we have worked out this example in the limited context of
linear elasticity but the derivation can be extended easily to deal, e.g., with a nonlinear constitutive law,
a nonlinear geometry, as demonstrated in the following section. While previous work has been focused on
deriving order by order solutions to the 3d equilibrium equations, our relaxation method makes use of the
variational structure of these equations by working directly on the energy. This, together with the fact that
the hard work underlying the derivation of the general method in section 2 and Appendix A has been done
once for all, simplifies the reduction of any particular structural model considerably.
The only gradient effect present in equation (4.8) comes from the gradient of bending. The absence
of a gradient term for stretching is a peculiarity of the linear elasticity model which we started from: a
gradient effect involving the axial strain h1 is restored if we start instead from a finite-elasticity theory, as
the next example will show. In fact, it does not make much sense to derive a higher-order rod model , which
aims at identifying subdominant corrections to the elastic energy, starting from a linear elasticity model :
the linearization underlying the linear elasticity theory suppresses subdominant contributions to the strain
energy a priori. This fact has not been well appreciated, as most of the earlier work on higher-order beam
models has been done in the framework of linear elasticity.
5. Application to a hyperelastic cylinder in tension
In our third example, we address the axisymmetric deformation of a hyperelastic cylinder. This problem
is motivated by the necking of bars, a situation where deformations become localized. Necking typically
involves plasticity but it can be analyzed using an equivalent elastic constitutive law obtained by the J2
deformation theory, as long as the loading is proportional and monotonous: the equivalent hyperelastic
law is such that the curve for homogeneous traction displays a maximum of the force as a function of the
stretch, leading to localization. This section derives the 1d model obtained by Audoly and Hutchinson
(2016) using a dedicated expansion method, this time using the general method of section 2. This worked
example combines a continuous cross-section, a nonlinear elastic model, and kinematic constraints.
5.1. Finite-strain elasticity model for an axisymmetric bar
In its reference configuration, the bar is a cylinder with length L and radius ρ, and we denote by S, T
and Θ the axial, radial and azimuthal coordinates, respectively, with 0 ≤ S ≤ L, 0 ≤ T ≤ ρ, 0 ≤ Θ ≤ 2 π.
We consider a transversely isotropic material whose elastic properties are functions of T but not of Θ or
18
Figure 6: A nonlinearly elastic cylinder in (a) reference configuration and (b) current configuration.
S (isotropy is a particular case of transverse isotropy, so this includes homogeneous isotropic materials):
this warrants that axisymmetric solutions possessing cylindrical invariance exist when the bar is subject to
traction. In fact, we restrict attention to axisymmetric solutions, ignoring the possibility of localized modes
involving shear bands (Triantafyllidis et al., 2007).
The coordinates (S, T ) are used as Lagrangian coordinates, and we denote by x(S, T ) = (Z(S, T ), R(S, T ))
the axial and radial cylindrical coordinates of a material point initially located as (S, T ), see figure 6(b). In
3d space, the final position is Z(S, T ) dS(Θ) + R(S, T ) dT , where (dS(Θ), dT , dΘ(Θ)) is the local cylindrical
basis, as sketched in the figure. Denoting partial derivatives using commas in subscript, the deformation
T dΘ ⊗ dΘ, and the strain writes
gradient is F = Z,S dS ⊗ dS + Z,T dS ⊗ dT + R,S dT ⊗ dS + R,T dT ⊗ dT + R
E =(cid:0) E1 E2 E3 E4
(cid:1) where
,S + R2
,T + R2
E1 = 1
E2 = 1
E3 = 1
E4 = 1
2
2 (Z 2
2 (Z 2
2 (Z,S Z,T + R,S R,T )
,S − 1)
,T − 1)
(cid:17)
(cid:1)2 − 1
(cid:1) = (cid:0) ESS ET T EST EΘΘ
(cid:1) are the components of the Green -- St-Venant
(cid:16)(cid:0) R
(5.1)
T
.
2 (F T · F − I) from finite-elasticity theory (symbols bearing a bar on top are relevant to the full
Here, (cid:0) E1 E2 E3 E4
(cid:90) ρ
form
strain E = 1
model) and I is the identity matrix.
(cid:18) ds
dS
(cid:19)
(S)
.
h(S) = (h1(S)) =
19
For a transversely isotropic material, the strain energy of the bar per unit length can be written in the
W (E) =
W (E1(S), E2
3 (S), E2(S) + E4(S), E2
2 (S) + E2
4 (S), E2(S) E2
3 (S)) 2 π T dT,
(5.2)
0
(cid:82) ρ
where W (ESS, E
written in terms of a set of invariants relevant to the transverse isotropic symmetry.
2
ST , ET T + EΘΘ, E
2
T T + E
2
ΘΘ, ET T E
2
ST ) is the elastic potential of the material model,
5.2. Macroscopic and microscopic variables
Let us consider the coordinate s(S) of the center of mass of the deformed cross-section, i.e.,
s(S) = (cid:104)Z(cid:105)(S),
(5.3)
where (cid:104)f(cid:105) = 1
0 f (T ) 2 π T dT denotes the average of a quantity over the cross-section. This s(S) is
denoted by the orange dot in figure 6(b); it does not correspond to any material point. We use a single
macroscopic strain variable, defined as the apparent axial stretch
π ρ2
(a)(b) We use as microscopic degrees of freedom y the position of the current relative to the center of mass of
the cross-section,
y = ({y1}T ,{y2}T ),
such that the position in deformed configuration can be reconstructed as
x(S, T ) = (Z(S, T ), R(S, T )) = (s(S) + y1(S, T ), y2(S, T )).
In view of equation (5.3), one must enforce the kinematic constraint q(y(S)) = 0 for all S, where
(cid:18)(cid:90) ρ
0
(cid:19)
q(y) =
y1(T ) 2 π T dT
.
In terms of the macroscopic strain and microscopic displacement, the map of strain over a cross section
writes, from equation (5.1),
†
E(h, h
; y, y†, y‡) =
2 ((h1 + y
(cid:110) 1
(cid:111)
(cid:8) 1
2 ((∂T y1)2 + (∂T y2)2 − 1)(cid:9)
†
2)2 − 1)
(cid:110) 1
(cid:111)
(cid:17)(cid:111)
(cid:110) 1
†
1) ∂T y1 + y
(cid:16)(cid:0) y2
†
1)2 + (y
†
2 ∂T y2)
2 ((h1 + y
(cid:1)2 − 1
T
T
2
T
T
T
.
(5.4)
As earlier with the elastic block, and as implied by the {. . .}T notation, each component of E(S) is a function
defined over the cross-section.
5.3. Homogeneous solutions
A detailed analysis of homogeneous solutions is done in Appendix B.1. The main results are summarized
as follows.
At the microscopic level, the cross-sections remains planar and undergo a uniform dilation with a stretch
ratio µ(h1) depending on the (uniform) longitudinal stretch h1, i.e., the microscopic displacement is of the
form
1 = {0}T
y(h1)
2 = {µ(h1) T}T .
y(h1)
Due to the material symmetry, the microscopic stress is equi-biaxial, with a longitudinal stress Σ(cid:107)(h1, µ(h1)) =
ΣSS and a transverse stress Σ⊥(h1, µ(h1)) = ΣT T = ΣΘΘ given in terms of the elastic constitutive model by
Σ(cid:107)(h1, µ) = ∂1W(cid:0) 1
Σ⊥(h1, µ) = ∂3W(cid:0) 1
1 − 1), 0, µ2 − 1, 1
1 − 1), 0, µ2 − 1, 1
2 (h2
2 (h2
2 (µ2 − 1)2, 0(cid:1)
2 (µ2 − 1)2, 0(cid:1) + (µ2 − 1) ∂4W(cid:0) 1
1 − 1), 0, µ2 − 1, 1
2 (h2
2 (µ2 − 1)2, 0(cid:1) .
(5.5)
Here, ∂iW denotes the partial derivative of the strain energy with respect to the ith argument.
The equilibrium of the lateral boundary yields an implicit equation for the transverse stretch µ(h1) in
terms of the longitudinal stretch,
Σ⊥(h1, µ(h1)) = 0.
(5.6)
The homogeneous microscopic strain, strain energy density, microscopic stress and tangent elastic stiff-
(cid:110) 1
ness are then given by equation (2.7) as
Eh =
(cid:16) (cid:8) 1
1 − 1)(cid:9)
(cid:16) 1
Whom(h) = (cid:82) ρ
Σh · δE = (cid:82) ρ
4 K
δE · Kh · δE = (cid:82) ρ
2 (h2
0
2 (h2
(h1) − 1)
2 (µ2
(h1) − 1, 1
0 W
0 Σ(cid:107)(h1, µ(h1)) δE1(T ) 2 π T dT
1 − 1), 0, µ2
T
T
{0}T
(cid:110) 1
(cid:111)
(cid:17)
(h1) − 1)2, 0
δE1(T )
∗
·
2 (µ2
δE2(T )
δE4(T )
ST
ST (δE3(T ))2 +
(h1) − 1)
2 (µ2
2 π T dT
∗
∗
∗ K
T T
T T K
∗ K
ΘΘ
T T K
ΘΘ
T T
T T
T T
T
(cid:111)
·
(cid:17)
δE1(T )
δE2(T )
δE4(T )
2 π T dT
20
where the incremental shearing modulus reads
K
ST
ST =
1
4
(2 ∂2W + (µ2
(h1) − 1) ∂5W ).
ST ) =(cid:0) 1
2
(5.7)
2 (µ2 − 1)2, 0(cid:1), as in equa-
In the right-hand side, the derivatives of the elastic potential W must be evaluated in the homogeneous
solution (ESS, E
tion (5.5). The expressions of the other elastic moduli do not play any role in the 1d model.
1 − 1), 0, µ2 − 1, 1
2
ST , ET T + EΘΘ, E
2
ΘΘ, ET T E
2
T T + E
2 (h2
The symmetry of the material and of the homogeneous solution accounts for the particular form of the
T T
T T in factor of (δE4)2 = (δEΘΘ)2 in the
tangent moduli found above. For instance, the tangent modulus K
second variation of the elastic potential δE · Kh · δE is identical to that in factor of (δE2)2 = (δET T )2 due
to the transverse isotropy, i.e., K
ΘΘ
ΘΘ = K
T T
T T .
5.4. Structure coefficients, local optimization problem
Using the homogeneous solution just obtained and the expression of the strain E in equation (5.4),
one can calculate the structure coefficients relevant to the nonlinear cylinder as (details can be found in
Appendix B.2)
2
0 0
(cid:16)
000(h) · h
†
e10
=
000(h) · h
† · e20
†
h
† · e10
e01
000(h) = 0 h
100(h) · (z1, z2) =
e00
(cid:111)
†
T h
1
†
1)2}T
(cid:18)
0 {µ(h1) ∂T z2(T )}T
(z1, z2) · e00
200(h) · (z1, z2) =
†
010(h) · (z
†
1(T )}T
e00
1, z
(cid:110) µ(h1 ) ∇µ(h1)
(cid:17)
0 0 0 (cid:1)
=(cid:0) {(∇µ(h1))2 T 2 (h
100(h) · (z1, z2) =(cid:0) 0 0 ∗ 0 (cid:1)
(cid:16)
(cid:8) h1
2 ∂T z1(T )(cid:9)
0 ∗ 0 (cid:1) .
2) =(cid:0) {h1 z
0 {(∂T z1(T ))2 + (∂T z2(T ))2}T
0
†
T
T
(cid:110)
0
(cid:111)
(cid:17)
(cid:26)(cid:16) z2(T )
(cid:17)2(cid:27)
z2(T )
T
T
T
T
µ(h1)
This yields the first set of operators as
(cid:19)
(5.8)
(5.9)
Ah = 0
C(0)
h = 0
h · z = Σ(cid:107)(h1, µ(h1)) h1
C(1)
†
h · h
† · B(0)
†
1)2 2 π ρ4
= 1
1
2 (h
2 h
†
† · B(1)
h · z = h
K
h
1
†
−h
1
0 h2
+ 1
2
(cid:82) ρ
(cid:82) ρ
(∇µ(h1))2 (µ2
(cid:82) ρ
ST µ(h1) ∇µ(h1) h1
d(Σ(cid:107)(h1,µ(h1)) h1)
(cid:19)
(cid:82) ρ
h · z = 1
(cid:16)
(cid:82) ρ
2 z · B(2)
ST
ST (∂T z1(T ))2 2 π T dT
· Q(h1) ·
1 K
ST
2
1
4
0
0 z1(T ) 2 π T dT
ST
ST + Σ(cid:107)(h1, µ(h1)))
(h1) K
0 T ∂T z1(T ) 2 π T dT
(cid:17)
0 z1(T ) 2 π T dT
dh1
(cid:18) ∂T z2(T )
(cid:32)
z2(T )
T
(cid:19)
2 π T dT
(cid:18) ∂T z2(T )
(cid:33)
z2(T )
T
Q(h1) = µ2
(h1)
K
K
T T
T T K
ΘΘ
T T K
ΘΘ
T T
T T
T T
where
and we recall that ∇ is a gradient with the respect to the macroscopic strain h, i.e.,
This quantity ∇µ(h1) can be found by differentiating the implicit equation (5.6).
∇µ(h1) =
dµ(h)
dh
(h1).
21
The local correction z = zopt is obtained by making the quantity 1
h · z stationary for
, subject to the constraint q(z) = 0. This variational problem is solved in Appendix
†· B(1)
(S) announced in equation (2.15) with
†
given values of h and h
B.3. The solution can be cast in the form zopt(S) = Zh(S)
opt
µ(h1) ∇µ(h1 )
(cid:32) (cid:110)− 1
· h
(cid:48)
(cid:16)
2
h1
T 2 − ρ2
2
Zh(S)
opt =
(cid:8)0(cid:9)
T
2 z· B(2)
h · z + h
(cid:17)(cid:111)
(cid:33)
(cid:16)
(cid:17)
2
Stated differently, the corrective displacement is purely longitudinal (zopt
− 1
µ(h1) ∇µ(h1)
T 2 − ρ2
h(cid:48)
1(S).
2
h1
2 = 0), and parabolic: zopt
1 =
The following integral, which is required later on, can be calculated based on the expressions of zopt
just
1
found as
T
.
(5.10)
A necessary condition for the stability of this microscopic solution (see §2.7) is that K
(cid:18) µ(h1) ∇µ(h1)
(cid:19)2
h1
h(cid:48)2
1 .
(T )(cid:1)2
(cid:90) ρ
0
(cid:0)∂T zopt
(cid:32)
1
2 π T dT =
π ρ4
2
(cid:33)
K
K
ΘΘ
ΘΘ K
ΘΘ
T T K
ΘΘ
T T
T T
T T
is positive.
(5.11)
ST
ST ≥ 0 and the
submatrix of the tangent moduli
5.5. Regularized model
We can finally calculate the two operators entering in the 1d model as
(cid:48) · Bh · h
(cid:48)
1
2 h
= 1
2 h
2 h(cid:48)2
= 1
= h(cid:48)2
1
2
4
1
π ρ4
2
(cid:48) · B(0)
h · h
2 π ρ4
(cid:48) − 1
2 zopt · B(2)
h · zopt
(∇µ(h1))2 (Σ(cid:107)(h1, µ(h1)) + µ2
(∇µ(h1))2 Σ(cid:107)(h1, µ(h1))
(h1) K
ST
ST ) − 1
2 h2
1 K
(cid:82) ρ
0 (∂T z1(T ))2 2 π T dT
ST
ST
where we have used the equilibrium condition (5.6) and the identity (5.11).
The operator Ch reads
Ch · h
(cid:48)
= C(0)
h · h
(cid:48)
+ C(1)
h · zh
opt = Σ(cid:107)(h1, µ(h1)) h1
(cid:90) ρ
0
zopt
1
(T ) 2 π T dT = 0.
Switching to the more familiar notation λ(S) for the apparent stretch λ = h1, we finally obtain the 1d
energy governing the cylinder as
(cid:90) L
(cid:90) L
0
1
2
(cid:17)
where Whom(λ) =(cid:82) ρ
Φ(cid:63)[λ] =
(cid:16) 1
2 (λ2 − 1), 0, µ2
0
(λ) − 1)2, 0
(cid:19)2
solution per unit length, the strain-gradient modulus is given by
(λ) − 1, 1
2 (µ2
0 W
(cid:18) dµ(λ)
π ρ4
Whom(λ(S)) dS +
B(λ(S)) λ(cid:48)2(S) dS,
(5.12a)
2 π T dT is the energy of the homogeneous
B(λ) =
2
dλ
Σ(cid:107)(λ, µ(λ)),
(5.12b)
and the transverse stretch µ(λ) is found by solving the transverse equilibrium of a homogeneous solution,
Σ⊥(λ, µ(λ)) = 0.
22
5.6. Comments
We have recovered in equation (5.12) the energy functional derived by Audoly and Hutchinson (2016).
As earlier with the membrane model, see §3, the strain-gradient modulus B(λ) is directly proportional to
the pre-stress Σ(cid:107) of the homogeneous solution, and does not depend on the elastic moduli. This can be
explained as follows. The expression of the operators B(i)
h in equation (2.12), reveal that the contribution to
2 e[1]·Kh·e[1],
the strain-gradient modulus coming from the elastic moduli arises fully from the expansion of 1
100 · z is the first-order correction to the strain, see also equation (A.6). If the
where e[1] = e10
correction z manages to cancel out entirely the strain e10
arising from the gradient effect, then e[1] = 0
and the strain-gradient modulus arises from the pre-strain Σh only. This is what happens with both the
axisymmetric membrane, for which e10
= 0 from equation (3.3), and for the axisymmetric cylinder, for
which e10
is a shear strain that is canceled out by the out of plane deformation of the cross-section, as
discussed by Audoly and Hutchinson (2016). For the bending of an elastic block, however, the corrective
displacement z does not fully suppresses the first order strain, e[1] (cid:54)= 0, and the elastic moduli enter into
the expression of the strain-gradient modulus.
000 · h
(cid:48)
000 · h
(cid:48)
000 · h
(cid:48)
+ e00
000 · h
(cid:48)
6. Conclusion and discussion
We have presented a systematic reduction method which, given a structural model representing a pris-
matic elastic solid, yields a 1d model that captures the strain gradient effect. The method implements a
two-scale expansion and is asymptotically exact.
It is based on a choice of macroscopic strain variables
which are retained in the 1d model, and a choice of microscopic variables which are relaxed during the
reduction process. It can be applied as a simple recipe, i.e., it requires one to follow a systematic sequence
of steps in order. The method retains the nonlinearity of the initial model, and can account for large and
inhomogeneous changes in the shapes of cross-sections.
As illustrated by the worked examples, this method can be used to recover known 1d models for structures
in a systematic and unified way. In future work, it will be used to derive original 1d models, e.g., for structures
possessing highly deformable cross-sections, such as tape springs, or having large and inhomogeneous pre-
stress (Liu et al., 2014; Lestringant and Audoly, 2017). The method can also be applied to the analysis of
localization, which is ubiquitous in slender structures. In the absence of 1d models capturing the gradient
effect, the various examples of localization have been addressed using equations that are specific to each
particular structure, such as the axisymmetric membrane theory for bulges in balloons (Fu et al., 2008;
Pearce and Fu, 2010); the authors have suggested recently that, by using dimension reduction, it is possible
to analyze the various examples of localization in a unified mathematical framework (Lestringant and Audoly,
2018).
T
We close this paper with a few general remarks.
In the reduction method, the choice of the macroscopic strain h has been left to the user. This choice
actually reflects how the applied load scales with the slenderness parameter (note that the external load
has not appeared in our reduction reduction method). The scaling assumptions for the load are a key
ingredient in dimension reduction, and different assumptions can lead to different 1d model (Marigo and
Meunier, 2006). Similarly, different choices of h ultimately lead to different 1d models using our method.
Consider for instance what happens if we add a 'strong' external shearing force g on the linear elastic block,
a3/12 ex. This external load induces a moment G(S) ez perpendicular to the plane of the
g(S, T ) = G(S)
block, in each cross-section. The kinematic quantity conjugate to g is the average rotation of the cross-
(cid:104)T u(cid:105)
a2/12 + V (cid:48). This external load can therefore
section, which in the linear setting reads
be handled by including an additional kinematic constraint (cid:104)T u(cid:105)(S) = ϕ(S) in the reduction method, and
by augmenting macroscopic strain h with this new internal variable, h3 = ϕ. The reduction method has
to be redone, yielding this time a Timoshenko beam model. The external load is then taken into account
a2/12 + V (cid:48)(S). Note that this modification of the reduction
simply by coupling the load intensity G(S) with ϕ(S)
procedure is required if, and only if, the external load is strong and inhomogeneous enough that it modifies
significantly the natural microscopic displacement. For mild applied force, the external moment G(S) can be
coupled directly to the macroscopic rotation V (cid:48), saving one from the need to amend the dimension reduction.
a2/12(cid:104)y · (T ex)(cid:105) =
1
23
In their analysis of a 1d model for stretched bars, Coleman and Newman (1988) have proposed a derivation
of the gradient effect starting from the assumption that the kinematics of the classical bar model without
gradient effect remains valid, i.e., that the cross-sections remain planar. A similar approximation has been
In our notation, this amounts
used by several authors in various derivations of strain gradient models.
to neglecting the microscopic correction z arising from the gradient effect, i.e., to set Zh(S)
opt = 0, see
equation (2.15). Equation (2.17) then yields Bh = B(0)
h . The inequality (2.7) shows that the strain gradient
modulus derived from this kinematic assumption has a larger strain-gradient elastic modulus B(0)
h than the
true modulus predicted by our method with due account for the microscopic relaxation z, as noted by Audoly
and Hutchinson (2016). This is not surprising, as our approach relaxes the strain energy optimally by design,
while the former does not.
For the sake of simplicity, we have assumed that the full model is invariant along the longitudinal
direction. It is quite simple to extend the method to the case where the geometry and/or elastic properties
of the full model vary slowly in the longitudinal direction, i.e., depend on the stretched variable S used
in Appendix A. An additional explicit dependence on S must then be added to the various quantities
entering into the analysis, such as W , Eh, Whom, Σh, etc. This brings in a single significant change:
· h
in the integrand of Φ[2] in equation (A.8), one needs to include two extra terms −
and −
· z[1] in order to cancel out the extra terms coming from the expansion of the total
dC(0)
h
d S
h=h( S)
(cid:20)
(cid:21)
(cid:20)
(cid:21)
( S)
dC(1)
h
d S
( S)
h=h( S)
( S))
derivatives
terms make their way into the final expression of the 1d energy.
( S))
·z[1], when the C(j)
· h+
d(C(1)
h( S)
d S
d(C(0)
h( S)
d S
h ( S)'s depend explicitly on S. These two additional
The illustration examples in sections 3 -- 5 were simple enough that they could be solved analytically. When
this is not possible, the proposed reduction method lends itself naturally to a numerical implementation: by
solving numerically a series of elasticity problem over the cross-section, it is possible to build a numerical
representation of the various operators Ah, Bh and Ch, i.e., to evaluate numerically the coefficients of the
1d energy Φ(cid:63).
This paper was prepared using TeXmacs (van der Hoeven et al., 2013), an outstanding and freely available
scientific text editor.
Appendix A. Proof of the main results
In this appendix, we offer a detailed justification of the results announced in §2.
Appendix A.1. Change of microscopic variable
We return to the relaxation problem (2.4) for a prescribed, non-homogeneous distribution of macroscopic
strain {h(S)}S, and seek the microscopic displacement achieving the optimum of the functional Φ[h, y]
in equation (2.2) subject to the constraint q(y(S)) in equation (2.3). We seek y(S) in the form (2.8),
y(S) = yh(S) + z(S). Since yh(S) satisfies the linear constraint by construction, the new unknown z(S)
must satisfy the constraint as well, ∀S q(z(S)) = 0.
In terms of z(S), the gradients of the original microscopic displacement write
(S) · ∇yh(S) + z(cid:48)(S)
(cid:48)
y(cid:48)(S) = h
(S) · ∇yh(S) + h
(cid:48)
(cid:48)(cid:48)
y(cid:48)(cid:48)(S) = h
(S) · ∇yh(S) · h
(cid:48)
(S) + z(cid:48)(cid:48)(S)
where ∇ denotes the gradient with respect to the macroscopic parameter h, see equation (2.10). In view
of this, the strain appearing in the strain energy in equation (2.2) can be expressed in terms of the new
unknown as
†
where eh(h
‡
, h
(S); z(S), z(cid:48)(S), z(cid:48)(cid:48)(S)),
; z, z†, z‡) is the strain function introduced in equation (2.9).
(cid:48)
E = eh(S)(h
(cid:48)(cid:48)
(S), h
24
Note that the principle of virtual work for the homogeneous solution (2.6) can be rewritten in hindsight
using the structure coefficient e00
100 as
∀y − Σh · (e00
q(yh) = 0.
100(h) · y) + f h · q(y) = 0
Appendix A.2. Expansion method
(A.1)
Given the distribution of macroscopic strain h(S), the strain energy of the full model is expressed in
terms of the corrective displacement z(S) as
(cid:90) L
0
Φ[h, yh + z] =
(cid:48)
W (eh(S)(h
(cid:48)(cid:48)
(S), h
(S); z(S), z(cid:48)(S), z(cid:48)(cid:48)(S))) dS.
(cid:90) L
This energy must be relaxed with respect to z, subject to the constraint
∀S q(z(S)) = 0.
This relaxation problem is treated by an asymptotic method, which assumes that the prescribed strain
h(S) is a slowly varying function of S. Accordingly, we introduce a stretched variable S = γ S where γ (cid:28) 1
is a small parameter. We denote by a dot the derivation with respect to the new variable,
f =
df
d S
.
The assumption of slow axial variations is implemented by requiring that the dependence of the various
functions on S encountered so far is replaced with a dependence on the slow variable S. This amounts to
assuming f = O(1), f = O(1), etc. while the original derivatives scale as
f(cid:48) = γ f = O(γ)
f(cid:48)(cid:48) = γ2 f = O(γ2)
etc.
We seek the microscopic correction as an expansion
z( S) = γ z[1]( S) + γ2 z[2]( S) + ···
(A.2)
where the absence of any term of order γ0 = 1 follows from the observation that in the homogeneous case,
corresponding to the formal limit γ → 0, we have z(S) = 0.
The strain energy to be relaxed can be rescaled as Φ = Φ/γ, where
Φ[h, z] =
W (eh(γ h( S), γ2 h( S); γ z[1]( S) + γ2 z[2]( S) + ··· , γ2 z[1]( S) + ··· , 0)) d S + O(γ3),
(A.3)
0
and L = γ L, subject to the constraint
∀i ≥ 1 ∀ S q(z[i]( S)) = 0.
(A.4)
We now proceed to derive an expansion of the energy Φ in powers of γ, and to solve this problem order by
order.
Appendix A.3. Strain expansion
The argument eh of W in (A.3) is the strain. It can be expanded as
eh(γ h( S), γ2 h( S); γ z[1]( S) + γ2 z[2]( S), γ2 z[1]( S), 0) = Eh( S) + γ e[1]( S) + γ2 e[2]( S) + O(γ3)
(A.5)
where term of order γ0 is the homogeneous strain Eh( S) = E(h( S), 0; yh( S), 0, 0) defined in equation (2.7),
and the linear and quadratic corrections can be written in terms of the structure coefficients as
e[1]( S) = e10
e[2]( S) = 1
000(h( S)) · h( S) + e00
2 ( h · e20
000(h) · h + 2 h · e10
100(h( S)) · z[1]( S)
000(h) · h + e00
100(h) · z[1] + z[1] · e00
010(h) · z[1] + e00
200(h) · z[1])···
100(h) · z[2].
+e01
(A.6)
All the quantities appearing in the right-hand side of e[2]( S) must be evaluated at S, like those in the
right-hand side of e[2]( S).
25
Appendix A.4. Energy expansion
The strain expansion (A.5) can be inserted into the energy in (A.3)
Φ[h, z] =
(W (eh( S) + γ e[1]( S) + γ2 e[2]( S)) + O(γ3)) d S
This yields an expansion of the energy as
Φ[h, z] = Φ[0][h] + γ Φ[1][h] + γ2 Φ[2][h, z[1]] + ···
(A.7)
where the first orders in the expansion read
Φ[0][h] = (cid:82) L
Φ[1][h] = (cid:82) L
Φ[2][h, z[1]] = (cid:82) L
0 Whom(h( S)) d S
(cid:17)
(cid:16) 1
0 Σh( S) · e[1]( S) d S
2 e[1]( S) · Kh( S) · e[1]( S) + Σh( S) · e[2]( S)
0
d S.
Our notation anticipates on the fact that Φ[1] does not depend on z, and that Φ[2] depends on z through
its dominant contribution z[1] only, as we prove in the next section.
Appendix A.5. Rearranging the energy contributions
Using the operator Ah introduced in equation (2.12), the energy contribution Φ[1] can be written as
Φ[1][h] =
(Ah( S) · h( S) + Σh( S) · (e00
100(h( S)) · z[1]( S))) d S.
(cid:90) L
0
(cid:90) L
0
The second term in the integrand can be simplified by using the principle of virtual work for homogeneous
100(h( S)) · yh( S)) = f h( S) ·
solutions (A.1): setting the virtual motion as y = yh( S), one has Σh( S) · (e00
q(yh( S)) = 0 since yh( S) satisfies the constraint q. This shows that the second term in the integrand above
vanishes, and that Φ[1][h] is actually independent of the microscopic displacement z, as anticipated in our
notation,
(cid:90) L
0
Φ[1][h] =
Ah( S) · h( S) d S.
Most structures are invariant by the reflection S ← (−S), which implies that the operator Ah, as well as
the first-order correction to the energy Φ[1][h], vanish. This makes it important to determine the expansion
of the energy to second order.
(cid:19)
e[1]( S) · Kh( S) · e[1]( S) + Σh( S) · e[2]( S)
d S
At order γ2, we have
Φ[2][h, z[1]] =
0
2
(cid:18) 1
(cid:90) L
C(0)
(cid:90) L
0
Φ[2][h, z[1]] =
Inserting the expression of e[1] and e[2] in the integrand, and using the operators defined in (2.12), we can
rewrite this as
h · h +
h · B(0)
+ 1
+Σh · (e00
2
)
d(C(0)
h( S)
d S
h · h + h · B(1)
100(h) · z[2])
· h + C(1)
h · z[1] +
h · z[1] + 1
d(C(1)
h( S)
d S
2 z[1] · B(2)
)
· z[1] . . .
h · z[1] ···
d S
(A.8)
where all quantities in the integrand must evaluated at S, namely h = h( S),
z[1] = z[1]( S). In this expression, we have made appear the term
)
d(C(0)
h( S)
d S
26
h = h( S), z[i] = z[i]( S) and
· h( S),
· h( S) = h( S) · ∇C(0)
h( S)
which cancels out with the last term introduced in the definition of 1
2
same holds for
)
d(C(1)
h( S)
d S
· z[1] and the last term in h · B(1)
h · z[1].
In addition, the last term in the integrand of Φ[2][h, z[1]] cancels by the same argument as earlier, i.e.,
using the homogeneous principle of virtual work (A.1) and the constraint (A.4) at second order. As a result,
Φ[2] depends on z[1] but not on z[2], as anticipated in our notation. Noting that the terms on the first line
in the integrand form an exact derivative, and can be integrated, we have
h · B(0)
h · h, see equation (2.12). The
(cid:90) L
(cid:18) 1
0
2
(cid:19)
(cid:19)
Φ[2][h, z[1]] = [C0
h · h + C1
h · z[1]] L
0 +
h · B(0)
h · h + h · B(1)
h · z[1] +
z[1] · B(2)
h · z[1]
1
2
d S
(A.9)
where all the quantities in the integrand side are implicitly evaluated at S in our notation.
The expansion of the energy Φ[h, yh + z] in non-scaled form, announced earlier in equation (2.13), is
readily obtained by inserting into equation (A.7) the expressions of Φ[0], Φ[1] and Φ[2] just derived, and
(cid:48)
, z(cid:48) and z(cid:48)(cid:48), as well as the unscaled longitudinal coordinate
restoring the original (scaled) gradients h
S.
(cid:48)(cid:48)
, h
Appendix A.6. Dominant order correction found by a local problem
We now proceed to minimize the energy (A.3) under the constraint (A.4) order by order, using the
expansion of Φ[h, z] just found. Thanks to the integration by parts, the dependence of the strain energy
Φ[2][h, z[1]] on the axial gradient z(cid:48)
[1] of the corrective displacement has been removed. Therefore, in every
cross-section S, z[1]( S) is the solution of a local optimization problem: z[1]( S) makes stationary the quantity
h · z[1] among all z's satisfying the constraint q(z) = 0. This leads to the variational
h · z[1] + 1
B(1)
problem stated in equation (2.14), where we use the notation zopt = γ z[1] for the dominant contribution to
the correction z, see equation (A.2), and f opt = γ f [1] for the scaled Lagrange multiplier. In terms of z[1], the
variational problem can be written equivalently as ∀z
· z− f [1]( S)· q(z) = 0
and q(z[1]( S)) = 0.
· z + z[1]( S)· B(2)
h( S)· B(1)
h( S)
2 z[1] · B(2)
h( S)
Appendix A.7. Relaxed energy
By using the particular virtual motion z = z[1]( S) in the above variational problem, we obtain the
identity
When the optimal displacement zopt
[1]
we obtain the second-order contribution to the relaxed energy as
· z[1]( S) = −z[1]( S) · B(2)
h( S) · B(1)
h( S)
is inserted into Φ[2][h, z[1]] in equation (A.9), and this identity is used,
· z[1]( S).
h( S)
Φ(cid:63)
[2][h] = Φ[2][h, zopt
[1] ] = [C0
h · h + C1
h · zopt
[1] ] L
0 +
h · B(0)
h · h − 1
2
z[1] · B(2)
h · zopt
[1]
d S.
This yields the final expressions (2.16 -- 2.17) of the relaxed model.
Appendix B. Detailed calculations for the nonlinear cylinder
Appendix B.1. Analysis of homogeneous solutions
In this section, we provide a detailed analysis of the homogeneous solutions for the hyperelastic cylinder,
with the aim to justify the main results announced in section 5.3.
The map of strain for homogeneous solutions writes, from equation (2.5) and (5.1),
E(h, y) =
1 − 1)(cid:9)
(cid:8) 1
(cid:8) 1
2 ((∂T y1)2 + (∂T y2)2 − 1)(cid:9)
(cid:8) 1
(cid:9)
(cid:16)(cid:0) y2
(cid:110) 1
(cid:17)(cid:111)
(cid:1)2 − 1
2 h1 ∂T y1
2 (h2
T
T
2
T
T
T
.
27
(cid:90) L
(cid:18) 1
0
2
In view of the material symmetries, we seek equi-biaxial solutions in the particular form y1 = {0}T and
y2 = {µ T}T . Then,
E(h, y) = E(h, ({0}T ,{µ T}T )) =
(h, ({0}T ,{µ T}T ))·y =
(cid:8) 1
(cid:9)
{0}T
{µ ∂T y2}T
2 h1 ∂T y1
(cid:110) µ y2
(cid:111)
T
T
T
.
∂ E
∂y
T
T
T
2 (h2
{0}T
(cid:8) 1
1 − 1)(cid:9)
(cid:8) 1
2 (µ2 − 1)(cid:9)
2 (µ2 − 1)(cid:9)
(cid:8) 1
·
ΣSS
ΣT T
2 ΣST
ΣΘΘ
δE1(T )
δE2(T )
δE3(T )
δE4(T )
2 π T dT
For this type of strain, the material symmetry warrants ΣST = 0 and ΣT T = ΣΘΘ. We denote the
longitudinal stress as Σ(cid:107)(h1, µ) = ΣSS and the isotropic transverse stress as Σ⊥(h1, µ) = ΣT T = ΣΘΘ.
Then, the first variation of the density of strain energy reads
dW
0
dE ( E(h, y)) · δE = (cid:82) ρ
= (cid:82) ρ
(cid:20)
2 = {µh T}T as
(cid:90) ρ
0 (Σ(cid:107)(h1, µ) δE1(T ) + Σ⊥(h1, µ) (δE2(T ) + δE4(T ))) 2 π T dT.
By combining these equations with equation (2.6), we obtain a principle of virtual work for the homogeneous
radial displacement yh
∀y1, y2
1 y1(T ) − µ Σ⊥(h1, µh)
f h
0
+ ∂T y2(T )
2 π T dT = 0
(cid:18) y2(T )
T
(cid:19)(cid:21)
(cid:82) ρ
It can be seen that the Lagrange multiplier is zero, f (h1)
= 0, which is a consequence of the fact that our
trial function satisfies the constraint already. The other term in the integrand above can be rewritten as
0 [−2 π µ Σ⊥(h1, µh) ∂T (T y2)] dT = −2 π µ Σ⊥(h1, µh)ρ y2(ρ) and so the principle of virtual work yields the
equation Σ⊥(h1, µ(h1)) = 0 as announced in equation (5.6).
1
We proceed to present a derivation of the stress and the tangent moduli in the homogeneous solution.
The stress is found by identifying the first variation of the elastic potential W (E1(S), E2
3 (S), E2(S) +
E4(S), E2
2 (S) + E2
4 (S), E2(S) E2
3 (S)), namely
δW = ∂1W δE1 +2 E3 ∂2W δE3 +(δE2 +δE4) ∂3W +2 (E2 δE2 +E4 δE4) ∂4W +(E2
3 δE2 +2 E2 E3 δE3) ∂5W .
elasticity, δW = (cid:80)
Identifying the stress components with the definition of Piola-Kirchhoff stress from the general theory of
IJ ΣIJ δEIJ , using of the definition of the EIJ 's in terms of (E1, . . . , E4) (see below
equation (5.1)) and evaluating these stress components in the homogeneous solution yields the equation (5.5).
The tangent moduli are found similarly, by identifying the second variation of the elastic potential
with the definition of the tangent moduli from the general theory of elasticity, δ2W =(cid:80)
δ2W = ∂11W × (δE1)2 + . . .
IJLM K
LM
IJ δEIJ δELM .
This yields the expression of K
ST
ST given in equation (5.7), together with the additional tangent elastic moduli
K
K
T T
ΘΘ
T T = 2 ∂4W + K
T T
ΘΘ
T T = ∂33W + 2 (µ2
(h1) − 1) ∂34W + (µ2
(h1) − 1)2 ∂44W .
The factor 1/4 in equation (5.7) comes from the fact that there are four possible sets of indices (I, J, L, M )
such that δEIJ δELM = (δEST )2.
28
Appendix B.2. Strain function in terms of the microscopic correction
The change of microscopic unknown is carried out by setting y(S) = yh(S) + z(S), that is y1 = z1
and y2 = y(h1)
2 + z2 = µ(h1) T + z2. A strain function eh defined in terms of the new unknown z has been
introduced in equation (2.9). Inserting the expression of E from equation (5.4) relevant to nonlinear elastic
cylinder yields
(cid:110) 1
(cid:110) 1
2 ((h1 + z
(cid:111)
†
†
†
2(T ))2 − 1)
1 ∇µ(h1) T + z
1(T ))2 + (h
(cid:19)(cid:27)
(cid:8) 1
2 ((∂T z1(T ))2 + (µ(h1) + ∂T z2(T ))2 − 1)(cid:9)
(cid:26)
†
†
1 ∇µ(h1) T + z
1(T )) ∂T z1(T ) + (h
T
(cid:18)(cid:16)
(cid:17)2 − 1
1
2
µ(h1) + z2(T )
T
T
†
2(T )) (µ(h1) + ∂T z2(T )))
.
(cid:111)
T
T
†
eh(h
‡
, h
; z, z†, z‡) =
2 ((h1 + z
†
, z, z† and
Calculating the Taylor expansion of eh about a homogeneous solution, i.e., for small h
z‡, we can then identify the coefficients in this expansion with the structure coefficients introduced in
equation (2.11).
‡
, h
Appendix B.3. Corrective displacement
In this section, we solve the local optimization problem for the nonlinear cylinder in traction. We start
from the expressions of the operators B(1)
h does
not couple the unknowns z1(T ) and z2(T ), the optimization problem for z obtained in equation (2.14) splits
into a problem for the axial corrective displacement z1
h obtained in equation (5.9). As the operator B(2)
h and B(2)
(cid:40) ∀z1 h
q(cid:0)zopt
1
h(S) · (z1, 0) +(cid:0)zopt
1
, 0(cid:1) · B(2)
h(S) · (0, z2) +(cid:0)0, zopt
2
(cid:48)
(S) · B(1)
(cid:48)
(S) · B(1)
, 0(cid:1) = 0,
(cid:110) ∀z2 h
and a problem for the radial corrective displacement z2,
(cid:1) · B(2)
h(S) · (0, z2) = 0.
h(S) · (z1, 0) − f opt(S) · q((z1, 0)) = 0
(B.1)
(cid:48)
In the latter, the first term h
(S)·B(1)
h(S)·(0, z2) vanishes in view of the definition of B(1)
h in equation (5.9).
The solution for the corrective radial displacement is therefore zero,
as it is proportional to the constraint q(cid:0)zopt
, 0(cid:1) = (cid:82) ρ
Observe that the second term in the right-hand side of the definition of B(1)
(T ) 2 π T dT .
0 zopt
1
1
h in equation (5.9) is zero,
Inserting the expressions of the
operators B(1)
h and B(2)
h from equation (5.9), we can rewrite the variational problem (B.1) as
∀z1
(cid:48)
1 K
[h1 h
ST
ST µ(h1) ∇µ(h1)] T + [h2
1 K
ST
ST ] ∂T zopt
1
(T )
∂T z1(T ) 2 π T dT ···
2 = {0}T .
zopt
(cid:17)
(cid:90) ρ
(cid:16)
0
(cid:104)− f opt
1
(cid:105)(cid:90) ρ
0
z1 2 π T dT = 0
+
where we have made appear a new Lagrange multiplier f opt
To solve this equation, we rewrite it in compact form as
1 = f opt
d(Σ(cid:107)(h1,µ(h1)) h1)
dh1
.
(cid:90) ρ
0
∀z1
(cid:0)a T + b ∂T zopt
(cid:1) ∂T z1 2 π T dT + d
1
†
1 − h
(cid:90) ρ
1
z1 2 π T dT = 0,
0
29
as
(cid:20)
∀z1
1
T
∂T
2 π T
0
(cid:18)
(cid:19)
(cid:19)(cid:21)
1 − d
T
2
a T + b ∂T zopt
where the coefficients a, b, d are identified with the square brackets in the equation above. An integration
by parts yields
(cid:90) ρ
(cid:2)2 π T (cid:0)a T + b ∂T zopt
(cid:1) z1
(cid:3)ρ
0 −
(cid:1) is a constant, which is found using of the boundary conditions
The solution is that T (cid:0)a T + b ∂T zopt
(cid:18)
1 − d T
(cid:0) d
2 − a(cid:1) − d ρ2
(cid:17)
We conclude ∂T zopt
must be chosen so as to
remove the logarithmic divergence for T → 0: this yields d = 0. We can integrate the remaining terms
and set the constant of integration such that the average constraint q1
zopt
1
displacement writes, after simplification, zopt
can be rewritten as
(cid:104)− f opt
(cid:105)
(cid:0)zopt
(cid:1) = 0 is satisfied: the result is
1
. In terms of the original quantities, the solution for the axial correction to the
1
T . The Lagrange multiplier d =
= − d ρ2
2
a T + b ∂T zopt
(T ) = − a
µ(h1) ∇µ(h1)
T 2 − ρ2
2
1 − d
T
2
T 2 − ρ2
2
= − 1
2
T 2 − ρ2
2
(cid:16)
h(cid:48)
1
1 = T
b
z1 dT = 0
(cid:16)
2 b
(cid:17)
. This
.
1
2
2 b
zopt
1 =
as captured in equation (5.10).
References
(cid:26)
− 1
2
1
(T ) = − a
µ(h1) ∇µ(h1)
2 b
h1
(cid:16)
(cid:18)
(cid:17)
(cid:19)(cid:27)
T 2 − ρ2
2
h(cid:48)
1,
T
h1
Acerbi, E., Buttazzo, G., Percivale, D., 1991. A variational definition of the strain energy for an elastic string. Journal of
Elasticity 25, 137 -- 148.
Agostiniani, V., DeSimone, A., Koumatos, K., 2016. Shape programming for narrow ribbons of nematic elastomers, arxiv
1603.02088v1.
Audoly, B., Hutchinson, J. W., 2016. Analysis of necking based on a one-dimensional model. Journal of the Mechanics and
Physics of Solids 97, 68 -- 91.
Bardenhagen, S., Triantafyllidis, N., 1994. Derivation of higher order gradient continuum theories in 2,3-d non-linear elasticity
from periodic lattice models. Journal of the Mechanics and Physics of Solids 42 (1), 111 -- 139.
Barenblatt, G., Joseph, D. (Eds.), 1997. Collected Papers of R.S. Rivlin. Springer, New York, Ch. Stability of an elastic
material.
Bermudez, A., Viano, J. M., 1984. Une justification des ´equations de la thermo´elasticit´e des poutres `a section variable par des
m´ethodes asymptotiques. Mod´elisation Math´ematqiue et Analyse Num´erique 18, 347 -- 376.
Buannic, N., Cartaud, P., 2000. Higher-order effective modeling of periodic heterogeneous beams. I. Asymptotic expansion
method. International Journal of Solids and Structures 38, 7139 -- 7161.
Calladine, C. R., 1983. Theory of shell structures. Cambridge University Press.
Cimeti`ere, A., Geymonat, G., Le Dret, H., Raoult, A., Tutek, Z., 1988. Asymptotic theory and analysis for displacements and
stress distribution in nonlinear elastic straight slender rods. Journal of Elasticity 19, 111 -- 161.
Coleman, B. D., Newman, D. C., 1988. On the rheology of cold drawing. I. elastic materials. Journal of Polymer Science: Part
B: Polymer Physics 26, 1801 -- 1822.
Freddi, L., Hornung, P., Mora, M.-G., Paroni, R., 2016. A variational model for anisotropic and naturally twisted ribbons.
SIAM Journal on Mathematical Analysis 48 (6), 3883 -- 3906.
Freddi, L., Morassi, A., Paroni, R., 2004. Thin-walled beams: the case of the rectangular cross-section. Journal of Elasticity
76 (1), 45 -- 66.
Fu, Y. B., Pearce, S. P., Liu, K. K., 2008. Post-bifurcation analysis of a thin-walled hyperelastic tube under inflation. Interna-
tional Journal of Non-Linear Mechanics 43 (8), 697 -- 706.
Geymonat, G., Krasucki, F., Serpilli, M., 2018. Asymptotic derivation of linear plate model for soft ferromagnetic material.
Chinese Annals of Mathematics, Series B 39 (3), 451 -- 460.
G'Sell, C., Aly-Helal, N. A., Jonas, J. J., 1983. Effect of stress triaxiality on neck propagation during the tensile stretching of
solid polymers. Journal of Materials Science 18, 1731 -- 1742.
Hamdouni, A., Millet, O., 2006. An asymptotic non-linear model for thin-walled rods with strongly curved open cross-section.
International Journal of Non-Linear Mechanics 41 (3), 396 -- 416.
Kyriakides, S., Chang, Y.-C., 1990. On the inflation of a long elastic tube in the presence of axial load. International Journal
of Solids and Structures 26 (9 -- 10), 975 -- 991.
30
Kyriakides, S., Chang, Y.-C., 1991. The initiation and propagation of a localized instability in an inflated elastic tube. Inter-
national Journal of Solids and Structures 27 (9), 1085 -- 1111.
Lestringant, C., Audoly, B., 2017. Elastic rods with incompatible strain: macroscopic versus microscopic buckling. Journal of
the Mechanics and Physics of Solids 103, 40 -- 71.
Lestringant, C., Audoly, B., 2018. A diffuse interface model for the analysis of propagating bulges in cylindrical balloons.
Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 474, 20180333.
Liu, J., Huang, J., Su, T., Bertoldi, K., Clarke, D., 2014. Structural transition from helices to hemihelices. PLoS ONE 9 (4),
e93183.
Mahadevan, L., Vaziri, A., Das, M., 2007. Persistence of a pinch in a pipe. Europhysics Letters 77 (4), 40003.
Marigo, J.-J., Meunier, N., 2006. Hierarchy of one-dimensional models in nonlinear elasticity. Journal of Elasticity 83, 1 -- 28.
Matsuo, E. S., Tanaka, T., 1992. Patterns in shrinking gels. Nature 368, 1735.
Mora, S., Phou, T., Fromental, J.-M., Pismen, L. M., Pomeau, Y., Nov 2010. Capillarity driven instability of a soft solid.
Physical Review Letters 105, 214301.
Ogden, R. W., 1972. Large deformation isotropic elasticity-on the correlation of theory and experiment for incompressible
rubber-like solids. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Science 326, 565 -- 584.
Pearce, S. P., Fu, Y. B., 2010. Characterization and stability of localized bulging/necking in inflated membrane tubes. IMA
Journal of Applied Mathematics 75, 581 -- 602.
Picault, E., Bourgeois, S., Cochelin, B., Guinot, F., 2016. A rod model with thin-walled flexible cross-section: Extension to 3D
motions and application to 3D foldings of tape springs. International Journal of Solids and Structures 84, 64 -- 81.
Sadowsky, M., 1930. Ein elementarer Beweis fur die Existenz eines abwickelbaren Mobiusschen Bandes und die Zuruckfuhrung
des geometrischen Problems auf ein Variationsproblem. In: Sitzungsberichte der Preussischen Akademie der Wissenschaften,
physikalisch-mathematische Klasse, 17. Juli 1930, Mitteilung vom 26. Juni. pp. 412 -- 415.
Sanchez-Hubert, J., Sanchez Palencia, E., 1999. Statics of curved rods on account of torsion and flexion. European Journal of
Mechanics. A. Solids 18, 365 -- 390.
Seffen, K. A., Pellegrino, S., 1999. Deployment dynamics of tape springs. Proceedings of the Royal Society of London. Series
A: Mathematical, Physical and Engineering Sciences 455 (1983), 1003 -- 1048.
Trabucho, L., Viano, J. M., 1996. Mathematical modelling of rods. Handbook of numerical analysis 4, 487 -- 974.
Triantafyllidis, N., Scherzinger, W. M., Huang, H.-J., 2007. Post-bifurcation equilibria in the plane-strain test of a hyperelastic
rectangular block. International Journal of Solids and Structures 44 (11 -- 12), 3700 -- 3719.
van der Hoeven, J., Grozin, A., Gubinelli, M., Lecerf, G., Poulain, F., Raux, D., 2013. GNU TEXmacs: a scientific editing
platform. ACM Communications in Computer Algebra 47 (1 -- 2), 59 -- 61.
Wunderlich, W., 1962. Uber ein abwickelbares Mobiusband. Monatshefte fur Mathematik 66 (3), 276 -- 289.
Xuan, C., Biggins, J., May 2017. Plateau-Rayleigh instability in solids is a simple phase separation. Physical Reivew E 95,
053106.
31
|
1809.04483 | 2 | 1809 | 2019-02-05T15:42:14 | Low-power optical beam steering by microelectromechanical waveguide gratings | [
"physics.app-ph",
"physics.optics"
] | Optical beam steering is key for optical communications, laser mapping (LIDAR), and medical imaging. For these applications, integrated photonics is an enabling technology that can provide miniaturized, lighter, lower cost, and more power efficient systems. However, common integrated photonic devices are too power demanding. Here, we experimentally demonstrate, for the first time, beam steering by microelectromechanical (MEMS) actuation of a suspended silicon photonic waveguide grating. Our device shows up to 5.6{\deg} beam steering with 20 V actuation and a power consumption below the $\mu$W level, i.e. more than 5 orders of magnitude lower power consumption than previous thermo-optic tuning methods. The novel combination of MEMS with integrated photonics presented in this work lays ground for the next generation of power-efficient optical beam steering systems. | physics.app-ph | physics | Letter
Vol. 44, No. 4 / 15 February 2019 / Optics Letters
855
Low-power optical beam steering by
microelectromechanical waveguide gratings
AND KRISTINN B. GYLFASON1
CARLOS ERRANDO-HERRANZ,1,* NICOLAS LE THOMAS,2,3
1Department of Micro and Nanosystems, KTH Royal Institute of Technology, Malvinas väg 10, SE-100 44 Stockholm, Sweden
2Photonics Research Group, INTEC Department, Ghent University-imec, Technologiepark-Zwijnaarde 15, 9052 Ghent, Belgium
3Center for Nano- and Biophotonics, Ghent University, Technologiepark-Zwijnaarde 15, 9052 Ghent, Belgium
*Corresponding author: [email protected]
Received 23 November 2018; revised 11 January 2019; accepted 13 January 2019; posted 14 January 2019 (Doc. ID 352248);
published 6 February 2019
Optical beam steering is key for optical communications,
laser mapping (lidar), and medical imaging. For these ap-
plications, integrated photonics is an enabling technology
that can provide miniaturized, lighter, lower-cost, and more
power-efficient
systems. However, common integrated
photonic devices are too power demanding. Here, we exper-
imentally demonstrate, for the first time, to the best of our
knowledge, beam steering by microelectromechanical
(MEMS) actuation of a suspended silicon photonic wave-
guide grating. Our device shows up to 5.6° beam steering
with 20 V actuation and power consumption below the
μW level, i.e., more than five orders of magnitude lower
power consumption than previous thermo-optic tuning
methods. The novel combination of MEMS with integrated
photonics presented in this work lays ground for the
next generation of power-efficient optical beam steering
systems.
© 2019 Optical Society of America
https://doi.org/10.1364/OL.44.000855
Optical beam steering is required in a wide range of applications,
such as high-speed optical communications [1], lidar for artifi-
cial vision [2], and medical imaging [3]. Traditionally, optical
beam steering systems use electrical motors to tilt mirrors
and scan a laser beam over a certain area, which suffer from large
size and weight, cost thousands of USD, and consume watts of
power [4]. These traditional systems are impractical for battery-
driven robots, mobile phones, or drones, for in vivo optical
coherence tomography (OCT) probes [3], and for miniaturized
and low-cost space division multiplexing (SDM) [1]. More
recently, optical beam steering systems have been scaled down
by using microelectromechanical (MEMS) mirrors and gratings,
resulting in significant reduction in cost and weight [5,6].
However, the parts of such systems (i.e., laser, scanning device,
detector, and electronics) are still fabricated independently, and
require costly assembly. Further miniaturization has the poten-
tial to provide smaller, lighter, and less power-consuming beam
steering at a low cost -- features that are required for the contin-
ued success of optical beam steering technologies [2].
0146-9592/19/040855-04 Journal © 2019 Optical Society of America
Integrated photonics, and silicon photonics in particular,
can potentially address these challenges by densely integrating
photonic devices for beam steering and optical signal process-
ing, optical sources, and detectors [7], with electrical processing
and control [8]. This results in integrated photonics systems
outperforming free-space optics not only in size and weight,
but also in cost, integration density, and robustness.
Integrated photonics approaches to beam steering have fo-
cused mostly on optical phased arrays. An optical phased array
consists of an array of emitters (usually grating couplers), result-
ing in a diffraction pattern in the far field highly dependent on
the relative phases of the emitted waves. By tuning the relative
phases of such waves using waveguide phase shifters, the output
beam angle is tuned. These systems allow a very tight control
over the beam shape and direction, and previous work has
shown 1D steering [9], very high angular beam resolution 2D
steering [10], and lidar measurements [11]. However, the com-
monly used thermo-optic phase shifters have an important
drawback: very high power consumption. This is caused by
the need for one phase shifter per emitter, requiring hundreds
of devices packed in a tight chip space. This has led to power
consumption on the order of watts (0.5 W for 1D steering in
[9], and about 4 W in [10]), which necessitates active cooling
and thus limits severely the applications of this technology.
Recently, a low-power lidar based on reverse-biased electro-
optic phase shifters achieved 2 μW power consumption per
phase shifter, amounting to about 1 mW for the required array
of 512 [12]. The optical loss was up to 4 dB per phase shifter.
A different approach to beam steering used a single thermo-
optically tuned grating coupler, and achieved a limited steering
of 2.7° while consuming 130 mW of electrical power in static
operation [13]. Along this line, a recent approach combined
thermo-mechanical actuation with thermo-optic tuning in a
grating coupler, and used it to improve the efficiency of
thermo-optic spectral tuning of the grating transmission [14].
MEMS tuning of photonic waveguide devices can provide
more than five orders of magnitude lower power consumption
compared to traditional tuning methods (sub-μW for MEMS
[15] compared to 30 mW per thermo-optic phase shifter [16]),
with prospects for upscaling photonic integrated circuit (PIC)
technologies [17]. In our previous work, we demonstrated a
856
Vol. 44, No. 4 / 15 February 2019 / Optics Letters
Fig. 1.
(a) Schematic showing the working principle of our MEMS
tunable grating before, and under actuation. (b) Scanning electron
microscope (SEM) image of our device. The grating is part of a soft
spring, stretched via a comb drive actuator, which changes the spacing
between grating teeth. Note that the warped grating is due to early
failure after actuation. (c) Simulation results (color: emitted light
intensity) with overlaid analytical estimate (white line) of the effect
of increased grating teeth separation on beam steering. (d) Analytical
actuation estimate for a comb-drive actuated device.
MEMS tunable grating for on-chip optimization of
light
coupling between an optical fiber and a silicon waveguide
by using vertical displacement of a grating embedded in a
cantilever [18].
In this Letter, we experimentally demonstrate, for the first
time, low-power beam steering using a MEMS tunable wave-
guide grating. Our results show beam steering up to 5.6° at a
wavelength of 1550 nm with actuation voltages below 20 V and
sub-μW static power consumption.
Our beam steering device is based on changing the spacing
between the teeth of a waveguide grating coupler using MEMS
actuation. Figure 1(a) shows a schematic view of the device. We
designed a suspended grating forming a folded spring, con-
nected on one side to a waveguide taper for light coupling,
and on the other side to a MEMS comb drive actuator.
Horizontal actuation of the comb drive pulls and stretches
the suspended grating, which changes the spacing between gra-
ting teeth, resulting in a change in the out-of-plane angular
emission of the grating. Figure 1(b) shows a scanning electron
microscope (SEM) image of a fabricated MEMS tunable
grating.
An analytical estimate of the effect of varying the gap in a
suspended grating can be obtained by using the standard gra-
ting equation assuming in-plane excitation, and estimating the
effective grating index neff as a weighted average between the
effective refractive indices of the air gaps (width g) and silicon
grating teeth (simulated nwg
2.4, width d):
Letter
, with neff
nwgd nairg
d g
:
(1)
sin θ neff
− λ
dg
nair
Thus, an equation relating the out-coupled light angle θ to the
gap width g was obtained. The white curve in Fig. 1(c) is the
graph of that analytical relation, assuming a silicon grating
tooth width d 300 nm at λ 1550 nm wavelength.
However, the analytical solution is only a rough approxima-
tion, due to the sub-wavelength scale of the structures involved.
To get a more accurate prediction of the effect of varying
grating spacing on the emission angle, optical simulations are
required. We performed such simulations using a finite-
difference time-domain solver (varFDTD, Lumerical Solutions),
commonly used for grating coupler simulations in silicon
photonics. We simulated the cross section of an air-cladded, sus-
pended grating (tooth width d 300 nm, device layer thick-
ness t 220 nm), including the under-etched buried oxide
layer (thickness 2 μm) and the silicon substrate. The input wave-
guide was excited with the fundamental TE waveguide mode,
and we investigated the out-coupled optical intensity in the
far field for a grating with 15 teeth [Fig. 1(c)].
The mechanical design is based on a suspended comb drive
actuator, with the attached tunable grating as an additional
soft spring. The grating was designed using a tooth width of
300 nm, with 300 nm wide initial gaps, and grating width
20 μm abutting a waveguide taper with a wider end of 12 μm
width [see Fig. 1(b)]. In our device, the change in gap between
teeth is the total MEMS displacement divided by the number
of teeth, and thus the number of teeth was chosen to be five to
ease observation of beam steering even at short MEMS dis-
placements. The grating design was 20 μm wide to minimize
the in-plane angular variation from tooth to tooth under actua-
tion (below 6° for 1 μm gap increase), so that along the central
area of the grating, i.e., where the optical mode is concentrated,
the effect of the variation in gap from tooth to tooth is negli-
gible. The grating spring is designed to be soft (spring constant
−4 N∕m), so that it is negligible in the MEMS
kgrat
actuation. The comb drive actuator is designed for fabrication
on a standard silicon-on-insulator (SOI) 220 nm thick device
layer, following [19]. The actuator uses four symmetric spring
0.44 N∕m), designed as folded beams
suspensions (ksprings
with beam width 300 nm, beam length 16 μm, and separation
between beams of 3 μm. The force balance equation is
3 × 10
kspringsx ϵ
0tN
w
D − x2
1
s
V 2,
(2)
with ϵ
0 the permittivity of vacuum, t the device layer thickness,
N the number of comb finger pairs, s the finger spacing, w the
width of each finger, D the initial distance between the end
of a finger and the beginning of the opposite one, x the
comb displacement, and V the applied voltage. Our designed
comb parameters are t 220 nm, N 36,
s 400 nm,
w 300 nm, and D 1.8 μm.
The maximum displacement for our designed actuator is
close to 900 nm, at a voltage of 40 V. Combining Eqs. (1)
and (2) leads to the actuation curve for the MEMS tunable
grating coupler in Fig. 1(d), with potential beam steering up
to 30° at 40 V actuation.
The device was fabricated using the simple process presented
in [20]. The process starts from a standard silicon photonic
SOI substrate (220 nm Si device layer on 2 μm buried
Letter
Vol. 44, No. 4 / 15 February 2019 / Optics Letters
857
Fig. 2.
(a) Experimental setup for measuring the angular emission
from our devices by Fourier imaging. The mirror and Fourier lens at
the rear end can be removed to view (b) the visible image of the grating
(blue optical path). (c) By inserting the mirror, we image the IR emis-
sion of the grating onto the IR camera. (d) By inserting also the Fourier
lens, we make the back focal plane visible, resulting in a k-space image
of the grating emission on the IR camera (outlined red optical path).
SiO2),
followed by two consecutive e-beam-lithography-
defined silicon etching steps to pattern two silicon thickness
levels. Then, a SiO2 free-etch using hydrofluoric acid (HF,
50% aqueous solution) and critical point drying (CPD) results
in the suspended MEMS structures [see Figs. 1(b) and 2(b)].
To characterize the beam steering, a Fourier imaging setup
was used [21], and a schematic can be found in Fig. 2(a). This
setup is based on an optical Fourier microscope using an
objective with large numerical aperture (NA 0.95), allowing
measurement of beam angles up to 71.8° with respect to the
normal.
Laser light at 1550 nm wavelength and 1 mW power is
edge-coupled via a polarization controller through a standard
optical fiber into the quasi-TE mode of the on-chip wave-
guides. On the chip, the light is guided to the tunable grating,
which emits light out of plane into a large NA microscope ob-
jective, part of the Fourier imaging setup. In order to visually
inspect the sample, visible light is sent into the objective using a
beam splitter, and the reflected light is routed through a set of
lenses for magnification, and finally onto a CCD camera [a vis-
ible light image is shown in Fig. 2(b)]. The infrared (IR) light
emitted from our gratings follows the same optical path.
However, optionally placing a mirror [labeled removable mirror
in Fig. 2(a)] in front of the CCD camera reflects the light into
an IR camera, and a real space image of the grating IR emission
is formed, and can be seen in Fig. 2(c).
Fig. 3.
(a) k-space images showing the change in beam directionality
with actuation. The white dashed lines show the cross-sectional axis
used for plotting (b). (b) Beam steering for a range of actuation
voltages along the MEMS actuation axis kx. The results show up
to 5.6° steering, using an analytical curve to fit the emission maxima
(black curve). (c) Cross section along kx direction at the beam maxima
for each voltage, showing negligible perpendicular beam distortion
with actuation.
To obtain an image of the IR k-space, an additional lens can
be placed in front of the IR camera [labeled removable lens in
Fig. 2(a)], resulting in back focal plane (Fourier) imaging. An
example of a k-space image of measured grating emission is
shown in Fig. 2(d), which consists of a circle with its radius
defined by the largest angular emission that the system can
detect, i.e., 71.8°. The emission angle along the radial direction
is defined by r sin θ, with r being the radial distance.
The substrate of the sample is electrically grounded via a
copper plate, and two soft electrical probes, in direct contact
with the silicon device layer, connect the comb drive electrodes
to a voltage source for MEMS actuation. Actuation of the comb
drive then results in pulling forces on the grating, changing the
tooth spacing and resulting in beam steering.
Figure 3 shows our beam steering measurement results.
The k-space images in Fig. 3(a) show the effect of increasing
actuation voltage on the beam directionality along the MEMS
actuation axis ky for 0 V and 20 V actuation. The absence of
features other than beam steering, which would appear as
different shapes in the k-space images, illustrates the absence
of stray light or added noise in the k-space under MEMS
858
Vol. 44, No. 4 / 15 February 2019 / Optics Letters
Letter
actuation. A cross section along ky in Fig. 3(b) shows the evo-
lution of the beam angle for a range of actuation voltages from
0 V to 20 V. Our device achieves 5.6 0.3° of beam steering
with a full width at half maximum (FWHM) divergence of 14°
along ky and 9° along kx. The beam steering angles were ob-
tained by fitting the maxima at each voltage to the actuation
curve shape in Fig. 1(d), and the fit is plotted in Fig. 3(b), with
the error value on beam steering angle calculated as one stan-
dard deviation. We can extract a tuning rate of Δθ∕ΔV
0.56°∕V from 10 V to 20 V. The cross section along the kx
axis at the light intensity maxima, shown in Fig. 3(b), shows
minor distortions in the direction perpendicular to the actua-
tion, which can be originated from non-parallel displacement of
the grating teeth. The efficiency of the device was estimated by
simulating the suspended to unsuspended waveguide transition
and the grating, and yielded <0.01 dB and 35% efficiency
(varying between 30% and 40% with actuation), respectively.
The static power consumption was in the sub-μW range, below
our measurement capabilities. The nanogram mass of our
MEMS structure makes gravitational forces negligible, and sets
the mechanical resonance frequency around 200 kHz, which
is far from mechanical noise sources, and sets a limit to the
steering speed. The maximum power consumption for the
device, estimated as 200 kHz full charge -- discharge cycles, is
below 10 nW.
The large beam divergence is due to the high-index contrast,
and thus the limited number of grating teeth contributing to
the diffraction pattern. This can be overcome by designing a
lower-refractive-index-contrast grating by, e.g., thinning down
the silicon device layer, or by choosing a different low-index
material platform such as silicon oxide or silicon nitride.
The limited actuation range, less than half of the theoretical
prediction, is due to premature collapse of the MEMS actuator,
stemming from the grating spring, and can be observed in the
tunable grating area in Fig. 1(b). We believe there are two rea-
sons for this effect: (i) large displacements of the grating result
in local stiffening and softening, resulting in bending momenta
and contact between grating teeth, and (ii) the asymmetry of
the grating spring generates an off-axis horizontal force under
actuation, resulting in lateral drift of the comb drive and con-
tact between grating teeth. We believe these problems are not
fundamental, and can be solved by (i) design of grating teeth
joints to avoid strain concentration, and (ii) design of symmet-
ric gratings, either by mirroring the current design, or by
designing a suspended grating connected on the sides.
Compared to free-space optics, the presented technology is
orders of magnitude smaller and lighter, lower in cost, less
prone to mechanical noise, and requires very limited assembly.
Integrated thermo-optic phased array systems have at least five
orders of magnitude (limited by measurement, seven orders of
magnitude based on our estimate) higher power consumption
than our device, and suffer from thermal cross-talk problems,
which our
technology inherently avoids. Compared to
elecro-optic tuning, our device features at least one order (more
likely three orders) of magnitude lower power consumption.
Furthermore, we achieve more than two times larger beam
steering than previously reported thermo-optic tunable single
gratings
large-angle tuning [see
Fig. 1(d)] with future improvements in MEMS actuator design.
We have introduced, for the first time, MEMS tunable
waveguide gratings for beam steering, and experimentally
[13], with potential
for
demonstrated beam steering up to 5.6°, with an actuation
voltage below 20 V. Our results show more than twice the
beam steering of previously reported thermo-optic tuning of
waveguide gratings [13], and more than five orders of magni-
tude lower power consumption.
The optical beam steering technology presented here can
provide the long-sought reduction in cost and power consump-
tion necessary to extend artificial vision to battery-powered
devices, including mobile phones or drones, to enable active
probes for in vivo medical imaging, and to grow the optical
communication bandwidth by SDM.
Funding. Vetenskapsrådet (VR) (621-2012-5364); H2020
European Institute of Innovation and Technology (EIT)
(780283).
Acknowledgment. N. L. T. acknowledges the support of
the MRP initiative at Ghent University through NB-Photonics.
REFERENCES
1. J. M. Kahn and D. A. B. Miller, Nat. Photonics 11, 5 (2017).
2. E. Ackerman, IEEE Spectr. 53, 14 (2016).
3. A. F. Fercher, W. Drexler, C. K. Hitzenberger, and T. Lasser, Rep.
Prog. Phys. 66, 239 (2003).
4. Velodyne LiDAR, "VLP-16."
5. H. Ra, W. Piyawattanametha, M. J. Mandella, P.-L. Hsiung, J. Hardy,
T. D. Wang, C. H. Contag, G. S. Kino, and O. Solgaard, Opt. Express
16, 7224 (2008).
6. O. Solgaard, A. A. Godil, R. T. Howe, L. P. Lee, Y. A. Peter, and H.
Zappe, J. Microelectromech. Syst. 23, 517 (2014).
7. Z. Zhou, B. Yin, and J. Michel, Light: Sci. Appl. 4, e358 (2015).
8. C. Sun, M. T. Wade, Y. Lee, J. S. Orcutt, L. Alloatti, M. S. Georgas,
A. S. Waterman, J. M. Shainline, R. R. Avizienis, S. Lin, B. R. Moss, R.
Kumar, F. Pavanello, A. H. Atabaki, H. M. Cook, A. J. Ou, J. C. Leu,
Y.-H. Chen, K. Asanović, R. J. Ram, M. A. Popović, and V. M.
Stojanović, Nature 528, 534 (2015).
9. K. V. Acoleyen, W. Bogaerts, J. Jágerská, N. L. Thomas, R. Houdré,
and R. Baets, Opt. Lett. 34, 1477 (2009).
10. D. N. Hutchison, J. Sun, J. K. Doylend, R. Kumar, J. Heck, W. Kim,
C. T. Phare, A. Feshali, and H. Rong, Optica 3, 887 (2016).
11. C. V. Poulton, A. Yaacobi, D. B. Cole, M. J. Byrd, M. Raval, D.
Vermeulen, and M. R. Watts, Opt. Lett. 42, 4091 (2017).
12. C. V. Poulton, P. Russo, E. Timurdogan, M. Whitson, M. J. Byrd,
E. Hosseini, B. Moss, Z. Su, D. Vermeulen, and M. R. Watts, in
Conference on
(2018),
paper ATu3R.2.
and Electro-Optics
(CLEO)
Lasers
13. S.-H. Kim, J.-H. Kim, S. Yeo, G. Kang, J. Kim, D. Yoo, H. Lee, D.-S.
Lee, and H.-H. Park, Proc. SPIE 10109, 101090Z (2017).
14. C. P. Ho, Z. Zhao, Q. Li, S. Takagi, and M. Takenaka, IEEE Photon.
Technol. Lett. 30, 1503 (2018).
15. C. Errando-Herranz, F. Niklaus, G. Stemme, and K. B. Gylfason, Opt.
Lett. 40, 3556 (2015).
16. A. Ribeiro, A. Ruocco, L. Vanacker, and W. Bogaerts, Optica 3, 1348
(2016).
17. S. Han, T. J. Seok, N. Quack, B.-W. Yoo, and M. C. Wu, Optica 2, 370
(2015).
18. C. Errando-Herranz, M. Colangelo, S. Ahmed, J. Bjork, and K. B.
Gylfason, in IEEE 30th International Conference on Micro Electro
Mechanical Systems (MEMS) (2017), p. 293.
19. K. Takahashi, E. Bulgan, Y. Kanamori, and K. Hane, IEEE Trans. Ind.
Electron. 56, 991 (2009).
20. C. Errando-Herranz, F. Niklaus, G. Stemme, and K. B. Gylfason, in
28th IEEE International Conference on Micro Electro Mechanical
Systems (MEMS) (2015), pp. 53 -- 56.
21. N. L. Thomas, R. Houdré, M. V. Kotlyar, D. O'Brien, and T. F. Krauss,
J. Opt. Soc. Am. B 24, 2964 (2007).
|
1903.09794 | 1 | 1903 | 2019-03-23T10:37:28 | Achromatic Non-Interferometric Single Grating Neutron Phase Imaging | [
"physics.app-ph"
] | We demonstrate a simple single grating beam modulation technique, which enables the use of a highly intense neutron beam for phase imaging and thus spatially resolved structural correlation measurements in full analogy to quantum interference based methods. In contrast to interferometric approaches our method is intrinsically achromatic and provides unprecedented flexibility in the choice of experimental parameters. Utilizing merely a macroscopic absorption mask unparalleled length scales become accessible. The results presented, including application to a variety of materials, establish a paradigm shift in phase imaging. | physics.app-ph | physics | Achromatic Non-Interferometric Single Grating Neutron Phase Imaging
M. Strobl,1,2* J. Valsecchi,1,$ R.P. Harti,1 P. Trtik,1 A. Kaestner,1 C. Gruenzweig,1 E. Polatidis,1 J.
Capek1
1Laboratory for Neutron Scattering and Imaging, Paul Scherrer Institut, 5232 Villigen,
Switzerland
2Niels Bohr Institute, University of Copenhagen, Nørregade 10, 1165 Copenhagen, Denmark
*[email protected]
[email protected]
Abstract
We demonstrate a simple single grating beam modulation technique, which enables the use
of a highly intense neutron beam for phase imaging and thus spatially resolved structural
correlation measurements in full analogy to quantum interference based methods. In
contrast to interferometric approaches our method is intrinsically achromatic and provides
unprecedented flexibility in the choice of experimental parameters. Utilizing merely a
macroscopic absorption mask unparalleled length scales become accessible. The results
presented, including application to a variety of materials, establish a paradigm shift in phase
imaging.
State of the art imaging with visible light as well as with electrons comprises numerous
contrast mechanisms, including phase and dark-field contrast schemes [1,2]. X-ray and
neutron imaging, on the other hand, have only recently seen the development of an
analogous variety of contrast modalities [3,4]. Today interferometric and non-interferometric
techniques play an important role in phase imaging with all these types of radiation in
providing contrast based on the real part of the refractive index, where conventional
attenuation contrast fails.
Neutron interferometry has created significant impact in neutron imaging in particular when
Talbot Lau grating interferometers enabled unprecedented high efficiency phase contrast
imaging with neutrons early in the millennium [5]. Attempts to utilize interferometry and its
outstanding sensitivity to quantum beam phase effects for spatially resolved measurements
have been reported earlier and in particular with single crystal Mach-Zehnder interferometers
[6]. However, their coherence and stability requirements as well as spatial restrictions
hindered practical applications. Also attempted non-interferometric techniques suffered
from high coherence requirements [7,8]. The ability of the grating interferometer to operate
at relatively low coherence requirements overcame these restrictions. Thus, it unlocked the
potential of the highly penetrating quantum beam of neutrons to visualize through phase
contrast e.g. magnetic structures. Studies of vortex structures, magnetic domains and domain
wall kinetics in the bulk of superconductors and ferromagnets which were not amenable to
any other technique have been reported [9-14]. In addition, the high sensitivity of the
apparatus was found to qualify it an excellent tool for the investigation of micro-structural
features based on coherent scattering length density variations in bulk materials and objects
[15-18]. In contrast to conventional instrumentation, the combination of microstructural
sensitivity with macroscopic real space resolution enables the study of heterogeneous
structures and processes in representative volumes [19-21]. Due to analogies to dark-field
microscopy, this phase imaging technique is in x-ray and neutron imaging referred to as dark-
field contrast imaging [15,22].
Today grating interferometers for imaging are available at many leading instruments at
neutron sources around the world [5,15,23-26]. Numerous successful studies [9-21] triggered
continuous developments with regards to varying implementations of analogue techniques
[10,25-30]. Advances targeted mainly efficiency, complexity and assessable correlation length
scales. Key progress was established in particular by the quantitative interpretation of dark-
field contrast providing correlation length information [17,18]. The first quantitative
characterization of microstructures was reported on the nano-scale already when a spin-echo
interferometer was successfully introduced to neutron imaging [28]. It operates in full
analogy to Talbot Lau grating interferometric imaging but builds on the quantum spin phase
and interference of the two spin states of the neutron wave function. The advantage of the
this method is access to the nano-scale, full remote control of the modulation as well as its
suitability for the most efficient time-of-flight wavelength dispersive measurement approach
with neutrons at pulsed spallation sources [28,29]. Disadvantages are the sophisticated and
elaborate set-up and technological barriers to assess the micro-scale, without excessive
sample to detector distances. This scale is however covered by the common Talbot Lau
interferometry [18]. A more recent implementation, referred to as far-field multi-phase-
grating interferometer [30], implies similar issues in the conventional range, but in addition
faces efficiency issues due to a high general collimation requirement. An advantage is meant
to be a lower wavelength dependence of the visibility achieved in the beam modulation.
Here we introduce a paradigm shift demonstrating that the interferometric nature of the
applied techniques is not relevant but only the spatial modulation of the beam, no matter
how it is achieved. In contrast to previous works underlining the virtues of quantum particle-
wave interferometry we chose the most basic approach to create spatial beam modulation
on a suitable length scale to retrievably encode small angle beam deviations superimposed
to real space images [17]. We prove that we are thus able to record and analyse the
corresponding multi-modal images providing attenuation contrast, differential phase and
dark-field contrast in full analogy to the discussed interferometric techniques. In addition, a
close analogy to non-interferometric phase imaging through structured illumination in light
microscopy [31,32] can be found in this approach.
The set-up is established by merely adding an attenuation grating in a conventional pinhole
collimated imaging instrument (Fig. 1). The image of the grating will be an accordingly
spatially modulated intensity profile. Due to the geometric blur the modulation image will
feature a sinusoidal intensity distribution, comparable to these of interferometric methods.
The achievable visibility V=(Imax-Imin)/(Imax+Imin) depends, for a perfect absorption grating, only
on the resolution capability of the set-up and is fully independent of the wavelength, hence
fully achromatic in contrast to any approach presented earlier (Fig.1b, c). This modulation of
the beam which is simply a projection image of an absorption grating will be used in analogy
to all interference based beam modulations before to provide quantifiable spatially resolved
measurements of attenuation, phase and scattered wave interference. The principle of the
latter is schematically depicted in Fig. 1a.
FIG. 1 Set-up and parameters (a) sketch of the basic set-up adding a simple attenuation grating at a
distance L from the pinhole and upstream of a variable sample position in a conventional neutron
imaging instrument. The principles of differential phase and scattering detection (dark-field contrast)
are depicted. The blue curve represents the undisturbed pattern and red is the refracted (sample sides)
and scattered (sample center) pattern. (b) visibilities V achieved for L=7m pinhole to grating distance
for different standard pinhole sizes D = 40, 20 and 10 mm and the utilized absorption grating period
365 µm. The red line at 10% marks the theoretical resolution limit and the lines provide a conservative
visibility estimate based on the calculated geometric blur d=l/(L/D). The point marked by a red circle
indicates the parameters used for the presented study. (c) measured proof of achromatic visibility with
data taken at the point marked with yellow circle in (b). (d) achievable correlation lengths according
to eq. (2)[17] for standard pinhole sizes available at the benchmark instrument ICON at PSI [33] and
D=0.2mm like utilized for far-field interferometry [30]. Minima of indicated ranges refer to LS = 10 mm
and a minimum wavelength of 2.8 Å, maxima to LS according to (b) and a wavelength of 5.6 Å. At D=20
mm a comparison with the use of a 222 µm period grating is provided and at D=10 mm the parameters
utilized in the presented study are shown.
The sensitivity of a modulated beam measurement technique depends on the modulation
period p and the sample to detector distance Ls (Fig. 1a) and can in the small angle
approximation be expressed by the characteristic angle or the probed scattering vector length
qc = p/2Ls
respectively. At this angle, or scattering vector, intensity is shifted from the maxima exactly
into the intensity depleted minima -- the dark-field - of the modulation pattern. This in turn
for the probed length scale equals [17]
x = lLs/p
For the intrinsically achromatic technique the wavelength l is a free parameter (Fig. 1c), but
it is limited by the typically used spectra in the thermal and cold energy range. Both other
parameters, Ls and p, are limited by the spatial resolution ability of the set-up (Fig. 1b). A
and qc = pp/lLs ,
(1)
.
(2)
larger distance and a smaller period provide access to larger structures (Fig. 1d). A typical
Talbot Lau interferometer provides modulation conditions enabling to resolve micrometer
sized structures with micrometer modulation periods. These require analyser gratings to be
resolved. Here we consider periods of some 100 micrometer which can be resolved spatially
up to some 100 mm distances by the detector directly. A workable compromise between
visibility and resolvable range has to be found (Fig. 1b & d). Because the spatial resolution
and hence the visibility at a certain grating to detector distance in a conventional pinhole
collimated set-up is indirect proportional to the available flux density, also efficiency
considerations have to be evaluated carefully. Theoretical approximations as well as
measurements confirm the ability to create modulations with substantial visibility and
compatible with outstanding correlation length ranges probed from the nanometer to
micrometer scale with standard instrument settings (Fig. 1b & d). This applies despite the
limited wavelength range of 2.8 to 5.6 Å accessible with the utilized instrumentation. Fig. 1 d,
in addition, displays the unparalleled extensive range our technique covers when employing
slits as reported to be applied for far field interferometry [30]. For our demonstration an
intermediate regime with the potential to cover nearly two orders of magnitude in correlation
lengths has been chosen while analyzed data are restricted to correlation lengths ranging
from about 10 to 300 nm. This range clearly complements that of conventional Talbot Lau
interferometry.
The measurements were performed with a 10 mm pinhole, common for conventional high
resolution imaging measurements and leading to a collimation ratio L/D of about 700. The
chosen L/D and grating to detector distance l of 0.26 m enabled a visibility of 18% (Fig. 1b)
comparable to such utilized in neutron Talbot Lau interferometry. The grating consisted of 20
µm high Gd lines on a quartz wafer with a duty cycle of 50% and a period of 365 µm. The
detector used was a common combination of a 20 µm thick gadolinium oxysulfide (Gadox)
scintillator screen and a commercial CCD camera (Andor, iKon-L) with a 100 mm Zeiss
photographic lens system. The corresponding field of view (FoV) was 70 x 70 mm2 with an
effective pixel size of 35 µm and an intrinsic spatial resolution of around 70 µm. The grating
had an effective size of 50 x 50 mm2 and hence did not cover the full FoV of the detection
system. The exposure time per image was 3 times 120 sec irrespective of the wavelength
used.
Fig. 2 Three imaging modalities are measured simultaneously and
in analogy to grating
interferometric imaging [9]; the images display a measurement of three samples, a 304L steel cylinder
and two quartz glass cuvettes containing aqueous solutions; (a) the conventional transmission image
(TI) corresponds to the ratio between transmitted and incident beam intensities A/A0 (eq. 3); (b) the
differential phase image (DPI) is established by the modulation phase difference between open beam
and sample q (eq. 3), the line profile contains a theoretical calculation for the steel cylinder (black
curve)(eq. 4); (c) the dark-field image (DFI) reflects the relative visibility loss through scattering from
the sample V/V0=BA0/(B0A) (eq. 3);
A macroscopic beam modulation superimposed to an image enables different routes of
measurement and analyses strategies for the contrast modalities [15,28,34,35]. While in
principle it is possible to fully analyse such images for all modalities from a single shot [28,35],
here, again in analogy to Talbot Lau interferometry, a grating scan has been performed with
11 steps over one period. This enables conventional pixel-wise extraction of all three contrast
parameters: transmission A, differential phase parameter q and visibility V=B/A constituting
the measured image as
Ii,j = Ai,j + Bi,j sin(Ci,jX + qi,j)
where i,j are the pixel indices, C is the phase of the open beam modulation and X is the grating
scan parameter.
For an individual measurement of a set of samples the images corresponding to the three
parameters: transmission image (TI), differential phase image (DPI) and dark-field image (DFI)
are depicted in Fig. 2. The samples in this measurement are a cylindric tensile test dogbone
sample of 304L steel and two aqueous solutions in quartz glass cuvettes placed in the beam.
While the attenuation by the water in the 5mm thick cuvettes is substantial as can be seen in
the attenuation contrast transmission image (TI) it affects the statistics in the corresponding
areas in the differential phase image (DPI) and in the dark-field image (DFI) significantly.
Nevertheless, all three contrast modalities display the typical corresponding features with
respect to the samples, which are well-known from corresponding interferometric imaging
approaches. Due to the good spatial resolution capability of the set-up and the significant
sample to detector distance even the signature of far field phase contrast [7] is visible upon
careful inspection in the attenuation contrast image (TI, Fig. 2a). The differential phase
contrast image depicts the influence of the distorted neutron wave-front on the spatial
modulation phase q (Fig. 1a). The respective local neutron wave phase shift can be extracted
from this according to the relation [5]
(3)
𝜃=#$%& '()*
(4)
where ∂Q/∂x is the gradient of the neutron wave-front perpendicular to the modulation. The
phase can be retrieved straightforwardly from corresponding integration [5]. For the phase
profile of the cylindric steel sample we compare the measurement to the according
calculation of the differential phase in Fig. 2b, underlining good quantitative agreement
within the spatial resolution limit.
Systematic quantitative reference measurements have been performed on a second set of
samples. The study focusses on quantitative dark-field imaging [15,17-21] which has
established as the dominant application of interferometeric modulated beam imaging in
material research. The first reference samples were two commercially available fractal
powders, namely Sipernat-350 and Sipernat-310 with characteristic particle sizes of 4 and 8.5
micrometers, respectively [36]. Sipernat-310 is a powder of silica featuring a large surface
area of 700 m2/g. The large surface area together with the micrometer sized particle
characteristics makes Sipernat-310 a well-suited material to investigate basic characteristics
of cohesive powders. It has recently been used in a study observing with spatial resolution
the heterogeneous breakdown of the fractal microstructure [21]. This study provided the
required reference data and model fit from more conventional spin-echo small angle neutron
scattering measurements with no direct spatial resolution.
FIG. 3 Mapping scattering length density correlations Quantitative dark-field imaging performed
through variation of wavelength and sample to detector distance at set-up parameters as indicated in
Fig. 1b & d and applied to a number of samples as shown in the TI in (d); (a) results of measurement
on a silica powder (Sipernat-310 [36]) combined with SESANS data and a model fit from literature [21];
the inset compares the applied model to the model used for the related powder sample presented in
(b). Results of correlation length scan on two thicknesses of Sipernat-350 [36] powder with model fit
(red line); insert displays the model applied separately to the results for the two different sample
thicknesses before normalization. (c) data and fit for ferrofluid sample (violet squares) and a porous
solid ferromagnet cube (red circles) and corresponding model fits. (d) an attenuation contrast TI
indicating samples with symbols and colors used in the corresponding plots. Note, that where the error
bars are not visible they are lying within the size of the used symbols.
Both powders were measured contained in quartz glass cuvettes with sample thicknesses of
5 mm for both Sipernat-310 and -350 and additionally a 2mm thick sample of Sipernat-350.
Distance scans were performed at wavelengths of 3.5 Å and 5.6 Å from Ls 20 to 250 mm and
+(-)+/ =𝑒12(3-45)
120 to 250 mm, respectively (compare Fig. 1b). The measured visibilities from grating scans
at each of these settings were analysed pixel-wise using our standard Talbot Lau data
reduction software [37]. The normalized visibilities can be written as [17]
(5)
where V0 is the visibility without sample, S is the total small angle scattering cross section and
G(x) is the projected real space correlation function of the microscopic sample structures [17].
The data is further reduced pixel-wise by computing the logarithm and dividing by the sample
thickness t and the wavelength square. The total scattering probability St depends on l2,
which needs to be normalized in order to combine scans at different wavelengths and to
achieve results independent of the utilized beam. Subsequently the data is modelled with
specific projected real space correlation functions G(x) representative of the microscopic
structure investigated. A fitting procedure finally enables to retrieve the respective structural
parameters.
According to Ref. [21] Sipernat-310 is best modelled with a randomly distributed two-phase
medium [38] corresponding to
𝐺𝜉 =-8𝐾5 -8
𝐺𝜉 =𝑒4- 8:
(6)
where K1 is the modified Bessel function of second kind and first order and a represents the
characteristic structure size in the two-phase medium. The model and parameters found
through small angle scattering measurements in Ref.[21], where a = 1.56 µm and S = 0.45, fit
the single grating data as depicted in Fig. 3a. This model, however, did not fit the data of
Sipernat-350, the second powder sample (compare inset Fig. 3a). This correlates with the fact
that the powders are significantly different in their specific structure and surface area, which
for Sipernat-350 is 55 m2/g, compared to 700 m2/g of Sipernat-310 [36]. The two data sets
from two different sample thicknesses of Sipernat-350, Fig. 3b inset, could both be fitted with
the same model
(7)
.
This model describes a simplified limited fractal taking into account an exponential cut-off
[39]. The two data sets after normalization with their respective thicknesses coincide well
(Fig. 3b). The fit yields a characteristic length scale a = 7 µm and a = 0.5, with a being related
to the structure of the phase boundary. Here a ranges in the lower domain of 0< a <1 which
corresponds to open and branched distributions with high specific surfaces.
Additional data on well characterized samples returned the porosity on the probed length
scale of a ferrite magnet produced through sintering as well as the poly-disperse structure
size distribution in a ferrofluid (Fig. 3c). The cubic ferrite sample of 10 mm side length
consisting of Fe2O3 (86%) and SrCO3 (14%) is revealed to have significant porosity in the
probed size range constituting about 1% of the volume. The pores are characterized with a
poly-disperse size distribution with predominant sizes of 20 and 210 nm diameter. Hard
sphere models [17] were used to approximate the pore structure, and remaining deviations
can be assumed to be due to deviation of average pore shape from the spherical model. The
volume fraction of the smaller pores is found to be more than a factor two lower than the
one of the dominating 200 nm pores constituting a volume fraction of 0.6%. In addition,
microscopy reveals, that pores can also be found on the scale of several micrometers, which
is, however, beyond the probed length scale and does not influence the presented results.
The ferrofluid can be described by a random two-phase medium (Eq. 6) consisting of a
polydisperse particle distribution (inset Fig. 3c) suspended in a liquid. Analyses provides a
characteristic size parameter of a = 250 nm, corresponding to the smallest particles found in
the liquid. In addition, the largest length scale derived from model fitting coincides with the
largest particle dimension in the ferrofluid of about 3 µm (Fig. 3c). The deviations of the
measured points from the fit at the lowest probed correlation lengths are understood to be
due to incoherent scattering from the hydrogenous liquid phase.
We conclude, that our most simple approach of a single attenuation grating with sub-
millimeter period that requires no specific and sensitive alignment, does overcome stability
requirements, design and fabrication issues related to fine structures as well as other
limitations of interferometers. In particular the intrinsically achromatic nature of our
technique enables straightforwardly multi-wavelength and polychromatic studies without the
corresponding drawbacks of interferometry. In contrast, it provides the capability for
efficient, simple, flexible and quantitative multi-modal imaging and especially phase imaging
in the form of differential phase and dark-field contrast. In addition, the method holds the
potential to straightforwardly cover an unprecedented correlation length range. The
capability to extend the range reported for Talbot Lau interferometers [17-21] by more than
an order of magnitude into the nanometer range has been demonstrated. Flux densities
available even at medium flux sources allow to further increase collimation ratios within
standard imaging settings (Fig. 1). This enables to substantially increase the total covered
correlation length scale range from the nanometer regime into and beyond the typically
assessed micrometer range of neutron Talbot Lau interferometry. In particular a comparison
with recent approaches of far-field interferometers [30] (compare Fig. 1d) suggests that
similar collimation conditions lead to significantly superior performance parameters. This
implies that our much simpler and in contrast non-interferometric and truly achromatic set-
up can provide higher efficiency with at least the same accuracy and range of experiments
without the need for careful alignment, stability and interferometry per se. Thus, the
introduced approach constitutes a paradigm shift for corresponding measurements and
methods.
Furthermore, especially the wider spectral range accessible continuously at high flux
spallation neutron sources generates substantial benefit from the achromatic nature of the
set-up for large range high flux measurements including kinetic studies. Instruments like the
imaging beamline ODIN under construction at the European Spallation Source [40] cover
wavelength ranges of up to one order of magnitude. This implies that a correlation length
range of the same order of magnitude can be probed simultaneously, enabling the spatially
resolved observation of microstructural changes not only in time-dependent but non-uniform
environments such as e.g. in shear fields, flow, inhomogeneous temperature, pressure and
magnetic fields in full field observations. This paves the way to foster entirely new analytical
capabilities for complex inhomogeneous and non-equilibrium structural states in materials
either opaque to or not providing sufficient contrast with other types of radiation.
An extension to 2-dimensional phase contrast sensitivity and resolution is similarly straight
forward through the utilization of 2D absorption patterns in contrast to recent elaborate
approaches with interferometry [27,41]. This enables to additionally study micro-structural
anisotropies without the need of several sample scans.
Extending this approach to x-rays is expected to be similarly rewarding because it is suited to
profit from the superior phase space densities available from x-ray sources. These allow
projecting small periods at significant distances, thus covering an outstanding correlation
length scale range.
References
[1] D. B. Murphy, Fundamentals of Light Microscopy and Electronic Imaging (Wiley, New
York, 2001).
[2] E. G. Ramberg, Phase contrast in electron microscope images. J. Appl. Phys. 20, 441 -- 444
(1949).
[3] J. Als-Nielsen and D. McMorrow, Elements of Modern X-ray Physics (Wiley, 2011).
[4] N. Kardjilov, I. Manke, A. Hilger, M. Strobl, and J. Banhart, Neutron imaging in material
science, Materials Today, 14, 6 (2011).
[5] F. Pfeiffer, C. Grünzweig, O. Bunk, G. Frei, E. Lehmann, and C. David, Phys. Rev. Lett. 96,
215505 (2006).
[6] H. Rauch and S. A. Werner, Neutron Interferometry: Lessons in Experimental Quantum
Mechanics, Wave-Particle Duality, and Entanglement, 2nd ed. (Oxford University Press,
Oxford, UK, 2015), Vol. 12.
[7] B. E. Allman, P. J. McMahon, K. A. Nugent, D. Paganin, D. L. Jacobson, M. Arif, and S. A.
Werner, Nature 408, 158 (2000).
[8] W. Treimer, M. Strobl, A. Hilger, C. Seifert, and U. Feye-Treimer, Refraction as imaging
signal for computerized (neutron) tomography, Appl. Phys. Lett. 83, 2 (2003).
[9] C. Grünzweig, C. David, O. Bunk, M. Dierolf, G. Frei, G. Kühne, J. Kohlbrecher, R. Schäfer,
P. Lejcek, H. Rønnow, and F. Pfeiffer, Neutron decoherence imaging for visualizing bulk
magnetic domain structures, Phys. Rev. Lett. 101, 025504 (2008).
[10] I. Manke, N. Kardjilov, R. Schäfer, A. Hilger, M. Strobl, M. Dawson, C. Grünzweig, G.
Behr, M. Hentschel, C. David, A. Kupsch, A. Lange, and J. Banhart, Three-dimensional
imaging of magnetic domains, Nature Commun. 1, 125 (2010).
[11] T. Reimann, S. Mühlbauer, M. Schulz, B. Betz, A. Kaestner, V. Pipich, P. Böni, and C.
Grünzweig, Visualizing the morphology of vortex lattice domains in a bulk type-II
superconductor, Nature Communications 6, 8813, (2015).
[12] B. Betz, P. Rauscher, R. P. Harti, R. Schäfer, H. Van Swygenhoven, A. Kaestner, J. Hovind,
E. Lehmann, and C. Grünzweig, Frequency-Induced Bulk Magnetic Domain-Wall Freezing
Visualized by Neutron Dark-Field Imaging, Phys. Rev. Applied 6, 024024 (2016).
[13] P. Rauscher, B. Betz, J. Hauptmann, A. Wetzig, E. Beyer, and C. Grünzweig, The
influence of laser scribing on magnetic domain formation in grain oriented electrical steel
visualized by directional neutron dark-field imaging, Scientific Reports 6, 38307 (2016).
[14] R. P. Harti, M. Strobl, R. Schäfer, N. Kardjilov, A. S. Tremsin, and C. Grünzweig, Dynamic
volume magnetic domain wall imaging in grain oriented electrical steel at power frequencies
with accumulative high-frame rate neutron dark-field imaging, Sci. Reports 8, 1 (2018).
[15] M. Strobl, C. Grünzweig, A. Hilger, I. Manke, N. Kardjilov, C. David, and F. Pfeiffer,
Neutron dark-field tomography Phys. Rev. Lett. 101, 123902 (2008).
[16] A. Hilger, N. Kardjilov, T. Kandemir, I. Manke, and J. Banhart, D. Penumadu, A. Manescu,
and M. Strobl, Revealing micro-structural inhomogeneities with dark-field neutron imaging,
J. Appl. Phys. 107, 036101 (2010).
[17] M. Strobl, General solution for quantitative dark-field contrast imaging with grating
interferometers, Scientific Reports 4, 7243 (2014).
[18] M. Strobl, B. Betz, R. P. Harti, A. Hilger, N. Kardjilov, I. Manke, and C. Gruenzweig,
Wavelength dispersive dark-field contrast: micrometer structure resolution in neutron
imaging with gratings, J. Appl. Cryst. 49 (2016).
[19] R. P. Harti, M. Strobl, B. Betz, K. Jefimovs, M. Kagias, and C. Gruenzweig, Sub-pixel
correlation length neutron imaging: Spatially resolved scattering information of
microstructures on a macroscopic scale, Scientific Reports 7:44588 (2017).
[20] M. Strobl, R. P. Harti, C. Grünzweig, R. Woracek, and J. Plomp, Small Angle Scattering in
Neutron Imaging -- A Review, J. Imaging 3, 64 (2017).
[21] R. P. Harti, J. Valsecchi, P. Trtik, D. Mannes, C. Carminati, M. Strobl, J. Plomp, C. P. Duif,
and C. Grünzweig, Visualizing the heterogeneous breakdown of a fractal microstructure
during compaction by neutron dark-field imaging, Scientific Reports 8, 17845 (2018).
[22] F. Pfeiffer, M. Bech, O. Bunk, P. Kraft, E. F. Eikenberry, Ch. Brönnimann, C. Grünzweig
and C. David, Hard-X-ray dark-field imaging using a grating interferometer, Nat Mater. 7, 2
(2008).
[23] S. W. Lee, D. S. Hussey, D. L. Jacobson, C. M. Sim, and M. Arif, Development of the
Grating Phase Neutron Interferometer at a Monochromatic Beam Line, Nuclear Instruments
and Methods in Physics Research A 605 (2009).
[24] S. W. Lee, Y. K. Jun, and O. Y. Kwon, A Neutron Dark-field Imaging Experiment with a
Neutron Grating Interferometer at a Thermal Neutron Beam Line at HANARO, J. Korean Phy.
Soc. 58, 730 (2011).
[25] T. Reimann, S. Mühlbauer, M. Horisberger, B. Betz, P. Böni and M. Schulz, The new
neutron grating interferometer at the ANTARES beamline: design, principles and
applications, J. Appl. Cryst. 49, 1488 (2016).
[26] Y. Seki, T. Shinohara, J.D. Parker, W. Ueno, T. Samoto, W. Yashiro, A. Momose, Y. Otake
and Y. Kiyanagi, Efficient phase imaging using wavelength-resolved neutron Talbot-Lau
interferometry with TOF method, EPL 123(1):12002 (2018).
[27] Y. Kim, J. Kim, D. Kim, D.S. Hussey, and S.W. Lee, Feasibility evaluation of a neutron
grating interferometer with an analyzer grating based on a structured scintillator, Rev. Sci.
Instr. 89, 033701 (2018).
[28] M. Strobl, M. Sales, J. Plomp, W. G. Bouwman, A. S. Tremsin, A. Kaestner, C. Pappas,
and K. Habicht, Quantitative Neutron Dark-field Imaging through Spin-Echo Interferometry,
Scientific Reports 5, 16576 (2015).
[29] M. Sales, J. Plomp, K. Habicht, A.S. Tremsin, W.G. Bouwman, and M. Strobl,
Wavelength-Independent Constant Period Spin-Echo Modulated Small Angle Neutron
Scattering, Rev. of Sci. Instr. 87(6):063907 (2016).
[30] D. A. Pushin, D. Sarenac, D. S. Hussey, H. Miao, M. Arif, D. G. Cory, M. G. Huber, D. L.
Jacobson, J. M. LaManna, J. D. Parker, T. Shinohara, W. Ueno, and H. Wen, Far-field
interference of a neutron white beam and the applications to noninvasive phase-contrast
imaging, Phys. Rev. A 95, 043637 (2017).
[31] M. Saxena, G. Eluru, and S.S. Gorthi, Structured illumination microscopy, Adv. Opt.
Photon. 7, 241-275 (2015).
[32] S. R. P. Pavani, A. R. Libertun, S. V. King, and C. J. Cogswell, "Quantitative structured-
illumination phase microscopy," Appl. Opt. 47, 15 -- 24 (2008).
[33] A. Kaestner, S. Hartmann, G. Kühne, G. Frei, C. Grünzweig, L. Josic, F. Schmid, and E.H.
Lehmann, Nucl. Instr. Meth. A 659, 1 (2011).
[34] M. Sales, J. Plomp, W. G. Bouwman, A. S. Tremsin, K. Habicht and M. Strobl, On the
analysis of time-of-flight spin-echo modulated dark-field imaging data, J. Phys.: Conf. Ser.
862, 012026 (2017).
[35] N. Bevins, J. Zambelli, K. Li, Z. Qi, and G.-H. Chen, Multicontrast x-ray computed
tomography imaging using Talbot-Lau interferometry without phase stepping, Med. Phys.
39, 1 (2012).
[36] http://www.thecarycompany.com/media/pdf/specs/tds-evonik-sipernat-310.pdf and
http://www.thecarycompany.com/media/pdf/specs/tds-evonik-sipernat-350.pdf (2018)
[37] R. P. Harti and J. Valsecchi, TaPy 0.2 - https://github.com/nGImagic/TaPy/releases,
10.5281/zenodo.1252995, 2017.
[38] R. Andersson, W. G.Bouwman, S. Luding, and I.M. de Schepper, Structure in cohesive
powders studied with spin-echo small angle neutron scattering, Granular Matter 10, 6
(2008).
[39] R. Andersson, L. F. van Heijkamp, I. M. de Schepper, and W. G. Bouwman, Analysis of
spin-echo small-angle neutron scattering. J. Appl. Cryst. 41, 868 (2008).
[40] M. Strobl, The Scope of the Imaging Instrument Project ODIN at ESS, Physics Procedia
69 (2015).
[41] M. Kagias, Z. Wang, P. Villanueva-Perez, K. Jefimovs, and M. Stampanoni, 2D-
Omnidirectional Hard-X-Ray Scattering Sensitivity in a Single Shot Phys. Rev. Lett. 116,
093902 (2016).
|
1901.09452 | 2 | 1901 | 2019-01-29T20:45:40 | Bayesian Inference-enabled Precise Optical Wavelength Estimation using Transition Metal Dichalcogenide Thin Films | [
"physics.app-ph"
] | Despite its ability to draw precise inferences from large and complex datasets, the use of data analytics in the field of condensed matter and materials sciences -- where vast quantities of complex metrology data are regularly generated -- has remained surprisingly limited. Specifically, such approaches could dramatically reduce the engineering complexities of devices that directly exploit the physical properties of materials. Here, we present a cyber-physical system for accurately estimating the wavelength of any monochromatic light in the range of 325-1100nm, by applying Bayesian inference on the optical transmittance data from a few low-cost, easy-to-fabricate thin film "filters" of layered transition metal dichalcogenides (TMDs) such as MoS2 and WS2. Wavelengths of tested monochromatic light could be estimated with only 1% estimation error over 99% of the stated spectral range, with lowest error values reaching as low as a few ten parts per million (ppm) in a system with only eleven filters. By step-wise elimination of filters with the least contribution toward accuracy, mean estimation accuracy of 99% could be obtained even in a two-filter system. Furthermore, we provide a statistical approach for selecting the best "filter" material for any intended spectral range based on the spectral variation of transmittance within the desired range of wavelengths. And finally, we demonstrate that calibrating the data-driven models for the filters from time to time overcomes the minor drifts in their transmittance values, which allows using the same filters indefinitely. This work not only enables the development of simple cyber-physical photodetectors with high accuracy color-estimation, but also provides a framework for developing similar cyber-physical systems with drastically reduced complexity. | physics.app-ph | physics |
Bayesian Inference-enabled Precise Optical
Wavelength Estimation using Transition Metal
Dichalcogenide Thin Films
Davoud Hejazi1, Shuangjun Liu2, Sarah Ostadabbas2, and Swastik Kar1
2Augmented Cognition Lab, Electrical and Computer Engineering Department
Northeastern University, Boston, USA
1Department of Physics
Abstract. Despite its ability to draw precise inferences from large and
complex datasets, the use of data analytics in the field of condensed mat-
ter and materials sciences -- where vast quantities of complex metrology
data are regularly generated -- has remained surprisingly limited. Specif-
ically, such approaches could dramatically reduce the engineering com-
plexities of devices that directly exploit the physical properties of mate-
rials. Here, we present a cyber-physical system for accurately estimating
the wavelength of any monochromatic light in the range of 325 -- 1100nm,
by applying Bayesian inference on the optical transmittance data from
a few low-cost, easy-to-fabricate thin film "filters" of layered transition
metal dichalcogenides (TMDs) such as MoS2 and WS2. Wavelengths of
tested monochromatic light could be estimated with only 1% estima-
tion error over 99% of the stated spectral range, with lowest error values
reaching as low as a few ten parts per million (ppm) in a system with
only eleven filters. By step-wise elimination of filters with the least con-
tribution toward accuracy, mean estimation accuracy of ∼99% could be
obtained even in a two-filter system. Furthermore, we provide a statistical
approach for selecting the best "filter" material for any intended spectral
range based on the spectral variation of transmittance within the desired
range of wavelengths. And finally, we demonstrate that calibrating the
data-driven models for the filters from time to time overcomes the mi-
nor drifts in their transmittance values, which allows using the same
filters indefinitely. This work not only enables the development of simple
cyber-physical photodetectors with high accuracy color-estimation, but
also provides a framework for developing similar cyber-physical systems
with drastically reduced complexity.
Keywords: 2D materials, layered materials, Bayesian inference, distri-
bution estimation, k-nearest neighbors, liquid-phase exfoliation, machine
learning, semiconductors, transition metal dichalcogenides (TMDs), trans-
mittance, wavelength estimation.
1
Introduction
The ability to perform wavelength-selective photodetection has remained one of
the most exciting areas of research in optoelectronics [14,9], owing to its appli-
2
D. Hejazi, S. Liu, S. Ostadabbas, and S. Kar
cations in advanced photonic circuits and systems [15]. In their simplest form, a
photodetector is a light-sensitive semiconductor that operating in a conductive,
diode, or transistor mode responds by generating a change in a measurable volt-
age or current to an incident light [35,30]. Most conventional semiconductor pho-
todetectors are broad-band, i.e. they respond to a broad range of wavelengths,
and hence are not intrinsically wavelength-selective. To achieve wavelength se-
lectivity in photodetection, in addition to the traditional use of color pre-filters,
various approaches have recently been proposed, such as the use of nanomate-
rials and nanostructures including quantum-dots [19], photonics/plasmonics ar-
rays [16,13], and cavity-based-resonators [20], each with characteristic responses
to specific wavelengths. These and other approaches have paved the way for a
variety of tunable photodetectors capable of responding selectively to incident
light with a specific wavelength [11,34].
The situation is significantly more complex, however, when a detection sys-
tem has to identify the wavelength of any incident light (and not just a specific
one). Such systems, that are capable of accurately discerning the wavelength
of incident light, have immense relevance for applications such as bionic vision
[32,31,22], robotic vision [6], and various industrial light detection [26,7,23,12], as
well as astronomical and military applications [17,24]. Typically, a wavelength
estimating system (e.g. in spectrometers) uses either a large number of pho-
todetectors or an intricate diffraction-grating based monochromator coupled to
one or two photodetectors to perform the task. To appreciate their complexity,
let us consider a photodetection system that is required to estimate the wave-
length of monochromatic light between the range of 325 -- 1100nm. A number of
notch/band-pass filters can be used to achieve this, depending on the desired
resolution [10,28]. For example, if wavelength estimation is desired with 1nm ac-
curacy, it will require a complex design with several hundred photodetectors with
1nm-width notch filters to achieve arbitrary wavelength identification. Beyond
the design complexity of such a system, developing such narrow-width notch
filters for each nanometer range could be a significant engineering challenge
by itself. Alternately, the incident light could be diffracted off via a diffraction
grating, but it will still require the same large number of detectors placed in
an array to obtain the desired precision. Other approaches, such as rotating
gratings, would lead to cumbrous addition of electronic and mechanical parts.
In other words, achieving high accuracy wavelength selectivity using a purely
mechanical set of detection systems can be complex, bulky, and expensive.
We show that by using a few easy-to-fabricate nanomaterial-based broad-
band thin film filters, and harnessing sophisticated statistical approaches on
large training datasets, it is possible to dramatically reduce the physical com-
plexity of an accurate wavelength estimator. Nanomaterials, due to their diverse
electronic and optical properties are constantly being explored and used for va-
riety of low-cost, sensitive, and scalable photodetection technologies [2]. In this
context, transition metal dichalcogenides (TMDs) are considered to be among
the leading candidates in sensing applications. TMD monolayers are atomically
thin semiconductors of the type M X2, with M a transition metal atom (Mo,
Bayesian Inference-enabled Precise Optical Wavelength Estimation
3
W, etc.) and X a chalcogen atom (S, Se, or Te). One layer of M atoms is sand-
wiched between two layers of X atoms [33]. TMD monolayers of M oS2, W S2,
M oSe2, W Se2, and M oT e2 have a direct band gap, and can be used in elec-
tronics as transistors and in optics as emitters and detectors [29,25,4]. In this
work, we used nanoscale TMDs to develop thin film broadband optical "filters".
Although these are not monolayer TMDs, their thin films provide wide regions
of variation of optical transmittance, and the broadband optical responses from
just a few filters turn out to be far more useful in wavelength identification than
using a large array of "notch" filters, as justified next.
In addition to the previously discussed complexity and cost issues, all of the
previously mentioned traditional methods share a common limitation: Except
for the readings corresponding to a specific filter or detector that senses the
incident wavelength, the data from the rest of the detectors/measurements are
usually discarded. This loss is an inefficient use of available data, especially since
the appropriate use of data science provides ways to harness all available data
to substantially increase the estimation accuracy from large datasets. Among
the more powerful estimation algorithms is Bayesian inference, which is a theory
in the field of statistics based on the Bayesian interpretation of the probability
where probability expresses a degree of belief in an event, which can change as
new information is gathered, rather than a fixed value based upon frequency or
propensity [5,21]. Bayesian inference uses Bayes' theorem to compute and update
probabilities after obtaining new data. Bayes' theorem describes the conditional
probability of an event based on the gathered data as well as prior information or
beliefs about the event or conditions related to the event. Since Bayesian infer-
ence treats probability as a degree of belief, Bayes' theorem can directly assign a
probability distribution to a parameter or set of parameters that quantifies the
belief in them [5,21].
In this study, we have utilized the Bayesian inference approach to show that
is possible to exploit the wavelength-dependence of broadband optical trans-
mittance of simple thin film optical filters (built using low-cost, liquid-phase
exfoliated TMDs) to accurately estimate the wavelength of any monochromatic
color (i.e. the midpoint of a very narrow-band light source with ∼1nm width,
over a wide spectral range, 325nm< λ <1100nm). To the best of our knowl-
edge, this is the first time the efficacy of such a powerful statistical analysis is
being utilized to "train" a set of physical sensing systems to provide high ac-
curacy estimation of "test" light sources, thereby developing the world's first
cyber-physical monochromatic color estimator. By using Bayesian inference on
optical transmittance data of up to 11 of such filters, the wavelengths of "test"
monochromatic light could be estimated with less than 0.1% estimation error
for 71% of the spectrum, and less than 1.5% error for the rest of the spectrum.
Furthermore, it is shown that even though using data from all available 11 filters
yields to the smallest estimation errors in general, a Greedy selection algorithm
[8] could be applied to reduce the number of filters and complexity of the sys-
tem while keeping the estimation accuracy in an acceptable range. Our selection
algorithm progressively discards the filters that have least contribution towards
4
D. Hejazi, S. Liu, S. Ostadabbas, and S. Kar
the accuracy of the wavelength estimation, hence achieves the lowest number of
filters for a given acceptable accuracy. The proposed algorithm would allow in
principle to perform filter selection on the desired specific spectrum ranges and
find the optimal filter combinations that work well on those ranges, which might
be different in other ranges.
A remarkable outcome of our investigation is how such high accuracies could
be achieved from a relatively simple photodetection system, as enabled by the
advanced data analytics. The physical part of our design required only two types
of TMDs (M oSe2 and W S2) to fabricate all of the 11 filters with sufficient filter-
to-filter variations, while liquid-phase exfoliation technique (which was used to
fabricate the thin film filters) can be considered to be one of the simplest fabri-
cation approaches. Since the efficacy of the estimation was found to be driven by
the monotonic, i.e. single-valued nature of the spectral transmittance (or, "filter
functions") and not on their specific values and other variations, their fabrica-
tion remained simple, scalable, and low-cost, requiring very little process-control.
Moreover, by re-calibrating these filters from time to time, it was possible to re-
tain the high accuracy wavelength estimation ability over extended period of
time. This aspect of our investigation addresses a huge challenge in many de-
vices fabricated using nanomaterials, whose properties often degrade with time,
rendering them impractical for real-world applications. In the following sections,
we discuss how these systems were built and characterized, outline the statisti-
cal data analysis employed for wavelength estimation, and present details of the
functional efficacy of our cyber-physical sensor.
2 Results and Discussion
Filter Design and Transmittance Data. The estimation of wavelengths was
done by fabricating a set of eleven thin film optical filters using a combina-
tion of two TMDs nanoflakes, in order to get sufficient variations in their filter
functions. Fig. 1(a,b) show digital images of all the filters, including microscope
images from three representative filters. Two out of the eleven filers (which we
labeled as f1 and f11) were made by drop-casting suspensions of liquid-phase ex-
foliated nanoflakes of the two types of TMDs, M oS2 and W S2, respectively, onto
the surface of two separate glass slides. The other nine filters were made by drop-
casting suspensions with gradually differing mixing proportions of the same two
TMDs onto surface of separate glass slides. The transmittance vs. wavelength
of these filters over the 325 -- 1100nm spectrum range was collected 120 times for
each filter, using a Perkin-Elmer Lambda 35 UV-vis-NIR spectrophotometer. In
this process, a broadband light source was converted into variable monochro-
matic light using a diffraction grating system and was made to pass through the
filters, and the transmitted light was measured using a silicon photodetector.
Fig. 1(c) schematically represents the apparatus measuring the transmittance,
where a plain glass slide was used to remove the background transmittance of the
slides. The mean value of the glass-background-subtracted transmittance for all
of the filters are shown in Fig. 1(d). In each case, the overall transmittance values
Bayesian Inference-enabled Precise Optical Wavelength Estimation
5
Fig. 1. (a) Eleven filters drop-casted on glass slide; f1 is 100% W S2, but f2, . . . , f10 are
made by gradually adding M oS2 and decreasing W S2, and finally f11 is 100% M oS2.
(b) Microscopic image of three filters f1, f6, and f11: nanomaterials on glass substrate.
Atomic force microscopy (AFM) and scanning electron microscopy (SEM) images can
be found in the Supporting Information. (c) Schematics of transmittance measurement
as measured inside a Perkin-Elmer Lambda 35 UV-vis-NIR spectrometer. A broadband
light source (a combiation of deuterium and halogen lamps) that contains a spectrum of
different wavelengths passes through a diffraction grating based monochromator. The
monochromator isolates a narrow-band portion of the spectrum. This beam is split
in two, passing through filter position and reference position, the beams are incident
on photodetectors. A pure glass slide is placed in reference position and its spectral
transmittance is removed from the total transmittance. (d) Background-subtraction
transmittance vs. wavelength for all 11 filters. The excitonic peaks get modified grad-
ually from f1 to f11 as a results of changing proportion of mixing two TMDs.
were found to grow with increasing wavelength of incident light corresponding
to the gradually reducing density of states near the Fermi level of these materi-
als with growing wavelength (or decreasing energy values) tending towards zero
close to the band gap. In addition, there are characteristic "dip" features that
correspond to various excitonic resonances in these systems [18]. Mixing two
different TMDs in different amounts enabled us to get gradually evolving trans-
mittance curves with changing magnitudes, slopes, and feature positions. As we
show, this allowed us to examine which features of the transmittance curve were
responsible for higher wavelength estimation confidences. We next discuss how
the data obtained using these filters were analyzed using the Bayesian inference.
Wavelength Estimation Using Bayesian Inference: The statistical anal-
ysis of our data were performed over a set of transmittance values measured
discretely over the entire range of wavelengths, for each filter, as well as 120
repetitions of wavelength-dependent data. The repeated data was acquired to
account for drifts, fluctuations, and other variations commonly observed in phys-
ical measurements especially in nanomaterial-based systems, which tend to be
sensitive to their environments. Using this data referred to as our "training
(b)Concentration of MoS2Concentration of WS270 µm(a)1100 nm800 nm550 nm300 nmDiffraction Grating based Monochromatorλj, I0λj, ItFilter fKGlassReferencePhotodetectorSample Beam Photodetector(c)300400500600700800900100011000.800.820.840.860.880.900.920.940.960.981.00(d) Transmitance of LightWavelength (nm)Transmitance f1 f2 f3 f4 f5 f6 f7 f8 f9 f10 f11 6
D. Hejazi, S. Liu, S. Ostadabbas, and S. Kar
data", we formulated the wavelength estimation problem as follows: Let λ =
{λ1, ..., λi, ..., λN} be N different wavelengths in desired spectral range and
with specified granularity (i.e. 325 -- 1100nm with 1nm step in this study), and
T = {t1, ..., ti, ..., tK} be the transmittance vector of K filter values (i.e. K = 11
when all of the filters are used in this study). Employing the Bayesian inference,
the probability of the monochromatic light having the wavelength λj based on
the observed/recorded transmittance vector T data is called posterior proba-
bility P (λj T ), which is the probability of a hypothesis given the observed
evidence:
P (λj T ) =
P (T λj)P (λj)
P (T )
,
(1)
where P (λj) is the prior probability of the specific wavelength λj being present in
the monochromatic light of interest, which is defined as the estimate of the prob-
ability of the hypothesis before the current evidence is observed. In this study,
we assumed all of the wavelengths are equally-likely to happen, so we considered
a uniform distribution function for the prior probability as P (λj) = 1
N , where
N is the total number of quantifiable wavelengths in the range under study.
Moreover, P (T λj) is the probability of observing transmittance data T given
wavelength λj, and is called the likelihood, which indicates the compatibility of
the evidence with the given hypothesis. Although, the filters are related due to
having the same two materials with different mixtures, for computational pur-
pose, we assume independence between their outcomes, and model them using
with Naive Bayes algorithm [27]. As such, the likelihood of all filter readings T
can be calculated as the product of each filter value ti in a given wavelength
k=1 P (tk λj). To compute individual P (tk λj) values,
a Gaussian normal distribution for each filter at each wavelength was assumed,
and their mean values and standard deviations were calculated from the training
data (i.e. the 120 measured transmittance spectra) collected from each filter at
each wavelength. P (T ) is called marginal probability of measured transmittance
i=1 P (T λi)P (λi). Since P (T )
is the same for all possible hypotheses that are being considered, so acts as a
normalization factor to keep the posterior probability in the range of 0 to 1.
Finally, given the measured transmittance sample T (a vector of K elements --
one transmittance value per filter at an unknown wavelength, see Section 4 for
more details), the target wavelength λ∗ of the monochromatic light is estimated
by choosing the value of λj that maximizes the posterior probability P (λj T ):
λj, as P (T λj) = (cid:81)K
vector T , which can be calculated as P (T ) =(cid:80)N
λ∗ = arg max
λj
P (λj T ),
(2)
in which this optimisation called the maximum a posteriori (MAP) estimation
[3,5,21].
The efficacy of our wavelength estimator was tested both using test samples,
i.e transmittance value for a test monochromatic source that were collected sep-
arately, and which were not used in the training data and hence were not seen by
Bayesian Inference-enabled Precise Optical Wavelength Estimation
7
the model before; as well as for training samples which were generated randomly
from the same Gaussian distributions that were assigned to each wavelength
for each filter. The training samples were utilized to check how well the model
works on the training set itself, while the test samples are used to investigate
how the model can estimate the truly unknown wavelengths. Fig. 2 provides an
typical example of applying Bayesian inference for estimating the wavelengths.
The Fig. 2(a) plots 30 instances of overlapping transmittance spectra of the filter
f1, with the inset showing a magnified view of the transmittance data set for
a single wavelength (shown here for λ = 700 nm), the collected transmittance
values creating a distribution around the mean value of transmittance for that
wavelength. Fig. 2(b) shows the histogram of the same 30 recorded transmittance
data of f1 at 700nm with the calculated mean µ and standard deviation σ. The
red curve shows the Gaussian fit on the data, which justifies the assumption of a
normal distribution for the P (tk λj) probabilities. In order to perform Bayesian
inference for wavelength estimation of a test monochromatic light, we needed to
calculate the posterior probabilities of different wavelengths P (λj T ), when
transmittance data T is collected from our 11 filters. When testing the efficacy
of our wavelength estimator, a new "test transmittance data" set is collected
separately from the training transmittance data. Fig. 2(c) shows the posterior
probability as a function of wavelength. It is apparent that the maximum pos-
terior probability is close to 1 around 700nm, and is almost zero for the rest of
the spectrum, which indicates the reliability of the Bayesian inference. Several
other cases are presented in Supporting Information. The same procedure was
performed to estimate all of the wavelengths (test and training).
Wavelength Estimation Accuracy To discuss the efficacy of our wave-
length estimator, we define the estimation error as difference between average
estimated wavelength and real wavelength, divided by the real wavelength, times
100 to find average estimation error percentage. The percentage average estima-
tion error (when using all 11 filters) is plotted as a function of wavelength in
Fig. 3(a). In this figure, we plot the estimation error from both the "training"
(in green) as well as the "test" (in black) data. We see that using only eleven
filters and two photodiodes, our cyber-physical color estimator is able to achieve
extremely high accuracy color-estimation, with just a very few estimation data
points lying above 1% error, with most of the error values being far lower in com-
parison. To appreciate how high the accuracy values reached, the semi-log plot
of the same data has been shown in the inset. We find that not only a significant
portion of the estimated wavelengths from the training dataset was better than
0.1%, the lowest errors are arrive close to 0.001%, or a few tens parts per million.
In particular, the test data, which could be performed only on a smaller set of
source wavelengths, appear to fall well-within these low-error accuracies. This
fact that errors for the estimation of the test samples are well-within the range of
training errors, highlights an extremely important feature of our cyber-physical
wavelength estimator, i.e. the efficacy of our Bayesian inference approach is gen-
eralizable from training set to test data. In other words, a lab-trained system
is very likely to continue providing high-confidence wavelength estimations un-
8
D. Hejazi, S. Liu, S. Ostadabbas, and S. Kar
Fig. 2. (a) Transmittance spectra of f1 measured 30 separated times, such each point
on the curve is actually 30 dots as shown for 700nm in the inset. The mean value and
standard deviation for each filter at each wavelength is calculated from this repeated
training data. (b) Histogram of the same 30 transmittance data of f1 at 700nm. The
red curve shows a Gaussian function fit to the data. The parameters of the Gaussian
distribution for each filter at each wavelength (i.e. their µ's and σ's) are found using the
training data. (c) The posterior probability calculated using Bayesian inference applied
on transmittance data collected from our 11 filters when a test 700nm monochromatic
light was shine on them. The wavelength with maximum posterior is chosen as the
estimated wavelength λ∗, which is equal to 700nm in this case.
der field-testing as well. We conclude that our cyber-physical approach using 11
filters and the Bayesian inference system is not only able to estimate unknown
wavelengths with a high degree of accuracy, but also do so with equal efficacy
under both training and testing conditions.
In order to compare the estimation results of our Bayesian inference with
another data-driven approach, we chose the k-nearest neighbor (k-NN) as one
of the most straightforward machine learning algorithm, which is widely used in
supervised classification applications [1]. For training and testing the k-NN, we
used the same training and testing datasets we used in our Bayesian inference,
respectively. Using the k-NN method, the wavelengths that had nearest trans-
mittance values to the test samples were found, and the result for test data from
all 11 filters has been plotted in Fig. 3(b). It is noted that while in some in-
stances, the k-NN approach provides estimations of nearly the same accuracy as
the Bayesian approach, overall the estimation accuracy with the Bayesian meth-
ods is superior over the entire spectral range, especially at the lower wavelength
values where the deviation of estimation from the real values are larger. For
this reason, we present Bayesian inference as the the primary analysis approach
for the rest of this study. We next investigate, in a step-wise manner, how the
wavelength estimation efficacy changes as the number of filters are reduced.
(a)(b)= 700nm(c) Bayesian Inference-enabled Precise Optical Wavelength Estimation
9
Fig. 3. (a) Average training (in green) and test (in black) wavelength estimation error
percent of Bayesian inference using all 11 filters. The inset is semi-log plot of the same
figure. Each data point is averaged over 100 estimated values. (b) Average test wave-
length estimation error percent by applying Bayesian inference (in black) compared to
k-NN algorithm (in red) using 11 filters.
Filter Selection and its Effect on Estimation Accuracy. A key ad-
vantage of our cyber-physical system is that its ultimate estimation accuracy
depends on both the efficacy of the Bayesian inference approach as well as the
total number of filters used. In other words, if such high accuracy is not required
for any specific application, it is possible to further reduce the physical complex-
ity of the system. With all the training data available to us, it was possible to
investigate the estimation accuracy of the system by identifying and removing
the filters that were least effective, in a step-by-step manner. Understandably,
by using a fewer number of filters for estimation, the error tends to increase.
The estimation error vs. wavelength when using only 1, 2, or all 11 filters are
shown in Fig. 4(a). We found out that using only two or one filter(s), the highest
estimation errors grows by a factor of ∼5 and ∼25, respectively, at the most
error-prone region between 300 -- 500nm. In most of the remaining parts of the
spectrum, the estimation error remains much lower, as seen more clearly in the
the inset that shows a semi-log plot of the test and training error when using
filters f1 and f11. Further, starting with all 11 filter functions, using the Greedy
algorithm [8] as a filter selection approach, the number of filters could be sequen-
tially reduced, by discarding the filters with least contribution towards accuracy
of estimation one by one. This way, the complexity of the system could be re-
duced systematically, while minimizing the cost of reducing overall accuracy at
each step. Fig. 4(b) represents the results of this filter selection. Here, on the left
side of the representation, each cell represents a filter function being used (gray
box) or discarded (white space). Starting from the top, where all the filters are
present, the Greedy algorithm was used to drop the least useful filter and this
is represented in the next row. In this way, in each row, the least useful filter is
300400500600700800900100011000.00.20.40.60.81.01.21.41.61.82.02.22.42.640060080010001E-30.010.11 Average % error in estimationWavelengths (nm)Bayesian, with 11 filters training error test error(a) Log10(% Error)Wavelengths (nm)300400500600700800900100011000.00.51.01.52.02.53.03.54.0 Average % error in estimationWavelength (nm)Test error with 11 filters k-NN Bayesian(b) 10
D. Hejazi, S. Liu, S. Ostadabbas, and S. Kar
Fig. 4. (a) Average training error percent of Bayesian estimation using one filter (f1),
two filters (f1 and f11) and all filters (see text for definition of error percent). The
inset semi-log plot of training and test percent errors when only two filters are used
for estimation. Each data point is averaged over 100 estimations (see Section 4). (b)
Filter selection via Greedy algorithm: horizontal dark-grey blocks represent the vector
of chosen filters to use, and corresponding blue-purple pair of bars on the right show
the mean average test or training error when the chosen filters are used for Bayesian
estimation.
dropped and the mean value of the average error (from the entire spectrum) is
plotted as a horizontal bar-graph on the right end of the row, for both training
and test errors. We found out the very encouraging result that the mean average
error (presented in nanometers instead of percentage values) does not change
much until it gets to the last few filters, suggesting that the filter-to-filter vari-
ation of transmittance functions using our simple approach provides was quite
effective. Indeed, even with two filters the error is significantly small, with an
effective average error of ∼6nm, which reflects less than ∼1% error at the center
of the spectrum. Applying this feature selection method when only two filters are
desired reveals that the filters f1 and f11 would give the best wavelength estima-
tion results, which was expected because these two were the most independent
filters being fabricated using completely independent nanomaterials, while the
other filters are mixtures of both materials. The increase in error is more ob-
vious when switching from 2 to 1 filter, which establishes that a single-filter
photodetector would not be enough for reliable wavelength estimation.
Sources of Estimation Error. We next discuss factors that affect the ac-
curacy of estimation as related to the curve-shapes of the transmission functions.
There is an interesting correlation between the positions (wavelength values) of
local maxima/minima of transmittance curves (which arise from variations of
the density of state and presence of excitonic peaks, fairly well-known features
of the spectral absorption curves of TMDs [18]), as seen in Fig. 1(d)), and where
the errors tend to increase. Large error occurs across multi-valued regions of
transmittance curve, i.e. regions where multiple wavelengths may have same or
very similar transmittance values. The estimations using only one filter is shown
Mean Average Error (nm) 0.02.04.06.08.010.012.014.016.0mean average test errormean average training errorf1f2f3f4f5f6f7f8f9f10f11c #11 c #10c #9c #8c #7c #6c #5c #4c #3c #2c #1 Selected FiltersCombination(a)(b) Bayesian Inference-enabled Precise Optical Wavelength Estimation
11
Fig. 5. (a) Average training error percent of Bayesian estimation when using only
f1 (in dark-blue) and transmittance of the same filter (black curve). Vertical dashed-
lines indicate that co-existing different wavelengths with same transmittance values in
the same neighborhood leads to inaccuracy in estimating wavelength; but when the
transmittance is single-valued the error is small. The two multi-valued regions have
obvious effects in increased error. Apart from the main single-valued region that is
shown, two more of such regions exist between 500 -- 600nm. The flat region causes
only small increase in error. (b) Selected range of previous figure showing 665 -- 1100nm
that has monotonic (increasing only) transmittance (black curve). Also plotted in the
same range is the first derivative (slope) of the transmittance vs. wavelength (in blue)
is the percentage estimation error (in pink). As the derivative of transmittance be-
comes smaller (decreasing slope), the errors become larger. For better visualization,
the normalized RMS error% is given in Fig. 5(b) (see Supporting Information for the
equation).
in Fig. 5(a) for simplicity. The horizontal lines between two vertical dashed lines
clearly show when the horizontal lines cut more than once through the transmit-
tance curve, the error becomes larger; but when a horizontal line passes through
only one point, the error is smaller. This result was expected because multi-
ple wavelength will presents similar posterior values in predictions, which are
hence prone to wrong estimations. Thus, materials with monotonic responses
(e.g. without excitonic or other absorption peaks, or in other words with single-
valued spectral transmittances) are better choices for fabricating such filters.
Furthermore, even in the parts of the spectrum that filter function is monotonic
(in this case only increasing) the errors are smaller when the slope (derivative)
of the transmittance curve is larger. Fig. 5(b) which refers only to a part of the
spectrum between 665 -- 1100nm reveals that as the slope (derivative) of trans-
mittance decreases the error increases. For better capturing the the deviations
in estimation visualizing the errors, the root-mean-square (RMS) % error is
used here (see Supporting Information for the equation). From these results, we
conclude that ideal transmittance curves should be monotonic with adequately
changing transmittance values. We note that while conceptually this is not diffi-
300400500600700800900100011000.810.840.870.900.930.960510152025(a) Transmittance of f1 Bayesian, with f1 Wavelengths (nm)single-valued regionmulti-valued regionmulti-valued regionflat regionTransmittanceAverage % error 6507007508008509009501000105011001150-0.020.000.020.040.060.00.30.60.91.2Wavelengths (nm)Derivative, dT/dl (1/nm)Average % error Transmittance of f1 Derivative of f1 Transmittance Bayesian, with f1(b) 12
D. Hejazi, S. Liu, S. Ostadabbas, and S. Kar
Fig. 6. (a) Average test error percent of Bayesian estimation with initial training on
test samples collected at the same time (in black; same points as in Fig. 3) compared
to error on test samples collected after 6 month but still using the initial training set
(in magenta). (b) Average test error percent of Bayesian estimation with the initial
training data on test samples collected at the same time (in black) compared to the
estimations with new calibration (6 month after the first calibration) using new training
and test samples (all collected 6 month after first calibration - in magenta).
cult to understand, in a real-world situation, it is challenging to "pre-order" the
transmittance curves of any material, once again pointing towards the usefulness
of the characteristic transmittance of the TMDs used in this study.
Filter Stability and Reusability Over Time. Finally, we explored the
estimation reproducibility of these easy-to-fabricate physical filters, which would
be an extremely important consideration from a practical viewpoint. These filters
were simply drop-casted onto the surface of regular glass without any additional
protection, and the typical time-lapse between first calibration of filters estima-
tion was 1∼100 days, which demonstrates the physical stability of the filters
despite being left in ambient conditions for 10% of the time and under nominal
vacuum for 90% of the time. Still, gradual change of the optical properties in
these nanomaterials is expected, as they absorb various gaseous species from the
ambient. To check the stability of the filters, six months after the first calibra-
tion, a new test set was collected and was estimated using the original 6 month
old training data. It was interesting to see that while the estimation errors were
found to have become larger for smaller wavelengths values, but for most of the
higher-wavelength portion of the spectrum the estimation error remained better
than 3% as seen in Fig. 6(a).
To see whether the filters are reusable in longer time spans or not, a new
training set also was collected along with the new test set (6 month after original
calibration). Despite some minor changes observable in transmittance (optical
response or filter function) of the filters, by calibrating the filters using the new
set of training data it was possible to estimate the new unknown wavelength
30040050060070080090010001100051015202530354045(b) Average % error in estimationWavelength (nm)Bayesian error % of estimation with 11 filters on initial test samples using initial training new test samples collected after 6 month using initial training300400500600700800900100011000.00.20.40.60.81.01.21.41.61.82.02.22.42.6(b) Average % error in estimationWavelength (nm)Bayesian error % of estimation with 11 filters on initial test samples using initial training new test samples where both training & test data are collected after 6 month Bayesian Inference-enabled Precise Optical Wavelength Estimation
13
as accurate as before (Fig. 6(b)). This, not only suggests fair stability of these
nanomaterial filters but also shows that by calibrating the filters from time to
time, it would be possible to continue using these same filters over extended
periods of time, and the efficacy of estimations does not suffer from wear or
minor scratches, since the calibration will overcome the gradual changes of the
filters.
3 Conclusions
In conclusion, we have successfully demonstrated a new approach that applies
data analytics (i.e. Bayes's theorem) to the optical transmittance information
collected from two extremely low-cost nanomaterial filters to estimate the wave-
length of a narrow-band unknown incident light with high accuracy. Using more
number of filters that are created from the same two nanomaterials it is possible
to considerably improve the accuracy of estimation, and with a feature selec-
tion algorithm the minimum number of filters needed for an acceptable value of
average accuracy can be determined by retaining the only the "most relevant"
filters. Even though the experiment was performed over the range 325 -- 1100nm,
in principle this approach can be extended beyond in both the UV as well as
NIR directions, thereby opening up the possibility of developing next generation
wavelength-estimators for both visible as well as beyond-visible regions of the
EM spectrum. The filters performed robustly even after many months without
additional protection and only low-maintenance storage, and by re-calibrating
the Bayesian inference model used for estimation from time to time it is possible
to continue using these same filters with high accuracy over extended periods
of time. In the ranges of spectrum that filter function (transmittance) has a
monotonic dependence on wavelength, the estimation accuracy is higher, and
furthermore, there is a positive correspondence between slope of filter function
and accuracy of estimation. Hence, based on the application and desired spectral
range, the highest accuracy values will be obtained by using materials (either
TMDs or other transparent films) whose transmittance values show large but
monotonic changes with wavelength. In addition to the Bayesian approach, the
k-NN analysis was also successfully applied, though with comparatively lower
wavelength estimation efficacy. We believe that our findings open up a com-
pletely new path for designing next generation sensors and detectors that can
harness the power of data analytics to reduce the physical complexity of detec-
tors in general, and in particular for future works on generic non-monochromatic
lights using more advanced data analyzing methods and state-of-the-art machine
learning techniques. We believe that this significantly transforms the field of
high-accuracy sensing and detection using simple cyber-physical approaches.
4 Materials and Methods
Sonication-Assisted Liquid-Phase Exfoliation. Bulk M oS2 and powder of
W S2 were purchased from ACS material. Bulk M oS2 was grinded using pestle
14
D. Hejazi, S. Liu, S. Ostadabbas, and S. Kar
and mortar, but the powder of W S2 was used as received. 80mg of M oS2 powder
and 8mL of Isopropyl alcohol (IPA) was added into a beaker and stirred until
the dispersion became dark, then was exfoliated in liquid phase via sonicating
using 30kHz and 80% of the power of a UP100H Hielscher ultrasound processor
for 8hours while the beaker was placed in cool water to avoid overheating all
the time. Afterwards, the dispersion was left still for a minute for the bulk
materials to settle down; supernatant (top half) of the dispersion was collected
and centrifuged for 2 minutes with 1000rpm using a Thermo Scientific centrifuge.
The supernatant (top one-third) of the centrifuged dispersion was moved to
another container and was centrifuged again with 1000rpm for 5 minutes. Finally
the supernatant was collected and stored. The same method was used to produce
more amounts of 2D nanomaterials of both M oS2 and W S2. The schematics of
exfoliation and drop-casting can be found in supporting documents.
Before drop-casting it was useful to know the relative concentration of M oS2
and W S2. For this purpose, the absorbance of the two dispersions was measured
for a few different wavelengths using a Perkin-Elmer Lambda 35 UV-vis-NIR
Spectrometer. By adding some amounts of IPA into the denser dispersion the
relative concentration of the two dispersions was equalized; this would make
the gradual mixing task much easier since the goal was to create a vector of
different combinations of these two materials by gradually changing the relative
proportions, being 100% W S2 (for f1), gradually adding M oS2 and reducing
W S2 in steps of 10% to create new combinations (for f2, ..., f10) and finally
reaching to 100% M oS2 (for f11). Altogether 11 of such combinations were made,
stored in separate sealed containers and later were drop-casted using micro-liter
onto surface of separate clean glass slides (Fig. 1(a)). The number of drops for
each glass slide were kept the same to create almost the same thickness and area
of drop-casted materials on glass. The IPA dried out in a few seconds. The slides
with nanomaterials on them (called "filters") were annealed in nominal vacuum
for 12 hours to stabilize them and eliminate any trace of IPA.
Transmittance. Since the glass itself was not part of the nanomaterial fil-
ters, by placing a clean glass slide as reference in reference beam position of
UV-vis-NIR the effect of glass itself was removed and the outcome was trans-
mittance of the 2D nanomaterials only (Fig. 1(b)). The transmittance spectrum
of each filter was measured about 120 times over the 325 -- 1100nm range with
1nm precision of UV-vis. Fig. 1(b) shows the transmittance spectrum of 11 filters
where each curve is averaged from 120 measurements of the same filter on the
scale of 0-1, with 1 being 100% transparent. As it can be seen all filters have
finite non-zero transmittance over a fairly large wavelength range.
Atomic Force Microscopy (AFM), Scanning Electron Microscopy
(SEM), and Ramman Spectrum. In order to obtain the nanomaterial prop-
erties of the exfoliated TMDs, a single drop of M oS2 and W S2 was drop-casted
on two separated silicon/silicon dioxide slides. The reason behind this was first,
to do layer-thickness investigation via AFM since the glass slide does not posses
as smooth surface as Silicon dioxide wafer does; second, Characteristic Ram-
man Peak of Silicon dioxide is a standard measure to study the nanomaterial
Bayesian Inference-enabled Precise Optical Wavelength Estimation
15
k, σi
k, so P (tk λi) ∼ N (µi
properties. AFM and SEM investigations revealed that a typical nanoflake of
M oS2 was about 500nm long and 30nm thick. The SEM and AFM images with
the corresponding line profiles of AFM'ed areas and Ramman spectrum of the
exfoliated samples can be found in supporting documents section.
Statistical Model; Generating Large Synthetic Training Samples
for the Inference Model. To calculate the individual filter likelihood P (tk
λj), it was assumed that transmittance data tk of filter k at wavelength λi
comes from a Gaussian normal distribution with the mean value of µi
k and
standard deviation of σi
k). This likelihood was used as
a generative model to synthesize large amount of training samples, which were
used in our training error reported in Fig. 3(a) and Fig. 4. For each wavelength
using the mentioned likelihood, 100 vector of 11 elements (1 transmittance per
each of 11 filters) were generated. The corresponding wavelength of each of these
100 synthesized examples per wavelength was estimated, and the 100 estimated
wavelengths were averaged and used to find the average error. The same was
performed for all wavelengths in the mentioned spectrum. This is where the test
samples were collected independent from the training data using UV-vis-NIR;
in another word, the test samples were not part of training set seen by the code.
For each test light, 100 test vectors of 11 elements (1 per filter) were sampled
and collected. Before doing estimation, every 10 samples were averaged, which
means only 10 vectors per test sample were obtained. The wavelength of these
10 (averaged) samples was estimated using Bayesian inference, and the average
test errors are presented on the same plots as the training errors are.
k-Nearest Neighbor. By averaging the 120 spectrum per filter, a single
spectrum per filter was obtained which was a 775 × 11 matrix of transmittance
values (11 filter and 775 spectrum elements between 325 -- 1100nm). With this T
matrix at hand, the sum of squares of absolute errors between the test vector
(1×11 elements) and each row (wavelength) of T was calculated. The wavelength
with smallest square value was picked as the best estimation.
5 Acknowledgment
DH and SK acknowledges financial support from NSF ECCS 1351424, and a
Northeastern University Provost's Tier 1 Interdisciplinary seed grant.
6 Supporting Information Available
Supporting Information Available: 1. Fabrication of filters; 2. Atomic force mi-
croscopy (AFM) images and line-profile of the samples. 3. Scanning electron
microscopy (SEM) images of the samples. Posterior probability estimation ex-
amples for test samples at some wavelengths. 4. Normalized root-mean-square
error percentage (RMS%) equation. This material is available free of charge via
the Internet at http://pubs.acs.org.
16
D. Hejazi, S. Liu, S. Ostadabbas, and S. Kar
References
1. Altman, N.S.: An introduction to kernel and nearest-neighbor nonparametric re-
gression. The American Statistician 46(3), 175 -- 185 (1992)
2. An, X., Liu, F., Jung, Y.J., Kar, S.: Tunable graphene -- silicon heterojunctions for
ultrasensitive photodetection. Nano letters 13(3), 909 -- 916 (2013)
3. Bassett, R., Deride, J.: Maximum a posteriori estimators as a limit of bayes esti-
mators. Mathematical Programming pp. 1 -- 16 (2016)
4. Berg, M., Keyshar, K., Bilgin, I., Liu, F., Yamaguchi, H., Vajtai, R., Chan, C.,
Gupta, G., Kar, S., Ajayan, P., et al.: Layer dependence of the electronic band
alignment of few-layer mo s 2 on si o 2 measured using photoemission electron
microscopy. Physical Review B 95(23), 235406 (2017)
5. Bernardo, J.M., Smith, A.F.: Bayesian theory (2001)
6. Boyer, K.L., Kak, A.C.: Color-encoded structured light for rapid active ranging.
IEEE Transactions on Pattern Analysis & Machine Intelligence (1), 14 -- 28 (1987)
7. Copeland, K.G., Toerne, K.A.: Color infrared light sensor, camera, and method
for capturing images (Jun 25 2009), uS Patent App. 11/960,302
8. Cormen, T.H., Leiserson, C.E., Rivest, R.L., Stein, C.: Greedy algorithms. Intro-
duction to algorithms 1 (2001)
9. De Sanctis, A., Mehew, J., Craciun, M., Russo, S.: Graphene-based light sensing:
fabrication, characterisation, physical properties and performance. Materials 11(9),
1762 (2018)
10. Decker, D.L., Tolles, W.M.: Optical notch filter utilizing electric dipole resonance
absorption (Jul 11 1978), uS Patent 4,099,854
11. Goossen, K.W.: Voltage-tunable photodetector (Jul 12 1994), uS Patent 5,329,136
12. Goudjil, K., Sandoval, R.: Photochromic ultraviolet light sensor and applications.
Sensor Review 18(3), 176 -- 177 (1998)
13. Grigorenko, A., Polini, M., Novoselov, K.: Graphene plasmonics. Nature photonics
6(11), 749 (2012)
14. Hu, X., Liu, H., Wang, X., Zhang, X., Shan, Z., Zheng, W., Li, H., Wang, X.,
Zhu, X., Jiang, Y., et al.: Wavelength selective photodetectors integrated on a
single composition-graded semiconductor nanowire. Advanced Optical Materials
p. 1800293 (2018)
15. Johnston, M.B.: Optoelectronics: colour-selective photodiodes. Nature Photonics
9(10), 634 (2015)
16. Kim, J.T., Yu, Y.J., Choi, H., Choi, C.G.: Graphene-based plasmonic photodetec-
tor for photonic integrated circuits. Optics express 22(1), 803 -- 808 (2014)
17. Kipping, D.M., Spiegel, D.S.: Detection of visible light from the darkest world.
Monthly Notices of the Royal Astronomical Society: Letters 417(1), L88 -- L92
(2011)
18. Knox, R.S.: Theory of excitons. Solid State Phys. 5 (1963)
19. Konstantatos, G., Howard, I., Fischer, A., Hoogland, S., Clifford, J., Klem, E.,
Levina, L., Sargent, E.H.: Ultrasensitive solution-cast quantum dot photodetectors.
Nature 442(7099), 180 (2006)
20. Lai, K., Campbell, J.C.: Design of a tunable gaas/algaas multiple-quantum-well
resonant-cavity photodetector. IEEE journal of quantum electronics 30(1), 108 --
114 (1994)
21. Lee, P.M.: Bayesian statistics. Oxford University Press London: (1989)
22. Lippke, J.A., Gordon, L.K., Brash, D.E., Haseltine, W.A.: Distribution of uv light-
induced damage in a defined sequence of human dna: detection of alkaline-sensitive
Bayesian Inference-enabled Precise Optical Wavelength Estimation
17
lesions at pyrimidine nucleoside-cytidine sequences. Proceedings of the National
Academy of Sciences 78(6), 3388 -- 3392 (1981)
23. Mori, K., Nishida, T., Hashimoto, H.: Ultraviolet light measuring chip and ultra-
violet light sensor using the same (Apr 22 2003), uS Patent 6,551,493
24. Nun, I., Pichara, K., Protopapas, P., Kim, D.W.: Supervised detection of anoma-
lous light curves in massive astronomical catalogs. The Astrophysical Journal
793(1), 23 (2014)
25. Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, i.V., Kis, A.: Single-layer
mos2 transistors. Nature nanotechnology 6(3), 147 (2011)
26. Rieke, G.: Detection of Light: from the Ultraviolet to the Submillimeter. Cambridge
University Press (2003)
27. Sahami, M.: Learning limited dependence bayesian classifiers. In: KDD. vol. 96,
pp. 335 -- 338 (1996)
28. Shenoi, B.A.: Introduction to digital signal processing and filter design. John Wiley
& Sons (2005)
29. Splendiani, A., Sun, L., Zhang, Y., Li, T., Kim, J., Chim, C.Y., Galli, G., Wang, F.:
Emerging photoluminescence in monolayer mos2. Nano letters 10(4), 1271 -- 1275
(2010)
30. Tang, L., Kocabas, S.E., Latif, S., Okyay, A.K., Ly-Gagnon, D.S., Saraswat, K.C.,
Miller, D.A.: Nanometre-scale germanium photodetector enhanced by a near-
infrared dipole antenna. Nature Photonics 2(4), 226 (2008)
31. Wang, X., Xu, L., Sun, H., Xin, J., Zheng, N.: Bionic vision inspired on-road
obstacle detection and tracking using radar and visual information. In: Intelligent
Transportation Systems (ITSC), 2014 IEEE 17th International Conference on. pp.
39 -- 44. IEEE (2014)
32. Wilke, R.G., Moghaddam, G.K., Dokos, S., Suaning, G., Lovell, N.H.: Stimulation
of the retinal network in bionic vision devices: From multi-electrode arrays to
pixelated vision. In: International Conference on Neural Information Processing.
pp. 140 -- 147. Springer (2010)
33. Wilson, J.A., Yoffe, A.: The transition metal dichalcogenides discussion and inter-
pretation of the observed optical, electrical and structural properties. Advances in
Physics 18(73), 193 -- 335 (1969)
34. Wood, T., Burrus, C., Gnauck, A., Wiesenfeld, J., Miller, D., Chemla, D., Damen,
T.: Wavelength-selective voltage-tunable photodetector made from multiple quan-
tum wells. Applied Physics Letters 47(3), 190 -- 192 (1985)
35. Xia, F., Mueller, T., Lin, Y.m., Valdes-Garcia, A., Avouris, P.: Ultrafast graphene
photodetector. Nature nanotechnology 4(12), 839 (2009)
|
1911.12393 | 1 | 1911 | 2019-11-27T19:44:24 | Field emission microscopy of carbon nanotube fibers: evaluating and interpreting spatial emission | [
"physics.app-ph",
"cond-mat.mtrl-sci"
] | In this work, we quantify field emission properties of cathodes made from carbon nanotube (CNT) fibers. The cathodes were arranged in different configurations to determine the effect of cathode geometry on the emission properties. Various geometries were investigated including: 1) flat cut fiber tip, 2) folded fiber, 3) looped fiber and 4) and fibers wound around a cylinder. We employ a custom field emission microscope to quantify I-V characteristics in combination with laterally-resolved field-dependent electron emission area. Additionally we look at the very early emission stages, first when a CNT fiber is turned on for the first time which is then followed by multiple ramp-up/down. Upon the first turn on, all fibers demonstrated limited and discrete emission area. During ramping runs, all CNT fibers underwent multiple (minor and/or major) breakdowns which improved emission properties in that turn-on field decreased, field enhancement factor and emission area both increased. It is proposed that breakdowns are responsible for removing initially undesirable emission sites caused by stray fibers higher than average. This initial breakdown process gives way to a larger emission area that is created when the CNT fiber sub components unfold and align with the electric field. Our results form the basis for careful evaluation of CNT fiber cathodes for dc or low frequency pulsed power systems in which large uniform area emission is required, or for narrow beam high frequency applications in which high brightness is a must. | physics.app-ph | physics |
Field emission microscopy of carbon nanotube fibers: evaluating and interpreting
spatial emission
Taha Y. Posos,1, ∗ Steven B. Fairchild,2 Jeongho Park,2 and Sergey V. Baryshev1, †
1Department of Electrical and Computer Engineering,
Michigan State University, East Lansing, MI 48824, USA
2Materials and Manufacturing Directorate, Air Force Research Laboratory, WPAFB, OH 45433, USA
In this work, we quantify field emission properties of cathodes made from carbon nanotube (CNT)
fibers. The cathodes were arranged in different configurations to determine the effect of cathode
geometry on the emission properties. Various geometries were investigated including: 1) flat cut
fiber tip, 2) folded fiber, 3) looped fiber and 4) and fibers wound around a cylinder. We employ
a custom field emission microscope to quantify I-V characteristics in combination with laterally-
resolved field-dependent electron emission area. Additionally we look at the very early emission
stages, first when a CNT fiber is turned on for the first time which is then followed by multiple
ramp-up/down. Upon the first turn on, all fibers demonstrated limited and discrete emission area.
During ramping runs, all CNT fibers underwent multiple (minor and/or major) breakdowns which
improved emission properties in that turn-on field decreased, field enhancement factor and emission
area both increased. It is proposed that breakdowns are responsible for removing initially undesirable
emission sites caused by stray fibers higher than average. This initial breakdown process gives way
to a larger emission area that is created when the CNT fiber sub components unfold and align with
the electric field. Our results form the basis for careful evaluation of CNT fiber cathodes for dc or
low frequency pulsed power systems in which large uniform area emission is required, or for narrow
beam high frequency applications in which high brightness is a must.
I.
INTRODUCTION
New and novel cathodes are being investigated for
use as electron beam sources for next generation vac-
uum electronic devices (VEDs). Applications such as
electron microscopy, X-ray sources, and traveling wave
tube amplifiers require high current, high brightness elec-
tron beams with a narrow energy distribution. Cathodes
need to be robust and durable to protect against dam-
age from ion back-bombardment and heating (external
or self-induced) during operation. Cathode lifetimes of a
few 1000s of hours are required[1].
As VEDs progress towards higher frequency and higher
power operation the benefits of using field emission cath-
odes rather than thermionic cathodes becomes appar-
ent. This primarily stems from the fact that cathode
size scales as as 1/f , where f is fundamental operat-
ing frequency of the device. Higher frequency devices[2 --
4] therefore require smaller cathodes, and the excessive
heat generated by thermionic can result in severe ther-
mal stress placed on cathode assemblies which leads to
beam instability. Field emission cathodes also offer the
potential of fast ON/OFF switching, as compared to ex-
ternally heated thermionic sources which require a tem-
perature thermal ramp-up to reach maximum emission
current. This fast ON/OFF switching capability offers
the potential of more efficient gating techniques.
Fibers made from carbon nanotubes (CNT) have
demonstrated significant potential for use as field emis-
∗ [email protected]
† [email protected]
sion cathodes[1, 5]. CNT fibers have excellent electri-
cal and thermal conductivity, and produce high output
emission currents with good current stability for ultralow
turn-on voltage. To date, most data on the emission
properties of CNT fibers has been obtained by measur-
ing emission current in a simple diode configuration with
the voltage applied to a metallic anode positioned above
a vertically mounted fiber. However, there are numerous
examples demonstrating that field emission is often not
laterally uniform[6, 7]. Thus, there is a need to evalu-
ate emission area to realistically estimate current density
and cathode brightness.
To spatially resolve the emission properties of CNT
fibers we utilize a projection type field emission mi-
croscopy apparatus that can both measure and image
the emission current.. Four different CNT fiber cath-
ode designs were fabricated for this measurement. We
observe that CNT fibers undergo a conditioning process
that immensely improve formal emission characteristics
(turn-on field and field enhancement β-factor) but not
necessarily spatial uniformity/coherence of emission. We
find that the field emission area is responsible for uncon-
ventional emitter behavior, namely, emitter saturation
and self-heating. Results and conclusions are consistent
across all the tested geometries. The most promising
CNT fiber cathode design is emphasized.
II. SAMPLES AND EXPERIMENTAL
The CNT fibers used in these experiments were pur-
chased from DexMat, Inc.
in Houston, TX. The fibers
were fabricated using a wet spinning technique described
by Behabtu et al.[8]. This fabrication process ensures
2
FIG. 1. SEM images of sample A (flat cut sample); sample B (folded sample); sample C (the wound geometry sample); and
optical microscope image of sample D (looped sample). Bottom row: side camera views of the samples placed against the
imaging YAG screen. All these images are taken before starting the experiments. There were no visible signs of unfolded stray
fibrils on the samples.
that the CNTs comprising the fibers are closely packed
and highly aligned which ensures high electrical and
thermal conductivity[9] as well as optimal performance
when used as eiher wire conductors or field emission
cathodes[8, 10 -- 12]. Carbon nanotube yarns are made by
twisting or braiding together multiple CNT fibers. Both
individual fibers and twisted yarns were used in these
experiments.
The CNTs fibers were arranged in four different config-
urations which utilized either a single ∼90µm diameter
fiber or multiple fibers braided together into a larger di-
ameter yarn. These different configurations allowed us
to investigate the effects of surface geometry on electron
brightness, beam size, emission area and current density.
Sample A consisted of four yarns inserted together into
a 3 mm diameter metal tube. Each yarn consisted of
∼300 fibers braided together to make the total yarn di-
ameter ∼900 µm. The yarns are protruding from the end
of the tube where then mechanically cut in an attempt to
get a surface with uniform emitter heights. This was dif-
ficult to achieve due to the toughness of the CNT yarns
which makes them difficult to cut. The final results are
shown in Fig.1 which shows an SEM image of the cut
fibers as well as an optical image which shows a side
view of the cathode.
Sample B was made of CNT yarns that were ∼200 µm
in diameter. They consisted of 21 CNT fibers braided
together. Several yarns were folded together and then
shoved through the cylinder to make a somewhat rounded
tip that protruded through the end of the cylinder. Fig.1
shows an SEM image of the bunched yarns at the top of
the cylinder as well as an optical image which shows a
side view of the cathode.
Sample C was made by winding a CNT yarn around
the wall of a 3 mm diameter metal cylinder. The purpose
of this sample was to see if we could make a uniform
emission edge around the edge of the cylinder. Fig.1
shows an SEM image of the top of the cylinder showing
the yarns pulled over the edge. Also shown is an optical
image of the side of the cathode. Samples A,B, and C
were all attached to the steel cylinder with silver paint
to ensure an electrical contact to ground.
Sample D was a single looped CNT fiber of ∼90 µm
in diameter which was arched and attached from both
ends to the stainless steel base. The fiber was contacted
to the steel base with silver paint. Optical images of the
cathode are shown in Fig.1.
The experiments were performed using a field electron
microscopy technique given in Ref.[13]. The measure-
ment setup is shown in Fig.2. In place of a standard metal
anode, we used a custom scintillator anode screen. The
screen is yttrium aluminum garnet doped with cerium
(YAG:Ce) coated with Molybdenum (Mo). YAG:Ce has
diameter of 1 inch, and thickness of 100 µm. The Mo
coating was applied in house using magnetron sputter-
ing in UHV base pressure system, and resulting film had
thickness of 7-8 nm. Metal needs to be deposited on
YAG:Ce screen to make it conductive, then electric field
can be establish between anode and cathode to accelerate
electrons through ultra-high vacuum, and capture and
them into the ground. The coating is thin enough to let
electrons penetrate through Mo to YAG:Ge to produce
green light and thick enough to prevent YAG:Ce screen
from charging up[13]. Mo is chosen because it has high
melting point (2896 K), so it can sustain in high power
density electron beam. No visible electron bombardment
induced damage (burn through pinholes) was observed
on the screen upon completing measurements. Cathode
base, which samples are mounted on top of, was surface
polished 316 stainless steel cylinder and 4.4 mm in diame-
3
FIG. 2. Experimental setup cartoon
ter. The cathode base was then attached to a in-vacuum
micrometer used to set the interelectrode gap. Paral-
lelism of the screen and the sample surfaces is checked by
top and side view cameras when installing the cathode
(see Fig.1, bottom row). Samples and vacuum chamber
are grounded. The screens are positioned using another
translation arm that is attached to the system using a
custom quartz nipple and therefore electrically isolated
from the chamber. It is positively biased in the experi-
ment. Emitted electrons from the sample under effect of
bias voltage are accelerated toward the screen and strike
the screen with an energy equal to the applied voltage. In
such way, electrons arriving from different points of the
emitting cathode surface originated at different angles
create cathodoluminescence patterns (at 550 nm lumi-
nescence line) on the YAG:Ce screen. The patterns, cap-
tured by a Canon DLSR camera with CMOS full frame
sensor installed at viewpoint behind the screen, repre-
sent laterally resolved field electron emission. Applied
voltage, feedback current and feedback voltage readings
are enabled by Keithley 2410 electrometer.
In all ex-
periments, the electrometer was programmed to sweep
voltage up/down with 1 V step with 100 µA set as an
upper limit for the emission current. Dwell duration for
each voltage step is 5 s to sample and record current,
set and feedback voltage, vacuum pressure and calculate
statistical error bars. The system was programmed to
take field emission images every 10V step, such that tak-
ing images was synchronized with the electrometer. All
measurements were done in vacuum (2 to 5)×10−8 Torr.
III. RESULTS AND DISCUSSION
A. Conditioning micro-breakdowns
All samples were tested multiple times; each test in-
cluded the voltage sweep up and then down. Fig.3 sum-
marizes electric I − E characteristics comparing the first
and the last run; E-field is the actual field that is cal-
culated using the measured feedback voltage Vf and the
measured gap. One particular feature can be seen it
is the improved efficiency of the cathodes in that the
turn-on field decreased and field enhancement β-factor
increased. The main vehicle mechanism of the improve-
ment is the conditioning process that happens through
a series of igniting/quenching emitters that, in most ex-
treme cases, is accompanied by breakdowns of different
strength. The ignition/quench process appears as exten-
sive noise of the I − E curves of the initial run for all
samples, labeled as Ab, Bb, Cb, Db where b stands for
before. Aa, Ba, Ca, Da where a stands for after show
I − E curves upon completing 4 runs. The extensive con-
ditioning noise is visible because every point on the I −E
curve is collected for 5 seconds to gain enough statistics in
order to calculate average current, voltage, pressure and
their error bars[13]: such a long dwell time captures ups
and downs in the output current of the fibers turned on
for the first time. The current noise of a relatively large
amplitude (3 to 5 times) is associated with breakdowns
(if any) that have negligible strength, i.e. cannot be de-
tected in our system. Following our previous work[14],
the sizable micro-breakdown/discharge taking place can
be visualized by plotting the difference between the set
voltage Vs and the feedback voltage Vf versus the feed-
back voltage or the actual E-field. Such a plot traces
the voltage loss in the system due to arcing: since the
electrometer is power limited, the arc will cause Vf to
drop with respect to Vs. Note, since the dwell or inte-
gration time per point is 5 s, shorter surges will result
in smaller delta between Vf and Vs even if the break-
down/arc/discharge strength was of the same magnitude.
In that sense, we are looking for non-zero difference be-
tween Vf and Vs to mark off the breakdown rather than
evaluate its actual strength. In Fig.3, the I − E curves
are superimposed with Vs-Vf traces. As can be seen,
all four samples underwent through breakdowns of dif-
ferent strengths or lengths (or both). Upon the first turn
on, the samples A and B do not have ramp down curves
as the strength/lasting of the breakdowns was extensive
and the power supply was automatically shut down via
a safety interlock. Even though the breakdown is often
seen a damaging process, in the present case there were
significant emission property improvements. For exam-
ple, the sample A before (Ab) and after (Aa) experienced
2-fold decrease of the turn-on field, from about 0.5 to 0.25
V/µm, and 2-fold increase of the β-factor, from about
3,000 to 9,000.
B. Field emission microscopy and conditioning
To better understand the effects of conditioning and
fully characterize the fiber design, the presented I − E
curves are compared to the laterally resolved field emis-
sion micrographs that are compiled in Fig.4. They com-
pare the emission patters between the first and the last
tests. A few main features can be noted as follows:
1) Samples A and B improved their emission by means
of increasing the total number of strong emitters seen as
bright spots on the micrographs Ab/Aa and Bb/Ba. The
larger the number of strong emitters (higher β-factor)
the lower the turn-on field: the electrometer senses cur-
rents above the detection threshold and therefore larger
number of high β-factor emitters will deliver an output
current of a magnitude above the threshold at a lower
E-field.
2) Sample C behaved differently. As seen from Fig.4,
the run Cb demonstrated very slow response to the field
in that the output current remained ∼1-10 nA even
though the applied field significantly changed (corre-
sponding to the applied voltage of 100 V, out of en-
tire sweep ranging 0 to 375 V). Then a series of micro-
breakdowns took place (at least two were detected) and
the output current instantaneously inflated by over 3 or-
ders of magnitude. Concurrently with the breakdown at
0.3 V/µm one strong emitter (see Cb in Fig.4) appeared
as a red spot. Our imaging screens are semi-transparent
to the red, and in this case red light emission from the
emitting locations was bright to the extent that the green
light emission from the YAG anode screen was not seen.
The intense red light emission suggested that this spe-
cific emitter was delivering major portion of the detected
output current ∼100 µA. The small emitter size (single
nanotube or a cluster of single CNTs) resulted in ex-
4
tensive current density and therefore led to exceptional
thermal heating of this emitter. Unlike samples A and B,
sample C retained a very similar emission area (i.e. one
red emitter in the right bottom corner) in the following
runs. One can see, that the ramp down I − E curve of
the initial run (Cb) and I − E curves of the subsequent
run (Ca) are identical.
There was no quantification metric for emission area
of the sample D and its emission imaging results will be
discussed in more detail in subsection III E that follows.
Overall, the emission improved after fibers underwent
conditioning breakdowns: this is seen as improved effi-
ciency (lower turn-on field and enhanced β-factor) which
happened alongside with the improved spatial emissivity
of the fibers in that the emission area was increased. A
rough stepwise process can be described as 1) the break-
down increases the number of emitters (i.e. through me-
chanical unfolding); 2) the larger number of emitters de-
liver larger output current seen as lowered turn-on field
and larger β-factor of the FN like part of the I−E curves.
This model is further supported by the analysis of the sat-
uration sections of the I − E curves; saturation follows
the FN-like part when going to higher current range. The
FN-like and saturation regions are labeled with FN and
sat. respectively on the FN plots given in Fig.3.
C. Emission area: FN vs image processing
According to the Fowler-Nordheim (FN) law, the emis-
sion current as a function of applied electric field is given
by:
(cid:18) δS
(cid:19)
φ
(cid:18)−6.83 × 109 · φ3/2
(cid:19)
β · E
I = 1.54×10−6
(β·E)2·exp
(1)
where δS is effective emission area, β is unitless effec-
tive field enhancement factor, and φ is the work function
which is assumed as 4.8 eV for all the CNT fiber geome-
tries. When ln(I/E2) is plotted against 1/E, the slope
gives [−6.83 × 109 · φ3/2/β].
Although for metallic surfaces, the experimental data
show linear slope[15],
for the non-metallic and semi-
metallic surfaces there is deviation of the slope from lin-
ear trend[16]. For all geometries of the fiber samples
tested, after filtering out conditioning noise portion of
the FN plots, there were two distinct slope regions: one
for low applied field and the other for high applied field
(see FN plots in Fig.3). The curves have knee separat-
ing one slope region from another. The low applied field
linear region of a higher slope corresponds to FN-like
emission. The high applied field linear region of a lower
slope corresponds to saturation region. The β-factors
were calculated from the slope of FN-like portion of the
I − E curve using the following procedure: 1) noise data
filtration, shown by blue solid line in Fig.5a; 2) third or-
der polynomial fit, shown by red solid line in Fig.5a; 3)
5
FIG. 3. Semi-log I − E curves and FN plots for the studied fibers.
6
FIG. 4. The laterally resolved field emission pattern on Mo\YAG:Ce screen taken at the same electric field before and after
conditioning for sample A (0.67 V/µm), B (0.72 V/µm), C (0.36 V/µm) and D (0.16 V/µm). The white dashed circles and
line show actual position and orientation of the samples with respect to the YAG screen. The outstanding image at the bottom
illustrates the source of the halo background: it is a stray emitter pair projected to be nearly parallel to the screen plane thus
generating electron rays that have long path across the screen resulting in intense halo.
FIG. 5. a) Emission curve after noise reduction processing (blue), 3rd order polynomial fitting (red), deduced FN-like section
of the I − E curve used for calculation of µ factor (black); b) The vector family used in the knee point calculation.
(2)
knee point calculation[17], using a set of vectors as shown
in Fig.5b; 4) filtering out all the points above the knee
point; 5) fitting fist order polynomial of the remaining
low field curve and calculating its constant slope to ex-
tract effective field enhancement factor, the final slope
is shown by black dashed line in Fig.5a. More specifi-
cally, step 3) when the knee point is calculated, is done
−→
by finding a unique point on the third order polynomial
fit to find the maximum magnitude of vector
d (shown
in Fig.5b) defined as:
−→
d =
−→
b − b · cosθ · a =
−→
b − (
−→
b · a) · a
where
−→
b is a constant vector between two edges of the
curve, −→a is a variable vector from one edge of the curve to
each data point, and θ is angle between −→a and
−→
b . Final
β-factor values extracted for all samples are labeled in
Fig.3.
Effective emission area is then calculated using the
measured I − E data and calculated βs through Eq.1.
The calculated dependence δS(E) for sample B is shown
in Fig.6 as the decaying blue solid line. However, the
results obtained using a custom image processing algo-
rithm developed by our group before[16] show opposite
trend: δS is predicted to increase as the applied field
increases (red solid line in Fig.6). Field emission micro-
graphs taken concurrently with I − E curves and pro-
cessed in batches point out that local emitting maxima
multiply with the field. To keep the discussion to the
general level and compare the trends, we do not present
detailed analysis of the emission area and only calculate
local maxima (brightest emitter locations). Assuming
that the source behind every local maximum is a single
CNT, δS must grow with the field. The same issue was
first pointed out in an original AFRL study of a CNT
fiber[1]. Using a PIC simulation, it was shown that the
emission area has to be a growing function of E-field to
account for the observed emission characteristics. The
presented results are an experimental evidence that sup-
port the earlier PIC findings. This result is also sup-
ported by our earlier studies of nanodiamond emitters
in which δS grows nonlinearly with the electric field[16].
Together, this result adds to concerns raised in recent
literature[18, 19] about the validity of FN equation ap-
plication for extracting the emission area. This problem
is under intense investigations in our lab.
D. Current saturation
One of the quantitative ways[20] to describe saturation
current plateau of a nonmetallic field emitter, or the total
current limit that cannot be exceed, is:
I max
s
=
e · n2/3 · υ∞
l
· δS
(3)
7
FIG. 6. Comparison of trends of the emission area on the
applied electric field extracted from I − E curves using FN
equation versus from the field emission micrograph dataset
using an image processing algorithm developed elsewhere.[16]
the depletion length and δS is the emission area. Since
each sample fibers constituent CNT material is the same
in the before and after experiments, it can be speculated
that their properties are the same. The only parameter
that is changing then in formula 3 is the emission area.
From comparing the I − E curves and field emission mi-
crographs (given for before and after runs at the same
electric field) for samples A and B, it is seen that the
output current saturation plateau value increased by 5-7
times, and so did the emission area as suggested by image
processing illustrated in the previous section. Unlike A
and B, sample C showed no change of the output current
in the saturation regime (ramp down of the Cb and Ca
in Fig.3). From comparing to Fig.4 Cb and Ca, it can
be noted that the single emitter generated during break-
down remained and therefore no change in the output
current was observed; this is consistent with formula 3.
Using typical numbers for CNTs, n ∼ 1018 cm−3 (cal-
culated from σ = e · n · µ where σ=1 kS/cm [21], and
µ=104 cm2/V·s [22]), υ∞ ∼ 107 cm/s [23], l ∼ 890 nm
(calculated using Ref.[24]), it yields the diameter of the
emitter of 0.7 µm. This result is much smaller than the
lateral resolution of our microscope yet detecting that
single emitter as ∼0.1 mm spot on the YAG screen. The
reason for that is a fairly large magnification of the sys-
tem when the fibers are placed far away from the screen
(1 mm in this case). Magnification of a point like electron
source can be estimated as
mag = 2 · d · tan(α)
(4)
where d is the distance between the emitting surface and
the screen (d was 1 mm for samples A, B and C) and the
angle α is calculated as
where e is the electron charge, n is the bulk charge car-
rier concentration, υ∞ is the saturated drift velocity, l is
α =
px
pz
=
·
1
β · γ
(5)
(cid:114) 2 · M T E
m0 · c2
where px and pz are transverse and longitudinal mo-
menta, M T E is the mean transverse energy, m0c2 is the
rest energy (0.511 MeV), β is the ratio between elec-
tron velocity and the speed of light, and γ is the Lorentz
factor. By using β = 0.063 and γ = 1 at 1 kV and
M T E = 4.5 meV (corresponding to its Fermi energy),
we find that a point like emitter would appear as a 0.25
mm spot on the screen due to spreading electron rays
that have non-zero transverse momentum.
8
In Ref.[1], CNT fiber was found to saturate at about
250 µA. This result could not be explained by the vac-
uum space-charge (Child-Langmuir) effect. Extending
the application of Eq.(3), we find that the emission area
in saturation had to be ∼0.02% of the total cross section
area of the fiber. PIC results suggested ∼0.3%. The or-
der of magnitude discrepancy could be explained within
the series ballast resistor model[20, 25], by adding ex-
tra terms (in addition to the basic resistivities associated
with the transport through the depletion region and tun-
neling barrier transparency) in the following form:
Ri ∝(cid:88)
i
m∗
e2 · n
· 1
τi
(6)
where τi is a characteristic scattering time that should be
associated with microscopic bundle/alignment structure.
An effect of this sort, earlier observed in Ref.[12], can
change the onset of saturation by many folds and adds
an uncertainty to emission area calculation and was not
included before into basic/simplified models.
E. Emission uniformity and directionality
Additional analysis of Fig.4 shows that four fiber de-
signs demonstrated very different emission patterns that
can be further discussed and interpreted as follows:
1) There is the glowing background that exists on ev-
ery image set in Fig.4. They come from tangent electron
rays that penetrate the anode screen at shallow angles.
Sometimes they can be visualized by moving the screen
such that the sample is at the edge of the screen, or
by increasing the distance between the screen and the
cathode. Then the background halo source can be seen
at the opposite edge of the screen as a streaked mag-
nified nanotube oriented more in parallel, rather than
perpendicular to the screen plane. This is exemplified
in the outstanding image in Fig.4 for sample C that had
the strongest halo. Typically, the halo becomes stronger
after the conditioning process, additionally confirming
mechanical untangling of CNTs comprising the fibers.
These CNT bunches are seen (highlighted by the dashed
circles in Fig.7) by the top view camera measuring the
interelectrode gap. One of them for sample C, bottom
CNT bunch perpendicular to the screen, can be identi-
fied as the major emitter on the laterally resolved im-
ages in Fig.4 when image is taken by the top camera
in dark, this location is glowing bright red corresponding
FIG. 7. Demonstration of various types of unfolded and dif-
ferently aligned stray CNT fibrils that form after conditioning
breakdowns all taken by the top view camera.
to a black body temperature of 1500-2000 K. Looking at
Fig.4, there is correlation:
if the fiber is enclosed into
a hollow cylinder like samples A and B, the background
is suppressed suggesting mechanical support somewhat
mitigates the untangling.
2) The physical dimensions are not directly related to
emission properties, i.e. emission area is not necessarily
large for a large size sample, such as exampled by sample
C. All samples demonstrated a counted number of strong
emitters during the before runs with emission area being
orders of magnitude lower that the physical area available
for emission.
3) Sample B, even improving the emission area upon
conditioning, shows very large distribution of emission
angles. When placed 1 mm away from the screen, the
emission envelope is three times larger than the actual
emitter size (dashed circle in Fig.4 Ba). This suggests
proper performance for MHz application unless the fiber
is placed in a solenoid field for focusing.
If X-band or
beyond applications are sought, the brightness of such a
design will deteriorate the performance of an VED.
4) Upon conditioning, sample D shows no stray emit-
ters (Fig.4) and the emission pattern is an arch showing
a coherent emission from a section of the looped fiber. At
this point, there is no good procedure of evaluating ex-
act field emission area for this type of geometry without
knowing the emitting section. One main complication
is the parallel shift with respect to the actual fiber loop
position due to a fairly large emission angle. Compared
to Fig.4 Db, showing spatially incoherent emission cen-
tered near the actual emitter location, the emission pat-
tern located away from the emitter (Fig.4 Da) is caused
by a mechanical bent when the front half section of the
loop bent down. Even establishing coherent emission,
9
bright point-like emitters appeared outside the emitter
physical size boundary due to untangled stray emitters,
but the core retained its shape and became brighter and
more uniformly distributed carrying more emission cur-
rent. Altogether, sample A demonstrated the best spatial
emission coherence. Therefore, this design may be fur-
ther optimized to achieve high brightness to be used as a
driving injector for miniature/small size VEDs operated
between X- and W- bands.
FIG. 8. Close-ups of emission patterns of sample A before
and after the conditioning runs. The thicker dashed line cir-
cles depict the actual fiber location with respect to the YAG
screen. Ultra-thin orange circles of the same diameter are to
illustrate that major emission pattern fits within the size of
the fiber even though there is a parallel shift caused by slight
misalignment.
this free-standing design shows weak resilience to condi-
tioning breakdowns that always take place.
5) The most remarkable and promising performance
was demonstrated by sample A. During the condition-
ing run, the emission is limited to ∼10 bright strongest
point-like emitters combined with more distributed lobes.
Most importantly all the emission locations are confined
within a circle of the size of the emitter A, see Fig.8. The
parallel shift is due to cut and installation angle imper-
fections. Electrons start at a small angle but soon after,
travel along uniform field lines yielding the projection
shift with respect to the actual emitter location. In Fig.8,
the white circle corresponds to the actual fiber position
and the vanishingly thin orange circle of the same diam-
eter is to emphasize that the diameter of the emission
core matches the fiber diameter. After the breakdown,
the parallel shift has changed due to the change in the
relative position between the fiber and the screen. More
IV. CONCLUSION AND OUTLOOK
Field emission microscopy of four different CNT fiber
designs is presented. Details of cathode conditioning
upon the initial turn on are outlined. It is emphasized
that the electrical breakdown plays critical role in es-
tablishing emission performance and operating point of
the emitter, typically improving performance in terms
of integral I − E characteristic in that the turn-on field
drops, field enhancement and emission area increases,
the saturation level
increases allowing for larger out-
put current. The flat cut fiber geometry enclosed in a
supporting tubing enclosure was found as a best design.
Folded and wound designs either demonstrated lower spa-
tial coherence or greatly suppressed area of emission due
to unfolded stray CNT emitters after undergoing con-
ditioning breakdowns, either would deteriorate perfor-
mance when driving a high frequency VED. The free
standing looped design showed weak mechanical stability
against breakdown: while still promising additional de-
sign considerations must be made to strengthen its stabil-
ity. Altogether, the new results support earlier findings
and provide new insights into performance of the CNT
fibers as the material-of-choice for future VED architec-
tures/platforms.
[1] D. Shiffler, S. Fairchild, W. Tang, B. Maruyama,
K. Golby, M. LaCour, M. Pasquali, and N. Lockwood,
IEEE Transactions on Plasma Science 40, 1871 (2012).
[2] A. J. Theiss, C. J. Meadows, R. Freeman, R. B. True,
J. M. Martin, and K. L. Montgomery, IEEE Transac-
tions on Plasma Science 38, 1239 (2010).
[3] S. S. Dhillon et al., Journal of Physics D: Applied Physics
50, 043001 (2017).
[4] S. Lewis et al., in Proc. IPAC, TUPTS077 (2019) pp.
2098 -- 2100.
[5] S. B. Fairchild, P. Zhang, J. Park, T. C. Back, D. Marin-
cel, Z. Huang, and M. Pasquali, IEEE Transactions on
Plasma Science 47, 2032 (2019).
[6] D. Shiffler, M. Ruebush, M. Haworth, R. Umstattd,
M. LaCour, K. Golby, D. Zagar, and T. Knowles, Review
of Scientific Instruments 73, 4358 (2002).
M. Sena, Applied Physics Letters 79, 2871 (2001).
[8] N. Behabtu et al., Science 339, 182 (2013).
[9] D. E. Tsentalovich, R. J. Headrick, F. Mirri, J. Hao,
N. Behabtu, C. C. Young, and M. Pasquali, ACS Applied
Materials & Interfaces 9, 36189 (2017).
[10] X. Wang, N. Behabtu, C. C. Young, D. E. Tsentalovich,
M. Pasquali, and J. Kono, Advanced Functional Mate-
rials 24, 3241 (2014).
[11] P. Zhang, J. Park, S. B. Fairchild, N. P. Lockwood, Y. Y.
Lau, J. Ferguson, and T. Back, Applied Sciences 8, 1175
(2018).
[12] S. B. Fairchild et al., Nanotechnology 26, 105706 (2015).
[13] S. S. Baturin and S. V. Baryshev, Review of Scientific
Instruments 88, 033701 (2017).
[14] S. S. Baturin, T. Nikhar, and S. V. Baryshev, Journal
of Physics D: Applied Physics 52, 325301 (2019).
[7] D. Shiffler, M. Ruebush, M. LaCour, K. Golby, R. Um-
and
stattd, M. C. Clark, J. Luginsland, D. Zagar,
[15] I. Brodie and P. Schwoebel, eds., in Field Emission in
Vacuum Microelectronics (Springer US, Boston, MA,
2005) pp. 1 -- 17.
[16] O. Chubenko, S. S. Baturin, K. K. Kovi, A. V. Sumant,
and S. V. Baryshev, ACS Applied Materials & Interfaces
9, 33229 (2017).
[17] V. Satopaa, J. Albrecht, D. Irwin, and B. Raghavan, in
2011 31st International Conference on Distributed Com-
puting Systems Workshops (2011) pp. 166 -- 171.
[18] H. Chen, V. Tagliamonti, and G. N. Gibson, Phys. Rev.
Lett. 109, 193002 (2012).
[19] R. G. Forbes, (2019), arXiv:1905.07585.
[20] S. S. Baturin, A. V. Zinovev,
and S. V. Baryshev,
(2017), arXiv:1710.03692.
10
[21] O. Chauvet, L. Forro, W. Bacsa, D. Ugarte, B. Doudin,
and W. A. de Heer, Phys. Rev. B 52, R6963 (1995).
[22] D. Estrada, S. Dutta, A. Liao, and E. Pop, Nanotech-
nology 21, 085702 (2010).
[23] B. Gao, Y. F. Chen, M. S. Fuhrer, D. C. Glattli, and
A. Bachtold, Phys. Rev. Lett. 95, 196802 (2005).
[24] A. F. Yatsenko, physica status solidi (a) 1, 333 (1970).
[25] J. W. Luginsland, A. Valfells, and Y. Y. Lau, Applied
Physics Letters 69, 2770 (1996).
|
1911.06230 | 1 | 1911 | 2019-11-14T16:41:18 | Bandgap Measurement of Reduced Graphene Oxide Monolayers through Scanning Tunnelling Spectroscopy | [
"physics.app-ph",
"cond-mat.mes-hall"
] | Most popular atomically thin carbon material, called graphene, has got no band gap and this particular property of graphene makes it less useful from the aspect of nanoscale transistor devices. The band gap can be introduced in the graphene if it is synthesized through chemical route. First, Graphene Oxide (GO) is made which further go under reduction and turns into Reduced Graphene Oxide (RGO). Band structure investigation of monolayer sheets of reduced graphene oxide (RGO) by Scanning Tunneling Spectroscopy (STS) has been investigated here. The GO sheets are 1-1.2 nm thick and become more thinner after reduction. The band gap of GO was found in the range of 0.8 eV. The RGO showed up a variety of band structure. RGO opens a new field of study of atomically thin layers of carbon because it has got non zero band gap which is not the case for graphene. | physics.app-ph | physics | Bandgap Measurement of Reduced Graphene Oxide
Monolayers through Scanning Tunnelling Spectroscopy
Pankaj Kumara,b,c,*
a. Thin Film Lab, Department of Physics, Indian Institute of Technology Bombay, India
b. Centre of Excellence in nanoelectronics, Indian Institute of Technology Bombay, India
c. L-NESS, Department of Physics, Politecnico di Milano, Via Anzani 42, 22100 Como, Italy
* Corresponding auther. Email: [email protected] Tel:+39-0313327375
Most popular atomically thin carbon material, called graphene, has got no band gap and this particular
property of graphene makes it less useful from the aspect of nanoscale transistor devices. The band gap can
be introduced in the graphene if it is synthesized through chemical route. First, Graphene Oxide (GO) is
made which further go under reduction and turns into Reduced Graphene Oxide (RGO). Band structure
investigation of monolayer sheets of reduced graphene oxide (RGO) by Scanning Tunneling Spectroscopy
(STS) has been investigated here. The GO sheets are 1-1.2 nm thick and become more thinner after
reduction. The band gap of GO was found in the range of 0.8 eV. The RGO showed up a variety of band
structure. RGO opens a new field of study of atomically thin layers of carbon because it has got non zero
band gap which is not the case for graphene.
Keyword: Graphene Oxide(GO), Reduced Graphene Oxide(RGO), Scanning Tunneling Spectroscopy(STS), Modified
Langmuir Blodgett (MLB), Scanning Electron Microscopy(SEM), Atomic Force Microscopy(AFM)
1. Introduction
Graphene has attracted the attention of a lot of researchers all
over the world. Graphene is a single sheet of carbon atoms
detached from graphite, an allotrope of carbon, having
carbon atom sheets stacked over each other. Each single
sheet, graphite is made of, is graphene itself. So, graphene
exists in two-dimensional space. The attraction toward
graphene has a reason. Graphene has shown extraordinary
electronic and electrical properties [1]. The mobility of
electrons is very high in graphene. But the synthesis of pure
graphene is still a challenge. A lot of methods have been used
to synthesize graphene. Some of them give low yield and
some of them give impure graphene. So, the purpose is not
served. Here, in this study we are going to investigate the
chemical route to make graphene. First, an intermediate
material graphene oxide (GO) will be synthesized [2]. Then
it will be reduced to make reduced graphene oxide (RGO).
The whole purpose of studying RGO is that it shows some
interesting properties like graphene [3]. The GO and RGO
have got some organic functional groups containing
s𝑝3carbon and oxygen. The purpose of reducing the GO, is
to reduce the contribution of s𝑝3 carbon and the oxygen,
present in GO [4]. The making of RGO is not as tough as that
of graphene. So, we are interested in RGO. It can be an
alternate to graphene.
In this report the chemically synthesized graphene, also
called graphene oxide (GO) and some different methods to
reduce have been presented. The reduced form of GO is
called reduced graphene oxide (RGO). GO sheet making
techniques and
their
disadvantages are also discussed. The main technique used
their advantages as well as
by us for making GO sheets is Modified Langmuir-Blodgett
(MLB) deposition. In the study, presented in this report, the
four individual methods of making RGO, used by us, are
explained thoroughly. The two different techniques of
reduction of GO through plasma are discussed. The
reduction through standard hydrazine treatment and through
thermal annealing in the presence of graphite powder is also
discussed. The theory of Scanning Tunneling Microscopy
(STM) and Scanning Tunneling Spectroscopy (STS) have
been discussed in detail. The purpose of doing STM study is
that one can get atomic level resolution. This is followed by
STS to get the band structure information of GO and RGO.
In 1983, Binnig and Rohrer succeeded in producing an
atomic resolution image of two-unit cells of the surface of
Silicon. This image received a lot of attention and
appreciation. For this wonderful discovery both were
awarded Nobel Prize in 1986 [5].
The information, we have about the behavior of material on
macroscopic scale is no longer true on the mesoscopic scale
(nanometer scale). Our mathematical modelling needs to be
redeveloped to understand the behavior of materials on
nanometer scale (mesoscopic scale). There is a possibility
that a particle can move from one region to another region
having its potential energy greater than its total energy. This
phenomenon is called tunneling. The study of electronic
structure and surface morphology of the different materials
has always been a challenge for scientists all around the
globe. The technique mentioned in this report, fortunately,
provides very fantastic information about the both. This
technique, I am talking about is Scanning Tunneling
Microscopy (STM). On the same instrument, Scanning
Tunneling Spectroscopy (STS) can also be performed. STM
gives the surface morphology information and STS gives the
information about local density of states (LDOS).
The second section of the report explains the theory of
Scanning Tunneling Microscopy and Scanning Tunneling
Spectroscopy. The methodology of the experiment is
presented in this section.
The third section of the report is about the synthesis of GO
using Hummer's Modified Method. The details of making
GO solution are given here. The pre-deposition substrate
cleaning recipe and transfer of GO onto the substrate have
been explained here. The transfer of GO was done using
Modified Langmuir Blodgett (MLB) which is given in this
section only.
the report contains
The fourth section of the report talks about the different
methods of reduction of GO. The four different methods used
to reduce GO are ammonia plasma treatment, hydrogen
plasma treatment, standard hydrazine treatment and thermal
annealing in presence of graphite powder. All these
techniques, of reduction of GO, are also explained in this
section.The fifth section of
the
experimental data of surface morphological studies of GO
and RGO sheets transferred on Silicon substrate, by
Scanning Electron Microscopy (SEM) and Atomic Force
Microscopy
(AFM). Raman Spectroscopy, X-ray
Photoelectron Spectroscopy (XPS) and the results of STS
measurements performed on GO, thermal annealed GO (in
presence of graphite powder), hydrazine treated GO,
ammonia plasma treated GO and hydrogen plasma treated
GO are also explained in the same section.The summary of
the work carried out and conclusions drown are presented in
the section six.
Scanning Electron Microscopy (SEM) and Atomic Force
Microscopy
(AFM). Raman Spectroscopy, X-ray
Photoelectron Spectroscopy (XPS) and the results of STS
measurements performed on GO, thermal annealed GO (in
presence of graphite powder), hydrazine treated GO,
ammonia plasma treated GO and hydrogen plasma treated
GO are also explained in the same section.The summary of
the work carried out and conclusions drown are presented in
the section six.
2. Scanning Tunneling Spectroscopy
Theory
2.1 Quantum Mechanical description of electron
tunnelling
STM is based on a quantum mechanical phenomenon which
is called tunneling. Small particles like electrons, in quantum
mechanics, show wave-like properties and do penetrate the
potential barriers. STM involves a very sharp conductive tip
which is arranged to come in the vicinity of the sample
surface within tunneling distance (sub-nanometer) [6]. This
makes a metal-insulator-metal setup. In the representation of
one-dimensional tunneling, a patch up of tunneling wave of
the sample electrons (𝜓𝑠) and a wave of a STM tip electrons
(𝜓𝑇) is shown in Figure 1. This overlap of wave-functions
allows the flow of current. In metals, electrons do fill the
continuous energy levels up to the Fermi level (𝐸𝐹).
Above 𝐸𝐹, the activated electrons are found [7]. To observe
the tunneling of electrons through the vacuum gap between
the sample and the tip, a bias voltage (𝑉𝑏𝑖𝑎𝑠) is applied. When
𝑉𝑏𝑖𝑎𝑠 is zero, the electrons cannot flow in either direction.
This reluctance of flow of current is because the tip and the
sample have got their Fermi levels at equal levels. During
positive bias (𝑉𝑏𝑖𝑎𝑠 > 0), the Fermi level of the sample is
raised up and the electrons in the occupied state of the sample
can tunnel into the unoccupied state of the tip. When biasing
is reversed (𝑉𝑏𝑖𝑎𝑠< 0), the electrons in the occupied state of
the tip tunnel into the unoccupied state of the sample.
STM images present the local measurement of the magnitude
of the tunneling current [8]. The tunneling current (I) decays
exponentially with the distance gap distance (d) and is
strongly affected by the density of states (DOS) of the sample
at the Fermi level, 𝜌𝑠(𝐸𝐹).
I α 𝑉𝑏𝑖𝑎𝑠𝜌𝑠(𝐸𝐹)𝑒
I α 𝑉𝑏𝑖𝑎𝑠𝜌𝑠(𝐸𝐹)𝑒− 1.025 𝑑√∅
2.2 STS measurement methodology
ħ
−2 𝑑 √𝑚(∅−𝐸)
The spectroscopy STM mode, involves either a bias voltage
(𝑉𝑏𝑖𝑎𝑠) sweep, or distance (d) ramping. A simplified form of
tunneling current equation can be used to estimate the barrier
height (∅) for the tunneling current,
Log(I) = − C (∅⋅ d) + k
C and k are constants. I-d spectroscopy is useful for the
characterization of the quality of the STM tip, its sharpness
and cleanliness. I-Vbias spectroscopy provides, with a first
order analysis, information about the electronics structure
(LDOS), and a second order analysis information vibrational
mobilities [9] [10]. In Tunneling Spectroscopy,
the
tunneling current I is continuously measured at each location
at a constant bias voltage (𝑽𝒃𝒊𝒂𝒔) [11]. This measurement
information generates a two-dimensional map of tunneling
conductance (I/𝑽𝒃𝒊𝒂𝒔) [12]. A normalized differential
tunneling conductance (dI/d𝑽𝒃𝒊𝒂𝒔)/(I/𝑽𝒃𝒊𝒂𝒔) = d[ln (I)] /
d[ln(𝑽𝒃𝒊𝒂𝒔)] is required by analyzing the obtained I-𝑽𝒃𝒊𝒂𝒔
data [13]. This normalization makes the final tunneling
conductance, independent of bias voltage (𝑽𝒃𝒊𝒂𝒔). Scanning
tunneling spectroscopy
local electronic
information (LDOS) rather than average LDOS. Metals do
not have a gap between the occupied states (valence band)
and the unoccupied states (conduction band). So, for metals,
the variation in LDOS is comparatively low and I-𝑉𝑏𝑖𝑎𝑠
curves do show linear behavior for the most part. This linear
behavior results in a very small dI/d𝑉𝑏𝑖𝑎𝑠 gradient. Semi-
metals also have not got any gap between the occupied and
unoccupied states. But there is a gap in the momentum space
because the waves are out of phase. This depresses the
tunneling conductance around the Fermi level (𝐸𝐹) and
bends the LDOS at low 𝑉𝑏𝑖𝑎𝑠. For semiconductors and
insulators, the tunneling conductance in the vicinity of 𝐸𝐹 is
zero. The band gap, 𝐸𝑔 = 𝑉+𝑏𝑖𝑎𝑠 + 𝑉−𝑏𝑖𝑎𝑠, is comparatively
low for semiconductors (< 3eV) [14] [15]. Semiconductors
do show a highly bend in LDOS while it is flat for insulators
at low 𝑉𝑏𝑖𝑎𝑠 [16]. Dopingof semiconductors does reduce the
resolves
the
band gap and can modify the DOS at 𝐸𝐹 such that it may
show a semi-metal behaviour. For scanning tunneling
spectroscopy (STS) the scanning tunneling microscope is
used to measure the number count of electrons against
electron energy. The electron energy is set by applying
voltage between the sample and the tip [17] [18].
Figure 1. (a) Schematic of STM one dimensional tunnelling configuration, (b)Schematic of a metal-insulator-metal
tunneling junction. The grey area represents electron filled states and the white area is empty states
3. Graphene Oxide Sample Making
Modified Hummer's method was used for the synthesis of
GO. GO sheets were synthesized by chemical exfoliation of
graphite powder using sodium nitrate (NaNO3), sulphuric
acid (98% conc. H2SO4), potassium per magnet (KMnO4),
hydrogen peroxide (H2O2) and milli-Q water. The final
supernatant solution (GO suspension) was collected and
called the stock solution, which was tested to standardize the
GO content. For this, 20 micro liter of master solution was
diluted by adding 3 ml of milli-Q water + methanol solution
(1:5). Its UV-vis absorbance spectrum was recorded was
recorded to observe the peak at ~ 230 nm along with a
shoulder at ~ 290 nm, corresponding to -- * transition of
C=C and n→* transition of the C=O, respectively. If the
absorbance at 230 nm was ~ 0.1, then the stock solution was
used as a spreading solution for MLB deposition. A typical
absorbance spectrum of the spreading solution is shown in
Figure 2. Before going for deposition, the substrate was
cleaned by RCA treatment. Ultimately the substrate will be
rinsed with milli-Q water and will be preserved in milli-Q
water only. The MLB process is shown schematically in
Figure 3. With the addition of spreading solution, the surface
pressure is found to increase. During MLB deposition in this
work, 2 - 4 ml of spreading solution (depending on the
concentration of GO solution) was used to obtain a surface
pressure of 6 -7 mN/m. The meniscus speed in the present
work was chosen as 3 -5 mm/min. During MLB transfer the
surface pressure does not change by more than 5%. For more
details on this section, one can refer to the supplementary
information [19] [20].
Figure 2. The UV-visible spectrum of as prepared
GO solution
Figure 3. Successive stages of transfer of GO monolayer sheets by MLB technique: (a) milli-Q water as
the subphase in a reservoir, (b) substrate inserted in the subphase, (c) GO solution is spread and (d)
transfer of GO sheets by draining of the subphase [20].
4. Reduced Graphene Oxide making
4.1 Chemical & thermal reduction
To carry out the chemical reduction, as -- transferred GO
sheets on the substrates were placed in a petri-dish, in which
a small container with 1ml of hydrazine monohydrate
(N2H4.H2O) 99% was also placed. The petri-dish was
covered and sealed with para-film. The petri-dish was placed
on a hot plate at 45±5oC for 18hrs. After the completion of
the chemical treatment with hydrazine vapor, the sample
were gently rinsed with milli-Q water several times and dried
under a stream of nitrogen gas.
The second step of the reduction process involves heat
treatment of the hydrazine treated GO sheets inside a quartz
tubular furnace. The heat treatment process consists of two
steps. Hydrazine treated GO samples were initially heat
treated at 800C for 30 min in vacuum (~10-6mbar) to remove
the adsorbed vapor/moisture. This was followed by another
heat treatment at 4000C for six hours in a stream of argon
(99.999%) gas. The rate of increase of temperature was fixed
at ~100C/min. after completion of the heat treatment, the
flow of argon gas was maintained, till room temperature was
reached [21] [19].
4.2 Thermal reduction in presence of graphite
powder
An evacuated and sealed system is an important requirement
in order to achieve the required result. As- transferred GO
sheets on Si substrates were reduced by graphite powder
under the condition mentioned here [22]. GO sheets inside
an evacuative & sealed enclosure with the presence of
graphite powder as carbon source at a pressure of 10-5 mbar
in the glass tube are placed properly. The heat treatment was
carried out for almost 6 hours on a stretch and the GO sheets
got heated to 1000 0C. At the temperature of 800 0C - 1000
0C, the GO sheets have been reduced in order to form
reduced graphene oxide sheets. This process plays an
important role in order to get rid of the defects and restoring
the graphitic carbon network in the GO sheet. The RGO
formed by above mentioned process was investigated to see
the change in property like stability, chemical & electronic
structure and electrical transport property under different
heat treatments. After investigating RGO sheets were found
to be highly conducting RGO which were never reported
earlier.
4.3 Ammonia plasma reduction
GO monolayers sheets were transferred on Si substrates by
MLB deposition method. These samples were subjected to
Ammonia plasma treatment under different conditions to
obtain nitrogen doped RGO [23]. The ammonia plasma
treatment system consists of the process chamber, cathode
assembly, DC power supply and vacuum gauge. A 12" water
cooled, stainless steel chamber was used, which was
evacuated by a diffusion pump backed by a rotary pump. A
Cu cathode of 3" diameter was used, and the GO samples
were placed on a grounded substrate holder-cum-heater
placed below, at the distance of ~7cm. The temperature of
heater could be varied up to 3000C. A shutter attached to a
rotary feed through was used to enable shielding of the
substrate. Ammonia gas was introduced into the chamber by
stainless steel tubes, and gas flow rate was controlled by
needle valve. A DC power supply with floating terminals
was used to obtain allow power plasma. The negative
terminal was connected to the cathode and the positive
terminal was grounded. The chamber was evacuated to a
base pressure of ~2×10-6 to 3×10-6 before generating plasma.
Then ammonia gas (99.999%purity) was sent into the
chamber to maintain a working pressure of ~0.5 mbar inside
the plasma chamber. The plasma treatment was carried out
typically at 10 W for 5 min at room temperature.
4.4 Hydrogen plasma reduction
Hydrogen plasma was also carried out in the same vacuum
chamber, as described above. GO monolayers sheets were
transferred on Si substrates by MLB deposition method. This
sample was subjected to hydrogen plasma treatment under
different condition to obtain RGO. Before creating the
plasma, the chamber was evacuated to get a chamber
pressure of 2×10−6 mbar to 3×10−6 mbar. Then nitrogen gas
(99.99 % purity) is introduced into the line for purging. Then,
hydrogen gas was sent into the chamber maintaining the
chamber pressure at 0.5 mbar. The hydrogen plasma
treatment was carried out at 15 W. The sample was exposed
the plasma for 30 sec at room temperature and 500C [24].
Figure 4. The Scanning Electron Microscopy images of (a) GO deposited on Si, (b) Ammonia plasma treated GO
deposited on Si,(c)Hydrazine treated GO deposited on Si and (d) Hydrogen plasma treated GO deposited on Si and (e)
Annealed GO in the presence of graphite powder(graphite mediated), deposited on Si. The Atomic Force Microscopy
images of (f) GO deposited on Si, (g) Ammonia plasma treated GO deposited on Si, (h) Hydrazine treated GO deposited
on Si and (i) Hydrogen plasma treated GO, (j) thermal annealed GO in presence of graphite powder (graphite mediated)
deposited on Si.
5. Measurement Results & Discussion
5.1 AFM & SEM measurement
Before carrying out the STS measurements, the morphology
of all the GO and RGO samples on Si substrate was studied
with SEM and AFM. The SEM images of as transferred GO
and the RGO samples on a large scale were obtained by
chemical reduction, graphite mediated reduction, ammonia
plasma reduction and hydrogen plasma reduction are shown
in Figure 4(a) to 4(e). All the samples showed a uniform
distribution of GO and RGO sheets of size in the range of 10
- 50 µm.
Figure 4(f) to 4(j) shows the surface morphology of GO and
RGO sheets on Si substrate obtained from AFM. It is clearly
displaying adherent and wrinkle free sheets in all the cases.
The thickness profiles are also shown in all the AFM images.
The thickness of the GO sheets on Si were found to be around
1.2 nm. All the RGO sheets showed a small reduction in their
thickness after the reduction treatments.
5.2 Raman & XPS measurement
Raman spectrum and XPS spectrum was taken for GO and
all the reduced GO sheets. Raman spectroscopy studies have
shown a red shift of G peak after both hydrogen and
ammonia plasma treatments, confirming the reduction of GO
sheets. Apart from plasma treatment hydrazine treated and
graphite powder reduced GO also showed shift in G peak.
XPS studies have shown that the hydrogen plasma reduction
results in increase of sp2 bonded carbon and substantial
removal of oxygen functional groups, comparable to that
observed for chemically reduced GO sheets. Ammonia
plasma reduction leads to simultaneous nitrogen doping of
GO sheets, but the quantification of N was difficult due to
overlapping of C-N peaks with those of some of the oxygen
functional groups. For more details on Raman and XPS,
supplementary information is suggested to refer.
5.3 STS Measurement
The tunneling current vs bias voltage data for all the samples
are shown in Figures 5 & 6. For every Itunneling -Vbias curve,
the dI/dV/I/V curve is also plotted against Vbias. As
mentioned above, dI/dV/I/V is independent of V (bias
voltage) while I/V is not, hence, the dI/dV/I/V, normalization
is usually done to get rid of the dependence on the bias
voltage. The dI/dV/I/V curves are shown in each figure for
all the cases. For each type of GO and RGO samples, 10
independent measurements were performed on each different
kind of sample and the typical results are presented here.
Figure 5(a)&5(c) shows typical tunneling current spectra for
two different GO samples. For all the GO samples, the
tunneling current curve showed a nonlinear behavior, which
is typically shown for both these samples. This kind of
nonlinear behavior by tunneling current is generally shown
by semiconductors. The normalized tunneling current
against the bias voltage for both the GO monolayer sheets
also shown in Figure 5(b)&5(d). The zero bias voltage
corresponds to Fermi energy level and the two bumps in the
vicinity of zero bias voltage give the information about
valance band and conduction band edges. The valance band
is present in the negative side of the bias voltage and the
conduction band is present in the positive side of the bias
voltage and the difference between the transition regions on
both sides of zero bias (Fermi level) is usually considered to
be the band gap. In both these cases, the values of band gap
are found to be in the range of 0.7 eV to 0.8 eV. According
to the available literature on optical measurements on GO
multilayers/films, the band gap of GO is expected to be in
the range of 2 - 4 eV [48]. Hence, the values of 0.7 -0.8 eV
obtained from STS measurements, appears to be on the lower
side, possibly due to the measurements being performed
under ambient conditions. This aspect needs
to be
investigated further.
Scanning Tunneling Spectroscopy is a powerful technique to
study the electronic structure at atomic resolution [25]. This
technique has till date not been used for the study of GO or
RGO monolayer sheets. The measurements were performed
on GO and four different kind of RGO samples. All the
samples for STS measurements were made on highly
conducting Silicon (100), having resistivity in the range of
0.001- 0.005 Ohm-cm. The GO sheets were transferred on
the silicon substrates and subjected to four different
reduction treatments (1) ammonia plasma, (2) hydrogen
plasma, (3) chemical reduction route using hydrazine and (4)
graphite mediated reduction. All of them are discussed in
section 3 of the report. RGO thus produced seems to have
conductivity in the range of (1) 33 S/cm, (2) 10 S/cm, (3) 2-
10 S/cm and (4) (2 − 3) × 103 S/cm. Conductivity values
were obtained from FET device measurement of the same
RGO sheets.
The STS measurements were performed on NanoScope𝑅-IV
di-Digital Instruments machine. The samples were placed on
to a sample holder (stub) and a contact was made with silver
paste from sample to sample holder. The assembly was
placed under the STM tip to perform the STS measurements
under the ambient condition. A platinum-iridium tip was
used for these measurements. The biasing voltage to
maintain the constant height or constant current was in the
range of 150 to 200 mV with a current set-point value of 1.0
nA. The sample was biased from -1.0 Volt to +1.0 Volt and
corresponding tunneling current was recorded.
The STS measurements of RGO samples obtained by
chemical/thermal reduction and graphite mediated reduction
are presented before those of the samples obtained by plasma
reduction methods. As mentioned above, chemical/thermal
reduction of GO by hydrazine vapor exposure is an
established method for obtaining RGO sheets of reasonable
conductivity (1 -100 S/cm) and the graphite mediated
reduction is reported to result in highly conducting RGO
sheets (1500 S/cm), hence, the STS measurements on these
samples serve as a reference for RGO sheets.
The plot of normalized differential tunneling current against
bias voltage for RGO sheets obtained by hydrazine vapor
reduction are shown in the Figure 6(c). The tunneling current
curve obtained, did not show any resemblance with the
tunneling current curve of the GO and tends towards a linear
behavior. This shows a nearly flat normalized tunneling
current, without bumps on either side of the Fermi level.
Figure 6(d) shows the normalized differential tunneling
curve for the graphite mediated sample. This curve also show
feature like those observed for the hydrazine vapor reduced
GO sheets, although the tunneling current behavior is much
closer to linear behavior. Such features are usually observed
in the case of the metals or semimetals and the improvement
in linearity of the tunneling current for RGO sheets obtained
by graphite mediated reduction is consistent with its higher
conductivity. Thus, STS data indicates that with increase in
conductivity of RGO sheets, the tunneling current behavior
closely approaches metallic or semi-metallic behavior.
Figure 6(a) presents the normalized differential tunneling
curve for the RGO sample obtained by ammonia plasma
reduction. These curves also show a nearly linear tunneling
current showing a metallic or semi metallic behavior. This is
expected, because the conductivity of the ammonia plasma
treated sample is in the same range as the chemically reduced
GO samples. The normalized differential tunneling curve for
the RGO sample obtained by hydrogen plasma reduction are
shown in Figure 6(b). In contrast to the ammonia plasma
reduced GO sample, this sample shows a completely
different behavior, which resembles, more like the behavior
of as deposited GO sample.
The semiconductor like behavior of hydrogen plasma
reduced GO is intriguing, since, as discussed in earlier in the
same section, the FETs fabricated with the same material
showed
the
chemically reduced GO, as well as an ambipolar behavior.
The only difference was that the FET devices were fabricated
reasonable conductivity, comparable
to
on SiO2/Si, while the STS measurements were carried out on
RGO sheets on highly conducting Si. It is not possible to
exactly point out the reason for this behavior, but it may be
pointed out that the hydrogenation of graphene and GO
sheets is known [49] result in attachment of hydrogen to
carbon atoms, leading to the formation of hydrogenated
graphene or graphene, which has a band gap. Such a
possibility cannot be
to excessive
hydrogenation of these samples, which needs to be explored
further.
ruled out due
Figure 5. The tunneling current against bias voltage for two different GO samples (a) & (c) and the
normalized differential tunneling current against bias voltage (b) & (d) for the GO, deposited on Si.
Figure 6. The normalized differential tunneling current against bias voltage for (a) ammonia plasma treated
GO, transferred onto the Si, (b) for hydrogen plasma treated GO, transferred onto the Si,(c) hydrazine
treated GO, transferred onto the Si & (d) for Graphite reduced GO, transferred onto the Si.
6. Conclusions
GO monolayer sheets of controlled surface density were
transferred by Modified Langmuir Blodgett Technique on
Si substrates. The monolayer sheets were subsequently
reduced by ammonia plasma treatment, hydrogen plasma
treatments, chemical reduction with hydrazine vapor and
graphite powder mediated reduction. The morphological
studies by SEM and AFM measurements showed that the
as-transferred GO sheets are uniformly distributed,
adherent and wrinkle free. The average sheet thickness of
GO sheets was in the range of 1 -- 1.2 nm, which confirms
their monolayer character. The RGO sheets display
morphological stability, subsequent to plasma reduction
treatments, and show no observable change in their
morphology and distribution, except for a small change in
sheet thickness to 0.8 -1 nm, which is attributed to the
removal of oxygen
functional groups. Raman
spectroscopy studies have shown a red shift of G peak
after both hydrogen and ammonia plasma treatments,
confirming the reduction of GO sheets. XPS studies have
shown that the hydrogen plasma reduction results in
increase of sp2 bonded carbon and substantial removal of
oxygen functional groups, comparable to that observed for
chemically
reduced GO sheets. Ammonia plasma
reduction leads to simultaneous nitrogen doping of GO
sheets, but the quantification of N was difficult due to
overlapping of C-N peaks with those of some of the
oxygen functional groups. As transferred GO sheets
displayed conductivity of about 2.9×10−2S/cm, which
reduced to 10 S/cm for hydrogen plasma reduced GO
sheets, 33 S/cm for ammonia plasma reduced GO sheets,
2-10 S/cm for hydrazine vapour reduced GO sheets and
(2 − 3) × 103 S/cm for graphite powder reduced GO
sheets. The observation of p-type conductivity, inspite of
n-type doping with nitrogen is attributed to the presence
of moisture/oxygen on GO sheets under ambient
conditions.
spectroscopy
measurement under ambient conditions were performed
for the first time on GO and RGO monolayer sheets on Si
obtained by hydrogen and ammonia plasma reduction. In
addition, these measurements were also performed on
RGO sheets obtained by chemical reduction as well as
graphite mediated
reduction. The STS
measurements showed that as transferred GO sheets
display a semiconductor like behaviour, although the
measured band gap was found to be ~ 0.8 eV, which is
smaller than the reported band gap of GO multilayers and
films, measured by optical methods. The RGO sheets
obtained by chemical reduction, graphite mediated
reduction and ammonia plasma reduction show a metallic
behaviour. In contrast, hydrogen plasma reduced GO
sheets display a semiconducting behaviour, despite their
high conductivity, which is may be due to excessive
hydrogenation, leading to formation of hydrogenated
graphene or graphene. These studies have opened the
scope of several future investigations on reduced GO
sheets. The STS measurements under ambient conditions
have shown interesting behaviour, particularly in the case
of hydrogen plasma reduced GO sheets, which also need
to be investigated under Ultra High Vacuum (UHV)
conditions.
tunnelling
Scanning
thermal
Conflicts of interest
There are no conflicts to declare.
Acknowledgements
The author would like to acknowledge Prof. S.S Major
from Physics Department at IIT Bombay, India. He also
wants to thank nanofabrication and characterization
facility in CRNTS, CEN and SAIF at IIT Bombay, India.
References
[1] K. S. Novoselov, A. K. Geim, S. V. Morozov, D.
Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva
and A. A. Firsov, Science, 2004, 306, 666-669.
[2] Ji Chen, Bowen Yao, Chun Li and Gaoquan Shi,
Carbon, 2013, 64, 225-229.
[3] Goki Eda, Giovanni Fanchini and Manish
Chhowalla, Nature Nanotechnology, 2008, 3, 270 --
274.
[4] Cristina Gòmez Navarro, R. Thomas Weitz,
Alexander M. Bittner, Matteo Scolari, Alf Mews,
Marko Burghard and Klaus Kern, Nano Letters,
2007, 7, 3499-3503.
[5] G. Binnig and H. Rohrer, Surface Science, 1983,
126, 236-244.
[6] K. Oura, V.G. Lifshits, A. Saranin, A.V. Zotov and
M. Katayama, "Surface Science:An Introduction,"
Berlin: Springer-Verlag, 2003.
[7] R. J. Hammers and D. F. Padowitz, "Methods of
Tunneling Spectroscopy with the STM" from
Scanning Probe Microscopy and Spectroscopy:
Theory, Techniques, and Applications," New York:
Wiley-VCH, 2001.
[8] C. Julian Chen, "Introduction
to Scanning
Tunneling Microscopy," Oxford University Press
New York, 1993.
[9] Z. Klusek, Z. Waqar, E. A. Denisov, T. N.
Kompaniets and I. V Makarenko, Applied Surface
Science, 2000, 161, 508-514.
[10] Y. Zhang, V. W. Brar, F. Wang, C. Girit, Y. Yayon,
M. Panlasigui, A. Zettl and M. F. Crommie, Nature
Physics, 2008, 4, 627-630.
[11] P. Yang, I. Arfaoui, T. Cren, N. Goubet and M.
Pileni, Phys. Rev. B, 2012, 86, 075409.
[12] R. Wiesendanger, "Scanning Probe Microscopy and
and Applications,"
Spectroscopy: Methods
Cambridge University Press, 1994.
[20] V. D. Botcha, G. Singh, P. K. Narayanam, S. S.
Talwar, R. S. Srinivasa and S. S. Major, Materials
Research Express, 2016, 3, 035002.
[13] R. Hamers, Annual Review of Physical Chemistry,
[21] Xingfa Gao, Joonkyung Jang and Shigeru Nagase,
1989, 40, 531-559.
J. Phys. Chem. C, 2010, 114, 832-842.
[14] S. Schintke, S. Messerli, M. Pivetta, F. Patthey, L.
Libioulle, M. Stengel, A. De Vita and W. Schneider,
Phys. Rev. Lett., 2001, 87, 276801.
[22] Gulbagh Singh, V Divakar Botcha, D S Sutar, S S
Talwar, RS Srinivasa and S S Major, Carbon, 2015,
95, 843-851.
[15] N. R. Wilson, P. A. Pandey, R. Beanland, R. J.
Young, I. A. Kinloch, L. Gong, Z. Liu, Ќ. K.
Suenaga, Ќ. J. P. Rourke, S. J. York and J. Sloan,
ACS Nano, 2009, 3, 2547-2556.
[16] R. M. Feenstra and J. A. Stroscio, J. Vac. Sci. &
Technology B, 1987, 5 ,923-929.
[17] R. J. Hamers, "STM on Semiconductors" from
Scanning Tunneling Microscopy I, Springer
Series in Surface Sciences, 1992.
[18] C. D. Ruggiero, T. Choi and J. A. Gupta, Appl.
Phys. Lett., 2007, 91, 253106.
[19] D. S. Sutar, P. K. Narayanam, G. Singh, V. D.
Botcha, S. S. Talwar, R. S. Srinivasa and S. S.
Major, Thin Solid Films, 2012, 580, 5991-5996.
[23] Gulbagh Singh, D S sutar,V Divakar Botcha, P K
Narayanam, S S Talwar, RS Srinivasa and S S
Major, Nanotechnology, 2013, 24, 355704.
[24] Gulbagh Singh, V Divakar Botcha, D S Sutar, P K
Narayanam, S S Talwar, RS Srinivasa and S S
Major, Physical Chemistry Chemical Physics, 2014,
16, 11708-11718.
[25] Akshay Mathkar, Dylan Tozier, Paris Cox, Peijie
Ong, Charudatta Galande, Kaushik Balakrishnan,
Arava Leela Mohana Reddy and Pulickel M.
Ajayan, J. Phys.Chem.Lett., 2012, 3, 986-991.
Supplementary Information
GO solution preparation
Modified Hummer's method was used for the synthesis of GO. GO sheets were synthesized by chemical exfoliation of graphite
powder using sodium nitrate (NaNO3), sulphuric acid (98% conc. H2SO4), potassium per magnet (KMnO4), hydrogen peroxide
(H2O2) and milli-Q water. A round flask having a magnetic bead inside it, was placed inside the bowl which was partially
filled with the ice. This bowl was placed over a magnetic stirrer and 0.5 g of graphite and 0.5 g of sodium nitrate (NaNO3)
were poured in the flask. After some time, the 23 ml of 98% conc.H2SO4 drop wise through wall was added, when addition of
H2SO4 got completed then 3 g of (KMnO4) was added slowly. After 15 minutes of the completion of KMnO4 the round flask
with solution was shifted on oil bath and the temperature was set to 40°C and maintained for 45-50 minutes. Then 40 ml milli-
Q water (resistivity 18.2 MΩ-cm) was added and the temperature was set to 90°C. Reaching the temperature at 95°C, the
mixture was diluted by adding the 100ml milli-Q water, followed by the addition of 3 ml of H2O2, which turns the colour of
solution from brown to yellow. This warm solution was filtered by using Teflon membrane filter of 0.5μm porousness and the
solution was washed by milli-Q water several times. The filter cake was then diffused in milli-Q water and the solution
(suspension) was slowly shaken with hand. The suspension was ultra-sonicated for 9-10 seconds. The suspension was
centrifuged four times (2 minute each) at 1000 rpm, during this process the supernatant was collected each time for subsequent
centrifugation and all the un-exfoliated as well as heavy GO particle were removed in this process. After that, the supernatant
was centrifuged two times at 8000 rpm for 15 minutes, and at end of each centrifugation, the sediment was collected and
diffused in 60 ml of milli-Q water. Further centrifugation of this solution two times at 8000 rpm for 20 minute and after this
step collected the precipitate, dissolved in 60 ml of milli-Q water and methanol in 1:5. The dispersion was centrifuged at 1000
rpm and 2500 rpm for 10 minute each, to remove the precipitate. The final supernatant solution (GO suspension) was collected
and called the stock solution, which was tested to standardize the GO content. For this, 20 micro liter of master solution was
diluted by adding 3 ml of milli-Q water + methanol solution (1:5). Its UV-vis absorbance spectrum was recorded was recorded
to observe the peak at ~ 230 nm along with a shoulder at ~ 290 nm, corresponding to -- * transition of C=C and n→*
transition of the C=O, respectively. If the absorbance at 230 nm was ~ 0.1, then the stock solution was used as a spreading
solution for LB and MLB deposition. In case, the absorbance was higher than ~ 0.1 (i.e in the range of 0.2 to 0.3), the master
solution was suitably diluted and then tested again as described above, to obtain an absorbance value of in the range of 0.08 -
0.14 0.1. The diluted solution was then used as the spreading solution. A typical absorbance spectrum of the spreading solution
is shown in Figure 2.
Figure 2. The UV-visible spectrum of as prepared GO solution
Substrate cleaning
Before going for deposition, one need to clean the substrates with proper standard procedure. The substrates used in deposition
is highly n-doped Si. The cleaning process followed is as follows.
First the substrate was cleaned ultrasonically with milli-Q water followed by ultrasonic treatment with acetone and isopropyl
alcohol. Then a 1:1:2 mixture of aqueous ammonia, hydrogen peroxide (30%) and milli-Q water was heated to 80 degree C
having substrate in it. It is called RCA treatment. The duration of RCA treatment is 20 minutes. Ultimately the substrate will
be rinsed with milli-Q water and will be preserved in milli-Q water only.
GO deposition
The deposition setup for MLB consists of glass beaker of nominal height of 10-11 cm and nominal diameter of 9-10 cm. This
beaker is called reservoir. This reservoir contains almost 500 cc volume and 64 𝑐𝑚2 top surface area. At the bottom of the
reservoir, a drain tube is connected. A plastic tube, having a control valve, is connected to the drain tube. This plastic tube is
meant for the controlled draining of subphase from the reservoir. To suspend the substrate vertically into the subphase, a glass
rod held across the rim of the beaker.
The reservoir was cleaned up with soap solution and rinsed with Milli-Q water. For, GO deposition the subphase is Milli-Q
water itself, so the reservoir is filled with Milli-Q water up to the 85-90% of its height. Then, the wilhelmly plate is inserted
into the subphase. After this the substrate is inserted into the subphase. Now, surface pressure is set to zero and the GO
spreading solution is spread over the subphase with a micro syringe and let it settle over the surface for 30-40 minutes. After
this, the water outlet is opened to let the meniscus go down. During this downward meniscus movement, the GO sheets will
get transferred onto the substrate. The MLB process is shown schematically in Fig.8.
Figure 3. Successive stages of transfer of GO monolayer sheets by MLB technique: (a) milli-Q water as the
subphase in a reservoir, (b) substrate inserted in the subphase, (c) GO solution is spread and (d) transfer of
GO sheets by draining of the subphase.
With the addition of spreading solution, the surface pressure is found to increase, and the corresponding calibration curve is
shown in Fig.7 (a). During MLB deposition in this work, 2 - 4 ml of spreading solution (depending on the concentration of
GO solution) was used obtain a surface pressure of 6 -7 mN/m. In MLB, a new parameter which comes into picture is the
meniscus speed. A calibration curve of the meniscus speed against the subphase draining rate is shown in Fig. 7 (b). The
meniscus speed in the present work was chosen as 3 -5 mm/min based on the optimization. During MLB transfer the surface
pressure does not change by more than 5%. This gives us a liberty to play with parameters like subphase pH, surface pressure
and meniscus speed.
Figure 7. The plots of (a) surface pressure vs volume and (b) meniscus speed vs flow rate, of the spreading GO
solution.
XPS spectrum after reduction
Figure 8. Survey scans of ammonia plasma treated GO.
Figure 9. The de-convoluted C-1s spectra of and (a) GO (b) Hydrazine treated GO. (c) 30 sec hydrogen plasma
treated GO with 15 watt power at 500C and (d) 5 min ammonia plasma treated GO with 10 watt power at
room temperature.
De-convoluted peak positions(eV)
𝑠𝑝2 − 𝐶
𝑠𝑝3 − 𝐶
C-O
or
C=O
or
N-(𝑠𝑝2 − 𝐶)
N-(𝑠𝑝3 − 𝐶)
COOH 𝜋 − 𝜋∗
O/C ratio
(%)
GO
284.4
(46)
285.3
286.4
(9)
(20)
287
(18)
288.5
289.4
(6)
(1)
Ammonia
Plasma
treated
GO
(5Min
Plasma)
Hydrogen
Plasma
treated
GO
(30 sec
Plasma)
Hydrazine
treated
GO
285
(44)
285.5
286.4
287.1
288.4
289.6
(17)
(14)
(9)
(11)
(5)
284.5
285.4
286.2
287.2
288.3
289.4
(60)
(18)
(12)
(5)
(3)
(2)
284.5
285.6
286.2
287.2
288.4
289.5
(74)
(9)
(8)
(4)
(3)
(2)
50
-
20
18
Raman Spectrum
Figure 10. The Raman spectrum of as prepared GO.
Figure 11. The Raman spectrum of hydrogen plasma (30 sec at 500C) treated GO.
Figure 12. The Raman spectrum of ammonia plasma (5 min at Room Temperature) treated GO.
|
1709.01789 | 1 | 1709 | 2017-08-29T14:45:07 | Novel magnetoelectric effects via penta-linear interactions | [
"physics.app-ph",
"cond-mat.other"
] | Magnetoelectric multiferroic materials, particularly with the perovskite structure, are receiving a lot of attention because of their inherent coupling between electrical polarization and magnetic ordering. However, very few types of direct coupling between polarization and magnetization are known, and it is unclear whether they can be useful to the design of novel spintronic devices exploiting the control of magnetization by electric fields. For instance, the typical bi-quadratic coupling only allows to change the magnitude of the magnetization by an electric field, but it does not permit an electric-field-induced switching of the magnetization. Similarly, the so-called Lifshitz invariants allow an electric-field control of complicated magnetic orderings, but not of the magnetization. Here, we report the discovery of novel direct couplings between polarization and magnetization in epitaxial perovskite films, via the use of first-principles methods and the development of an original Landau-type phenomenological theory. Our results feature penta-linear interactions involving the ferromagnetic and anti-ferromagnetic vectors as well as the polar distortions and oxygen octahedral tilting, and permit a number of striking effects. Examples include a continuous electric-field control of the magnetization magnitude and sign, and the discrete switching of the magnetization magnitude. Thus, the high-order, penta-linear couplings demonstrated in this work may open new paths towards novel magneto-electric effects, as well as, spintronic and magnonic devices. | physics.app-ph | physics | Novel magnetoelectric effects via penta-linear interactions
Hong Jian Zhao1, 2, M. N. Grisolia3, Yurong Yang2, Jorge Íñiguez4,5, M. Bibes3, Xiang
Ming Chen1,*, and L. Bellaiche2,*
1Laboratory of Dielectric Materials, School of Materials Science and Engineering,
Zhejiang University, Hangzhou 310027, China.
2Physics Department and Institute for Nanoscience and Engineering, University of
Arkansas, Fayetteville, Arkansas 72701, USA.
3Unité Mixte de Physique CNRS/Thales, 1 Av. A. Fresnel, Campus de l'Ecole
Polytechnique, 91767 Palaiseau and Université Paris-Sud, 91405 Orsay, France.
4Materials Research and Technology Department, Luxembourg Institute of Science
and Technology, 5 avenue des Hauts-Fourneaux, L-4362 Esch/Alzette, Luxembourg.
5Institut de Ciència de Materials de Barcelona (ICMAB-CSIC), Campus UAB, 08193
Bellaterra, Spain.
*email: [email protected] and [email protected]
Magnetoelectric multiferroic materials, particularly with the perovskite structure, are
receiving a lot of attention because of their inherent coupling between electrical
polarization and magnetic ordering. However, very few types of direct coupling
between polarization and magnetization are known, and it is unclear whether they can
be useful to the design of novel spintronic devices exploiting the control of
magnetization by electric fields. For instance, the typical bi-quadratic coupling only
allows to change the magnitude of the magnetization by an electric field, but it does
not permit an electric-field-induced switching of the magnetization. Similarly, the
so-called Lifshitz invariants allow an electric-field control of complicated magnetic
orderings, but not of the magnetization. Here, we report the discovery of novel direct
couplings between polarization and magnetization in epitaxial perovskite films, via
the use of first-principles methods and the development of an original Landau-type
phenomenological theory. Our results feature penta-linear interactions involving the
ferromagnetic and anti-ferromagnetic vectors as well as the polar distortions and
oxygen octahedral tilting, and permit a number of striking effects. Examples include a
continuous electric-field control of the magnetization magnitude and sign, and the
discrete switching of the magnetization magnitude. Thus, the high-order, penta-linear
couplings demonstrated
towards novel
this work may open new paths
magneto-electric effects, as well as, spintronic and magnonic devices.
in
1
I. Introduction
include
(1)
the Lifshitz
Discovering multiferroic materials possessing large or novel couplings between their
electric and magnetic orders is an important current research direction. It has the
potential to deepen the fundamental knowledge of condensed matter physics, and to
result in new technological applications in, e.g., the field of spintronics. Three main,
general types of magneto-electric (ME) coupling have been revealed so far in
multiferroic perovskites. The first one is the ``traditional'' direct bi-quadratic
coupling between the magnetization, M, and the electrical polarization, P, and its
associated energy is of the form E ~ P2 M2.[1] As a result of such a coupling,
application of an electric field modifies the magnitude of the magnetization via the
electric-field-induced linear change in the polarization. The second kind of ME
coupling gathers the terms that are linear in polarization and involves two different
invariants ΔE ~
magnetic quantities. Examples
P·[L(·L)+L×(×L)] and ΔE ~ P·[M(·M)−(M·)M)] [2-3] where L is an
antiferromagnetic vector, and (2) the spin-current model [4-6] for which ΔE ~
(P×eij)·(mi×mj), where eij is the unit vector joining the magnetic cations at sites i and j
whose magnetic moments are given by mi and mj, respectively. This second type is
responsible for non-trivial effects, such as the electric-field-driven change of the
propagation direction [7,8] and reversal of the chirality [9] of magnetic cycloids.
[Note that couplings being linear in P and potentially permitting magnetization
switching can occur in other less-studied structures, such as ilmenites [10].] The third
kind of ME effects is based on indirect couplings of magnetization and polarization.
An illustrative example recently reported relies on the existence of two different
trilinear terms: a first energy, E ~ PQ1Q2, couples the electrical polarization with
two other structural degrees of freedom, Q1 and Q2 (which can, e.g., be two oxygen
octahedral tiltings [11-14], or one octahedral tilting and an antiferroelectric mode
[15]); and a second trilinear energy, E ~ Q1 mP ms, that couples one of the two other
structural degrees of freedom with the predominant magnetic order parameter, mP, as
well as with a secondary magnetic order parameter, ms [16]. The existence of the first
2
trilinear energy implies that the switching of the polarization reverses Q1 (which is
assumed to be softer than Q2), which in turn reverses ms, because of the second
trilinear term [15]. This third type of coupling between polarization and a magnetic
quantity is therefore of indirect nature.
Today, after two decades of intense work, the possibilities for direct ME couplings
seem exhausted, and most of the research on novel ME effects tends to focus on
indirect effects as those described above [17]. Indeed, it seems hard to believe that
new energies can couple directly polarization and magnetization to yield useful
effects.
Here, we report results of first-principles calculations that show that such novel
couplings, in fact, exist. More precisely, our calculations and analysis of CaMnO3 thin
films reveal high-order (penta-linear) interactions that directly couple polarization and
magnetization, as well as the antiferromagnetic vector and oxygen octahedral tiltings
that characterize the investigated structures. Further, we predict that such couplings
permit electric-field-control of the magnitude and sign of the magnetization, by
inducing either discrete or continuous changes, providing a new and exciting
playground for ME effects.
This article is organized as follows. Section II provides details about the
first-principles method used here. Section III reports and describes the first-principles
results for CaMnO3 thin films, along with the development of a novel Landau-type
phenomenological model to analyze and understand them. Finally, a perspective about
the applicability of the present results to other systems, as well as about other related
phenomena, is given in Section IV.
II. Method
First-principles calculations are performed using the VASP code [18] within the
framework of local spin density approximations (LSDA) [19]. Note that, as in Ref.
[20], we introduce no Hubbard correction for the Mn4+ ion. The projector augmented
3
wave (PAW) scheme [18] is employed in the present calculations, with the following
electrons explicitly considered: calcium's 3p6 and 4s2, manganese's 3d6 and 4s1, and
oxygen's 2s22p4. The energy cutoff is chosen to be 500 eV. We presently consider four
different phases of epitaxial CaMnO3 films, namely the non-polar, so-called c-ePbnm
and ab-ePbnm phases [21,22], and the polar Pmn21 and Pmc21 states. The lattice
vectors of the c-ePbnm, Pmn21 and Pmc21 phases are given by a=aIP(x+y),
b=aIP(-x+y) and c=(2aIP+δ)z, where x, y and z are the unit vectors along the
pseudo-cubic [100], [010] and [001] directions, respectively, and where δ is a
coefficient to be relaxed while aIP is the in-plane lattice constant of the substrate (the
pseudo-cubic setting is used throughout in this paper). Figure 1 schematizes the
epitaxial growth of the c-ePbnm, Pmn21 and Pmc21 states on a cubic substrate.
Regarding ab-ePbnm, its lattice vectors are chosen to be a'=2aIPx, b'=2aIPy and
c'=δ'x+δ"y+2aIPz, where δ' and δ" are parameters to be optimized via structural
relaxation, and its growth on a cubic substrate is also depicted in Fig. 1. A k-point
mesh of 6×6×4 is used for c-ePbnm, Pmn21 and Pmc21 structures, while the selected
k-point mesh is 4×4×4 for ab-ePbnm (note its larger unit cell). Note that all four
phases considered exhibit an a-a-c+ pattern (in Glazer's notation [23]) for the O6
octahedral tilting, where the plus and minus super-scripts indicate anti-phase and
in-phase rotations, respectively. For the c-ePbnm, Pmn21 and Pmc21 states, the overall
anti-phase tilting occurs about the b axis, while the in-phase tilting is about the c-axis.
Regarding the ab-ePbnm state, the anti-phase tilting is about an axis being close to
a'+c' and the in-phase tilting is about a'. Note also that the electrical polarization in
the Pmc21 state lies along the pseudo-cubic [-110] direction (i.e., along the b axis),
that is, it is parallel to the axis of the anti-phase tilting. In contrast, the electrical
polarization in the Pmn21 phase is oriented along the perpendicular pseudo-cubic [110]
direction.
For our structural relaxations, we keep the in-plane lattice parameter (aIP) fixed and
optimize all the other degree of freedoms, i.e., the aforementioned δ, δ', and δ"
parameters and the ionic positions (until the force on each ion is converged within
4
0.005 eV/Å). The space groups of the resulting relaxed states are determined by using
the FINDSYM software [24], and the electric polarization is calculated by the Berry
phase method [25].
Regarding magnetic properties, we choose the predominant magnetic ordering to be
G-type antiferromagnetic, as in Ref. [20]. In other words, the spin vectors of
nearest-neighboring Mn4+ ions are anti-parallel to each other, as consistent with the
known magnetic configuration of bulk CaMnO3. [26] Spin-orbital coupling is also
included when determining the non-collinear magnetic structure. As shown in Fig. 2,
we consider two different cases: Case (1) for which the predominant G-type
antiferromagnetic vector, G, is lying along the pseudo-cubic [001] direction; and Case
(2) in which G is oriented along the pseudo-cubic [110] direction. The relaxed spin
configurations in Cases (1) and (2) are, in fact, the so-called Γ2(Fa, Cb, Gc) and Γ4(Ga,
Ab, Fc) magnetic states [27]. The Γ2 configuration possesses a weak magnetization
along the pseudo-cubic [110] direction (i.e., along the a vector defining the
orthorhombic cell) and a weak C-type antiferromagnetic vector lying along the
pseudo-cubic [-110] axis (i.e., along the b cell vector). On the other hand, for the Γ4
magnetic state, the weak magnetization is directed parallel or antiparallel to the
pseudo-cubic [001] (c cell vector), and the other (weak) antiferromagnetic vector is
now of A-type and is along the pseudo-cubic [-110] (b cell vector). These additional
weak magnetic vectors and their directions can be naturally explained by the universal
law given in Ref. [16], which couples dominant and weak magnetic orderings with the
in-phase and anti-phase tiltings of oxygen octahedra.
this work. Most
Note that, while (as indicated above) we used a plain LSDA functional to treat
CaMnO3, we also checked the effect that "Hubbard U" corrections have in the key
results of
the bare LSDA calculations yield
magnetizations that are considerably smaller (0.05 μB/f.u. and -0.0498 μB/f.u. for the
Γ2 and Γ4 spin configuration, respectively) than those obtained from a LSDA+U
calculation with an effective Hubbard Ueff of 3.0 eV (0.121 μB/f.u. and -0.118 μB/f.u.,
respectively) for bulk CaMnO3. Hence, the effects we discuss are predicted to be
importantly,
5
quantitatively stronger when a Hubbard correction is included in the simulations. Let
us also note that our work has similarities with the investigation reported in Ref. [20],
frequently cited in this paper. However, here we go one step further and discuss the
two aforementioned (as opposed to only one in Ref. [20]) polar structures that
CaMnO3 is predicted to present under tensile epitaxial strain, such a multi-stability
being the key to the remarkable magnetoelectric effects we have discovered.
III. Results
A) First-principles results
Let us first check the accuracy of our simulations to predict structural properties.
For that, we perform energy optimization of the CaMnO3 bulk in its Pbnm ground
state, and find that our a, b, and c equilibrium (orthorhombic) lattice parameters are
equal to 5.141, 5.200, and 7.292 Å, respectively. These computed equilibrium lattice
parameters are in excellent agreement with the first-principles predictions of Ref. [20],
which reports 5.161, 5.205, and 7.309 Å, respectively. On the other hand, as typical of
LSDA calculations, they underestimate the corresponding experimental data of Ref.
[26] (5.264, 5.278 and 7.455 Å), by about 2.3%, 1.5%, and 2.2%, respectively.
As in Ref. [20], we define the misfit strain experienced by the mimicked
CaMnO3 films as η(aIP-a0)/a0, where a0 is the average value of the a and b lattice
parameters (divided by √2) of the simulated ground state of bulk CaMnO3. Figure 3
reports the total energy of the studied four phases as a function of η, in case of tensile
strain (that is, for positive η) when adopting collinear magnetism. The ab-ePbnm
phase, which strictly speaking has monoclinic P21/m symmetry, is the ground state for
(weak) misfit strains having a magnitude smaller or equal to ~ 0.5%. Then, c-Pbnm is
the most stable phase for η between ~ 0.5% and ~ 3.9%. For higher strains, as
emphasized by the inset of Fig. 3, our calculations predict that the CaMnO3 films
become polar, the ground state adopting the Pmn21 space group. Further, a second
polar structure with Pmc21 symmetry is stabilized as well. On one hand, this finding is
consistent with the paraelectric-to-ferroelectric transition reported in Ref. [20], for
6
which the critical misfit strain is 3.2%. (The precise value of this critical strain
depends on technical details of density-functional calculations [20] and should likely
decrease as the temperature increases – as a result of the typical temperature-induced
reduction of the electrical polarization.) On the other hand, the predicted polar ground
state in Ref. [20] is reported to have Pmc21 symmetry, while our analysis rather yields
Pmn21. Interestingly, in an investigation of the related compound CaTiO3, Ecklund et
al. [21] found not only the same two polar structures (Pmn21 and Pmc21) but also the
same energy hierarchy between them under tensile strain that we find here for
CaMnO3, supporting the correctness of our results and analysis.
Interestingly, Fig. 3 also shows that the difference in energies between Pmn21
and Pmc21 is enhanced as the strain increases above its critical value of 3.9%.
However, this latter energy difference is only of the order of 0.01 eV/f.u. for a strain
as large as 7%, and both polar states Pmn21 and Pmc21 are more stable than the
non-polar ab-ePbnm and c-Pbnm phases for tensile strains larger than 3.9%. It is
therefore reasonable to wonder whether there is an "easy" structural path allowing the
Pmn21 phase to transform into the Pmc21 state, and vice-versa, by, e.g., the
application of electric fields. In order to resolve such issue, we first select a specific
misfit strain, namely 5.8%, and compute the energy vs polarization curves for the
Pmn21 and Pmc21 phases at this strain value. (For any η between 3.9% and 7% the
results are qualitatively the same.) These curves are displayed in Fig. 4, where we
take the c-ePbnm state as the reference phase for determining the polarization. They
indicate that the equilibrium Pmn21 ground state has an electric polarization of about
29.3 μC/cm2 oriented along the pseudo-cubic [110] direction, while the metastable
Pmc21 state exhibits an electrical polarization of about 24.8 μC/cm2 lying along the
perpendicular [-110] direction. The energy along the bridging path between the Pmn21
and Pmc21 minima is shown in Fig. 4 as well. These bridging structures therefore
possess a polarization having non-zero components along both the pseudo-cubic [110]
and [-110] axes, and thus adopt the monoclinic Pm space group. Figure 4 reveals that
an energy barrier of 6.5 meV/f.u. (respectively, 0.5 meV/f.u.) needs to be overcome to
7
transit from the Pmn21 ground state to the stable Pmc21 phase (respectively, when
going from Pmc21 to Pmn21). These energy barriers are therefore rather small. As a
result, applying realistic electric fields along the pseudo-cubic [-110] direction should
allow to switch from the Pmn21 ground state to the metastable Pmc21 state, and going
back to Pmn21 from Pmc21 should be feasible via the application of moderate fields
along the pseudo-cubic [110] axis. (Note that the application of in-plane electric fields
has been demonstrated in, e.g., Ref. [28].) We will come back to, and take advantage
of, this possibility of easily switching back and forth between Pmn21 and Pmc21 later
on.
Figure 5a shows the weak magnetization of the CaMnO3 films, at a tensile misfit
strain of 5.8%, when varying the magnitude and sign of the electrical polarization in
the Pmn21 and Pmc21 states (note that such variations can be practically done by
applying electric fields). One can see that, within the range of investigated
polarizations, the weak magnetization associated with Γ2 or Γ4 reaches its maximum
value when the polarization vanishes -- that is, when the corresponding structural
phase is, in fact, the paraelectric c-ePbnm. Moreover, an interesting magneto-electric
effect develops for the four different cases (Pmn21 and Pmc21 states with either Γ2 or
Γ4 spin structure), when the polarization is switched on and varies: the magnetization
significantly responds to the magnitude, but not the sign, of the polarization. As a
result and as depicted in Fig. 5a with horizontal arrows, reversing the polarization
from positive to negative between the two equilibrium Pmc21 phases– via, e.g., the
application of an electric field along the pseudo-cubic [1-10] direction– should not
change the direction nor the magnitude of the weak magnetization for both the Γ2 and
Γ4 spin configurations. As shown in Fig. 5a too, this insensitivity to a polarization
reversing of the weak magnetization in these two spin structures should also be found
for the equilibrium Pmn21 phases – by, e.g., applying an electric field along [-1-10].
As schematized in Fig. 5b and indicated by Fig. 5a (by comparing, at the equilibrium
polarization values, the solid and empty symbols having the same color), changing the
8
magnetic structure from Γ2 to Γ4 within the equilibrium Pmc21 (or Pmn21) structure –
via, e.g., the application of a magnetic field along [00-1] – has essentially no effect on
the magnitude of the weak magnetization. In that case, the magnetization "only"
rotates from the [110] to [00-1] pseudo-cubic directions. Furthermore, as also shown
in Fig. 5a, the aforementioned response of the magnetization to polarization leads to a
critical value of about 36 μC/cm2 at which a full vanishing of the magnetization in the
Pmn21 phase is obtained. In other words, we predict that the magnetization of the
equilibrium Pmn21 phase can be greatly controlled when applying an electric field
along the direction of its polarization. Such magnetization can even be reversed
depending on the magnitude of this electric field.
Interestingly, there is yet another magneto-electric feature that is revealed by Fig.
5a, and which may also be put in use to design novel devices. Note that, for both the
Γ2 and Γ4 magnetic configurations, the magnitude of the weak magnetization
decreases faster with the magnitude of the polarization in the Pmn21 phase than it does
in the Pmc21 phase. As a result, and as further stressed in Fig. 5a by means of oblique
arrows, the structural path of Fig. 4 going from the Pmn21 ground state to the
meta-stable Pmc21 state should result in a significant change of the magnitude of the
weak magnetization. This change is about 0.038 μB/f.u. for the Γ2 spin configuration
and about 0.035 μB/f.u. for Γ4, which correspond to an enhancement of the
magnetization by almost of a factor of 2. In other words, as sketched in Fig. 5b, an
electric-field-driven transition between the ground state and the metastable phase
should be accompanied by a large magnetization change. This novel magneto-electric
effect may also offer the opportunity to manipulate not only the magnetization
amplitude but also its dynamics by applying successively electrical pulses in the [110]
and [-110] pseudo-cubic directions, allowing to go back and forth between the ground
state and the metastable phase. One can also envision to apply a dc electric field in
one direction (e.g., [110]) and a pulse in the perpendicular direction (e.g., [-110]) to
control with time the change of the magnetization amplitude.
9
B) Development of a phenomenological model
In order to explain all the results summarized in Fig. 5, we further develop a
Landau-type phenomenological model with the energy given by:
2R,xGyFz Py
E F 2 F 2P2 (R,xGyFz R,xGzFy R,yGzFx R,yGxFz R,zGxFy R,zGyFx )
2R,xGzFy)
(Px
1(PxPyR,xGxFz PxPyR,yGyFz PzPxR,zGzFy PzPyR,zGzFx PyPzR,yGyFx PzPxR,xGxFy)
2(PxPyR,xGzFx PxPyR,yGzFy PzPxR,zGyFz PzPyR,zGxFz PyPzR,yGxFy PzPxR,xGyFz)
2R,zGxFy Pz
2R,yGxFz Pz
2R,zGyFx Py
2R,yGzFx Px
(1)
where x, y and z subscripts refer to Cartesian components along the pseudo-cubic
[100], [010] and [001] directions. F and G are the weak ferromagnetic vector and
dominant G-type antiferromagnetic vector, respectively, while P is the electric
polarization. ωR is the vector characterizing the anti-phase oxygen octahedral tilting
[29]: its direction is the axis about which the tilt occurs (i.e., the pseudo-cubic [-110]
direction) and its magnitude is the rotation amplitude. Furthermore, κ, λ, α, β, γ1 and γ2
are material-dependent coefficients to be determined. The first term of Eq. (1)
represents the usual harmonic effect associated with the magnetization, while the
second energy characterizes
interaction between
magnetization and polarization [1]. The third term arises from the application of the
universal law of Ref. [16] to a perovskite system possessing a dominant G-type
antiferromagnetic vector altogether with anti-phase oxygen octahedral tilting (which
leads to the occurrence of a weak magnetization). The last three energies of Eq. (1)
have never been considered before, to the best of our knowledge, even if they are
allowed by symmetry (in particular, they obey inversion- and time-reversal symmetry,
and the sum of the k-points associated with all these terms adds up to zero). They
characterize penta-linear interactions between different Cartesian components of the
electrical polarization, anti-phase oxygen octahedral tiltings, G-type antiferromagnetic
vector and magnetization.
typical bi-quadratic
the
Applying Eq. (1) to the four different cases depicted in Fig. 5 leads to the following
10
energies:
E(Pmn21,2) F 2 F 2P2 GF
2
E(Pmn21,4) F 2 F 2P2 GF
2
E(Pmc21,2) F 2 F 2P2 GF
2
E(Pmc21,4) F 2 F 2P2 GF
2
P2GF
P 2GF
P 2GF
P2GF
2
2
1
2
2
2
1
2
P 2GF
P 2GF
(2)
P 2GF
P 2GF
Here, F, , G and P are the projections of the magnetization, anti-phase tilting vector,
antiferromagnetic vector and electrical polarization along their corresponding
direction, respectively (e.g., F is the projection of the magnetization along the
pseudo-cubic [110] direction for the 2 magnetic configuration). Minimizing these
energies with respect to the magnetization, and assuming that P2/κ is small with
respect to unity, yields:
M (Pmn21,2) G
2
M (Pmn21,4) G
2
M (Pmc21,2) G
2
M (Pmc21,4) G
2
(1P2 /)
(2)P2G(1P 2 /)
4
(1P 2 /)
(1)P2G(1P2 /)
4
(1P 2 /)
(2)P2G(1P2 /)
4
(3)
(1P2 /)
(1)P 2G(1P2 /)
4
Remarkably, and as shown in Fig. 5 by means of solid and dashed lines, the
first-principles results for the magnetization can be fitted extremely well by these
Equations, the resulting fitting parameters being /κ=-2.85 m4C-2, α/κ=-0.008 deg-1,
β/κ =0.08 deg-1m4C-2, γ1/κ=0.04 deg-1m4C-2, and γ2/κ =-0.04 deg-1m4C-2. (These fits
are done by taking ωR and G to be constant and equal to 9.15 degrees and 2.39 μB/f.u.,
respectively, as given by our LSDA results.) Such an excellent fit therefore attests the
validity of the proposed Landau-type-model, which can thus be safely used to
understand the numerical results depicted in Fig. 5. For instance, the right-hand side
of Eqs. (3) only contain even orders (namely, second and fourth orders) of P, which
11
explains why the magnetization depends on the magnitude, but not sign, of the
polarization. The first term on the right-hand side of Eqs. (3) is also consistent with
the first-principles results of Fig. 5a showing that the magnetization decreases in
magnitude for increasing polarization values, since α/κ and /κ have the same sign.
Similarly, the second term on the right-hand side of Eqs. (3) also explains why the
magnitude of the magnetization is more strongly reduced when increasing the
polarization from zero in the Pmn21 state than in the Pmc21 phase for the Γ2
(respectively, Γ4) spin configuration, as β/κ is positive while γ2/κ is negative
(respectively, β/κ and γ1/κ are both positive). As a result, the penta-linear couplings of
Eq. (1), which lead to the 4th-order couplings of Eq. (3), are primordial to understand
the unusual magneto-electric effects displayed by CaMnO3 films.
IV. Further Discussion
Importantly, these penta-linear terms should also be valid in other perovskites
since their existence solely arises from symmetry considerations. In fact, these terms
become active in structures derived from the orthorhombic phases that are most
abundant among perovskite oxides, suggesting there are plenty of opportunities to
find alternative materials presenting the effects described here, including at more
moderate values of epitaxial strain. For example, CaTiO3 is predicted to display the
two polar phases of interest here at a moderate 2% epitaxial expansion [21]. This
observation led us to consider (CaTiO3)3/(CaMnO3)3 superlattices in an attempt to
reduce the epitaxial mismatched required to obtain the effects of interest. We found
that, in such superlattices, our two polar states can be stabilized at an epitaxial
mismatch of ~3.0%, therefore reducing the value of 3.9% obtained for pure CaMnO3.
Our results also clearly indicate which are the experimental fingerprints of the
presence and importance of such penta-linear couplings, i.e., large changes of
magnetization when switching between states with differently-oriented polarization
and/or a strongly anharmonic magnetoelectric response of a particular polar state.
Further, Eq. (3) also implies the existence of other features, such as the magnetization
12
reverting its direction when switching the anti-phase oxygen octahedral tilting or the
G-type antiferromagnetic vector. Subsequent first-principles calculations confirm such
features, further demonstrating the validity of our phenomenological theory. We thus
hope that the penta-linear couplings revealed in this work will lead to the
experimental discovery of novel magneto-electric materials, and that the variety of
novel static and dynamical effects they permit will motivate the design of the
next-generation of spintronic and magnonic devices.
Acknowledgements
This work was supported by the National Natural Science Foundation of China under
Grants Nos. 51332006 and 11274270 (X.M.C.), the U.S. Department of Energy,
Office of Basic Energy Sciences, under contract ER-46612 (L.B.), the ERC
Consolidator Grant #615759 MINT
(M.B.), FNR Luxembourg Grants
FNR/P12/4853155/Kreisel (J.I.) and INTER/MOBILITY/15/9890527 GREENOX
(L.B. and J.I.), and MINECO-Spain Grant MAT2013- 40581-P (J.I.).
References
[1] T. Kimura, S. Kawamoto, I. Yamada, M. Azuma, M. Takano, and Y. Tokura, Phys.
Rev. B 67, 180401 (2003).
[2] I. Sosnowska and A. Zvezdin, J. Magn. Magn. Mater. 140-144, 167 (1995).
[3] M. Mostovoy, Phys. Rev. Lett. 96, 067601 (2006).
[4] H. Katsura, N. Nagaosa, and A. V. Balatsky, Phys. Rev. Lett. 95, 057205 (2005).
[5] D. Rahmedov, D. Wang, J. Íñiguez, and L. Bellaiche, Phys. Rev. Lett. 109, 037207
(2012).
[6] A. Raeliarijaona, S. Singh, H. Fu, and L. Bellaiche, Phys. Rev. Lett. 110, 137205
(2013).
[7] D. Lebeugle, D. Colson, A. Forget, M. Viret, A.M. Bataille, and A.
Goukasov, Phys. Rev. Lett.100, 227602 (2008).
[8] S. Lee, T. Choi, W. Ratcliff, R. Erwin, S.-W. Cheong, and V. Kiryukhin, Phys. Rev.
B 78, 100101 (2008).
[9] S. Bhattacharjee, D. Rahmedov, D. Wang, J. Íñiguez, and L. Bellaiche, Phys. Rev.
Lett. 112, 147601 (2014).
[10] C.J. Fennie, Phys. Rev. Lett. 100, 167203 (2008).
[11] J. M. Rondinelli and C. J. Fennie, Adv. Mater. 24, 1961 (2012).
[12] J. Young and J. M. Rondinelli, Chem. Mater. 25, 4545 (2013).
[13] A. T. Mulder, N. A. Benedek, J. M. Rondinelli, and C. J. Fennie, Adv. Funct.
13
Mater. 23, 4810 (2013).
[14] H. J. Zhao, J. Íñiguez, W. Ren, X. M. Chen, and L. Bellaiche, Phys. Rev. B 89,
174101 (2014).
[15] Y. Yang, J. Íñiguez, A.-J. Mao, and L. Bellaiche, Phys. Rev. Lett. 112, 057202
(2014).
[16] L. Bellaiche, Z. Gui, and I. A. Kornev, J. Phys.: Condens. Matter 24, 312201
(2012).
[17] J. T. Heron et al., Nature 516, 370 (2014).
[18] G. Kresse, and D. Joubert, Phys. Rev. B 59, 1758 (1999).
[19] J. P. Perdew, and A. Zunger, Phys. Rev. B 23, 5048 (1981).
[20] E. Bousquet and N. Spaldin, Phys. Rev. Lett. 107, 197603 (2011).
[21] C.-J. Eklund, C. J. Fennie, and K. M. Rabe, Phys. Rev. B 79, 220101 (2009).
[22] Q. Zhou, and K. M. Rabe, arXiv:1306.1839.
[23] A. Glazer, Acta Crystallogr. B 28, 3384 (1972).
[24] H. T. Stokes, and D. M. Hatch, J. Appl. Cryst. 38, 237 (2005).
[25] R. D. King-Smith, and D. Vanderbilt, Phys. Rev. B 47, 1651 (1993).
[26] Z. Zeng, M. Greenblatt, and M. Croft, Phys. Rev. B 59, 8784 (1999).
[27] R. W. White, J. Appl. Phys. 40, 1061 (1969).
[28] Z. Chen, X. Zou, W. Ren, L. You, C. Huang, Y. Yang, P. Yang, J. Wang, T.
Sritharan, L. Bellaiche, and L. Chen, Phys. Rev. B 86, 235125 (2012).
[29] I. A. Kornev, L. Bellaiche, P.-E. Janolin, B. Dkhil, and E. Suard, Phys. Rev.
Lett. 97, 157601 (2006).
14
Figures and captions
Fig. 1 (color online). Schematization of the growth of the different, presently
considered structural phases of epitaxial CaMnO3 thin films on cubic substrates.
15
Fig. 2 (color online). Studied magnetic configurations in epitaxial CaMnO3 thin films.
Panel (a) depicts the different types of Mn atoms, while Panels (b) and (c) schematize
the Γ2 and Γ4 spin configurations, respectively.
16
6
7
3
2
5
Misfit strain (%)
4
)
2
m
c
/
C
(
n
o
i
t
a
z
i
r
a
l
o
P
40
30
20
10
0
0
1
0.6
0.5
0.4
0.3
0.2
0.1
)
.
u
.
f
/
V
e
(
y
g
r
e
n
E
0.0
0
1
2
3
4
Misfit strain (%)
c-ePbnm
ab-ePbnm
Pmc21
Pmn21
5
6
7
Fig. 3 (color online). Calculated total energy of epitaxial CaMnO3 thin films (adopting
a collinear G-type antiferromagnetic state) as a function of the tensile misfit strain, for
the four considered structural phases. The inset shows the magnitude of the electric
polarization versus misfit strain for the Pmc21 and Pmn21 states. The zero of energy
corresponds to the c-ePbnm phase at zero strain.
17
Fig. 4 (color online). Energy-versus-polarization curve for the Pmn21 (green line and
symbols) and Pmc21 (blue line and symbols) phases of epitaxial CaMnO3 thin films
for a misfit strain of 5.8%. The red line and symbol show the energy of the
intermediate structures connecting the Pmc21 and Pmn21 states, as a function of their
electric polarization. Pa and Pb represent components of the polarization along the
orthorhombic a and b axes, that are along the pseudo-cubic [110] and [-110] directions,
respectively. The zero of energy is arbitrarily chosen to correspond to the Pmc21 state
having a polarization of ~35 μC/cm2 in magnitude for the misfit strain of 5.8%
18
of
5.8%.
a misfit
Fig. 5 (color online). Predicted magnetoelectric effects in epitaxial CaMnO3 thin films
being
the
magnetization-versus-polarization for the four different cases corresponding to the
combination of the crystal structure among Pmn21 and Pmc21 and the choice of the
magnetic structures among Γ2 and Γ4. This magnetization is along the pseudo-cubic
[110] and [001] directions for the Γ2 and Γ4 spin configurations, respectively, and the
reports
Panel
(a)
under
strain
19
polarization is along the pseudo-cubic [110] and [-110] directions for Pmn21 and
Pmc21, respectively. The symbols represent the predictions from first-principles
calculations, while the blue and red (solid and dashed) curves correspond to the fitting
of these data by the Landau-type-derived Equation (3). The solid and dashed vertical
lines indicate the value of the polarization in the equilibrium Pmn21 and Pmc21 states,
respectively, while the horizontal dashed line depicts the zero in magnetization. The
horizontal (respectively, oblique) solid arrow emphasizes the starting and ending
magnetization, when the system undergoes a transition between two different but
symmetry-equivalent Pmn21 minima having opposite polarization (respectively,
between the equilibrium Pmn21 and Pmc21 states), in case of the Γ2 spin configuration.
Horizontal and oblique dashed arrows show similar data, but for the Γ4 spin
configuration. Panel (b) schematizes
the different possible electric-field or
magnetic-field driven transitions discussed in the text, and their effects on the
directions and magnitude of the magnetization.
20
|
1912.09394 | 1 | 1912 | 2019-12-16T14:37:31 | Harnessing near-field thermal photons with efficient photovoltaic conversion | [
"physics.app-ph",
"physics.optics"
] | A huge amount of thermal energy is available close to material surfaces in radiative and non-radiative states, which can be useful for matter characterization or for energy devices. One way to harness this near-field energy is to scatter it to the far field. Another way is to bring absorbers close to thermal emitters, and the advent of a full class of novel photonic devices exploiting thermal photons in the near field has been predicted in the last two decades. However, efficient heat-to-electricity conversion of near-field thermal photons, i.e. the seminal building block, could not be achieved experimentally until now. Here, by approaching a micron-sized infrared photovoltaic cell at nanometric distances from a hot surface, we demonstrate conversion efficiency up to 14% leading to unprecedented electrical power density output (7500 W.m-2), orders of magnitude larger than all previous attempts. This proof of principle is achieved by using hot graphite microsphere emitters (~800 K) and indium antimonide cells, whose low bandgap energy matches the emitter infrared spectrum and which are specially designed for the near field. These results pave the way for efficient photoelectric detectors converting thermal photons directly in the near field. They also highlight that near-field thermophotovoltaic converters, which harvest radiative thermal energy in a contactless manner, are now competing with other energy-harvesting devices, such as thermoelectrics, over a large range of heat source temperatures. | physics.app-ph | physics | Harnessing near-field thermal photons with efficient
photovoltaic conversion
C. Lucchesi1, D. Cakiroglu2, J.-P. Perez2, T. Taliercio2, E. Tournié2, P.-O. Chapuis1,
R. Vaillon2,1
1Univ Lyon, CNRS, INSA-Lyon, Université Claude Bernard Lyon 1, CETHIL UMR5008, F-
69621, Villeurbanne, France
2IES, Univ Montpellier, CNRS, Montpellier, France
A huge amount of thermal energy is available close to material surfaces in radiative and
non-radiative states1 -- 3, which can be useful for matter characterization4 -- 6 or for energy
devices7,8. One way to harness this near-field energy is to scatter it to the far field4 -- 6.
Another way is to bring absorbers close to thermal emitters9,10,2,3, and the advent of a full
class of novel photonic devices exploiting thermal photons in the near field has been
predicted in the last two decades7,11,12,13,14. However, efficient heat-to-electricity
conversion of near-field thermal photons, i.e. the seminal building block, could not be
achieved experimentally until now. Here, by approaching a micron-sized infrared
photovoltaic cell at nanometric distances from a hot surface, we demonstrate conversion
efficiency up to 14% leading to unprecedented electrical power density output
(7500 W.m-2), orders of magnitude larger than all previous attempts. This proof of
principle is achieved by using hot graphite microsphere emitters (~800 K) and indium
antimonide cells, whose low bandgap energy matches the emitter infrared spectrum and
which are specially designed for the near field15. These results pave the way for efficient
photoelectric detectors converting thermal photons directly in the near field. They also
highlight that near-field thermophotovoltaic converters16, which harvest radiative
thermal energy in a contactless manner, are now competing with other energy-harvesting
devices, such as thermoelectrics, over a large range of heat source temperatures17.
1
A significant number of experimental demonstrations of the enhancement of thermal radiation
heat transfer between a hot and a cold body18 -- 22 establish a firm basis for new photonic devices
based on thermal photons in the near field, involving for instance thermal rectification13 and
photonic cooling14. In particular, it was predicted 20 years ago that near-field effects would
drastically increase thermal radiation transfer between a hot emitter and a photovoltaic cell, and
in turn electrical power generation23 -- 25. Many theoretical results have further elaborated on this
idea8,26 -- 30. An early experiment31 showed some qualitative enhancement of the photogeneration
of electrical charges at microscale gaps between the emitter and the cell. More recently, three
experimental works32 -- 34 reported on an enhancement of the electrical power in the near field,
with estimations of efficiency (when available) of 0.02% and 0.98% and a maximal power
density of 7.5 W.m-2. These modest performances indicate that a clear proof of efficient
photovoltaic conversion of near-field thermal radiation, leading to both significant output
power and conversion efficiency, is still lacking. This is puzzling since thermophotovoltaics,
based on the conversion of radiative heat emitted in the far field, has recently been demonstrated
to be ultra-efficient35. Nevertheless, the electrical power density measured for these far-field
devices remains moderate (approaching the W.cm-2 level35,36 only for high-temperature emitters
above 1000°C) in comparison to other thermal energy-harvesting devices37. Near-field
thermophotovoltaic converters, where the radiative power transferred is much higher, have the
potential to mitigate this issue17.
In the present work, we have specifically designed and fabricated an indium antimonide (InSb)
cell having a very low bandgap energy (0.23 eV, i.e. 5.3 μm, at 77 K), which is well suited for
conversion of thermal radiation emitted from medium-grade heat sources (temperatures in the
range 250-650°C) and therefore resolves one of the obstacles to enhanced efficiency. We have
also identified cell design parameters that allow the maximization of radiative heat-to-
electricity conversion in the near field. In addition, we have chosen the configuration of a
spherical emitter exchanging radiation with a planar receiver18,19,38 -- 41, which both improves the
view factor between the emitter and the cell and allows probing a large emitter-cell distance
range, from millimeters down to nanometers.
2
Figure 1 Device and simulated radiative heat transfer for photovoltaic conversion of near-field
thermal photons. a) Scheme of the emitter-photovoltaic cell device. b) Scanning electron micrograph
of the emitter seen from below. A graphite sphere is glued on the tip of a SThM doped-Si probe with a
ceramic adhesive. c) Optical microscopy top view of an InSb cell. d) Simulated thermal radiation heat
transfer spectrum for the case of a 732 K emitter and 77 K cell for far-field (blue curve) and near-field
(red curve) configurations.
The experimental set-up (see Fig. 1a) involves a micron-sized photovoltaic cell made of a thin
InSb p-on-n junction (view from top in Fig. 1c) and a spherical graphite emitter glued to the
probe of a scanning thermal microscope (SThM, see Fig. 1b). A key feature is that the infrared
photovoltaic cell is specifically designed for harvesting thermal radiation fluxes of medium-
grade temperature sources, around 500°C, in the near field42. It is made of InSb, one of the III-
V semiconductors with the longest bandgap wavelength, to be able to convert photons of
wavelengths in the micrometer range (see Fig. 1d). The bandgap wavelength depends on
temperature and varies from 5.3 μm at 77 K to 7.3 μm at room temperature. The architecture of
the cell was optimized based on full calculations involving the coupling of charge transport and
radiative absorption in a 1D configuration42. The junction is located close to the cell top surface
(at few hundreds of nanometers only) so as to collect a large share of the near-field photons. In
order to avoid having zones of the cells which are not illuminated (generating a dark current
without any photocurrent), the size of the cells is matched to that of the emitter, which increases
the view factor. Thus the cell geometries involve mesas with active-area diameters of the order
of few tens of micrometers (see Methods and Ext. Data Fig. 7a). Gold layers are used as top
and bottom electrical contacts. It is important to underline that InSb can operate as a proper p-
3
n junction only at low temperature15, so in this experiment the cell is cooled down to 77 K by
placing it on the cold finger of a cryostat located in a vacuum chamber (see Methods).
The emitted radiative heat flux is measured by means of a scanning thermal microscope, which
is an atomic force microscope (AFM) where a temperature sensor is located on the cantilever43.
A resistive sensor made of doped silicon is used, the evolution of which as a function of
temperature is known. The temperature of the sphere is estimated to be equal to that of the
resistor and the thermal conductance from the tip temperature to the ambient was measured,
which allowed monitoring variations of the radiative flux lost by the emitter (see Methods and
Ext. Data Fig. 1). Most of the results presented here were obtained for a tip apex temperature
of 732 K (439 °C), which maximizes measurement sensitivity, but the emitter could be heated
up to temperatures larger than 1200 K. Graphite was selected for the emitter as the material
maximizing the near-field radiative exchange with InSb (see Supplementary Information (SI)
and Ext. Data Fig. 2). Fig. 1d indicates that about 40% of the overall power exchanged between
a planar emitter and a planar receiver is located in the wavelength range that can be converted
to electricity. It also shows that reducing the distance between the emitter and the cell strongly
increases radiative exchange. In the experiment, a sphere is therefore approached in the vicinity
of the cell by moving it vertically with a piezo-actuator. A spherical emitter was selected,
despite its reduced near-field exchange with respect to a planar one at the same distance (see
Ext. Data Fig. 2c and SI), in order to eliminate the parallelism issues inherent to planar
configurations. Such geometry was successful for Casimir force and near-field radiative heat
flux experiments18,19,38 -- 41,43 and allows probing a large emitter-cell distance range, from the
sub-100 nm region to millimeters.
4
Figure 2 Current-voltage characteristics of a cell at 77 K having a 20 µm active area diameter. I-
V curves under 300 K ambient illumination (blue), 732 K graphite sphere at d = 1 µm (red) and under
dark conditions (orange, inset). The dark and light gray rectangles represent respectively the maximum
output power configuration and the product VOC ISC, used for computation of the fill factor (FF). The
inset shows the current density in absolute values according to the applied voltage. The curves were
obtained for a 500 µm-thick InSb substrate and a 1017 cm-3 p-doping level.
The cells are first characterized at a fixed distance from the emitter (see Fig. 2). I-V curves in
the dark indicate a low noise level (mA.cm-2) in the experiments. It is remarkable that room-
temperature radiative emission leads to a significant electrical power generation by the cell. The
open-circuit voltage depends on the temperature of the emission source (see diverging values
in the logarithmic scale of the inset of Fig. 2) and is found to be larger than 100 mV at 77 K,
which is around half the bandgap energy of InSb (Eg = 0.23 eV). In contrast to previous
reports33,34, the I-V curves look like usual photovoltaic cell characteristics, i.e. rectangles with
a rounded corner. The short-circuit currents depend on the emitter temperature and reach
already microamps at a distance of 1 μm. Fill factors as high as 0.75 were found, leading to
electrical output powers up to 5.5 μW, depending on the distance.
The high sensitivity of the electrical power output measurement allows following the approach
of the hot sphere towards the cold cell from distances up to millimeters. It is observed that
variations of the short-circuit current as a function of the emitter-sphere distance superimpose
nicely the prediction of far-field view factor theory for these large distances (see inset in Fig.
5
3b and Ext. Data Fig. 8b), indicating that far-field thermophotovoltaic conversion depends only
on the level of illumination.
Figure 3 Near-field exchanged radiative power, maximum generated electrical power and near-
field conversion efficiency obtained simultaneously as a function of the emitter-cell distance. a)
Near-field radiative power according to distance. b) Generated electrical power, deduced from the
measurement of the cell short-circuit current and the application of the superposition principle, as a
function of distance. The total electrical power (red curve) is compared to the power increase due to the
far-field contribution estimated from the view factor (dashed-dotted black curve). The region between
the two curves provides the near-field electrical power. The inset data, obtained with another cell, show
the match between the view factor computation (plain line) and the experimental data in the far field. c)
Near-field conversion efficiency calculated by dividing the near-field electrical power (b) by the near-
field exchanged radiative power (a).
6
We now focus on the near field, which is approximately given by the distance range below
Wien's wavelength (λW.T= 2898 µm.K) for the emitter, i.e. below 4 µm. Our setup allows the
measurement of the radiative exchange as a function of the emitter-cell distance and the increase
of the short-circuit current at the same time. As expected, a strong enhancement of the flux is
observed for small distances, where the exchanged radiative power exceeds 14 μW. Almost all
this increase is due to the near-field contribution, since the gain from the far-field contribution
is much weaker in the last micrometers. While we cannot directly measure the far-field
contribution background, it is computed to be ~7.8 μW for a sphere at 900 K and a cell having
a 20 µm active area diameter (see theoretical analysis in SI and Ext. Data Figs. 2-3).
The electrical output power can be derived from the short-circuit current data obtained as a
function of emitter-cell distance at the same time as the radiative flux data. We show in Ext.
Data Fig. 9 that all I-V curves can be superimposed onto the dark I-V curve by shifting the
short-circuit current to zero ("superposition principle"). Hence by measuring only the short-
circuit current as a function of distance, one can recover all illuminated I-V curves and deduce
the maximum electrical power output as a function of distance (Fig. 3b, red curve). It reaches
almost 2 μW for the selected case, leading to an overall efficiency estimated to be about 9%.
This is one order of magnitude higher than the best previous experiment32. It is noticeable that
such a value competes with the efficiency of state-of-the-art thermoelectrics44,45. It is interesting
to estimate the contribution of the near field to this power increase. By matching the view factor
computation as a function of distance to the largest distances in the curve (dashed-dotted black
curve in Fig. 3b), we are able to single out a stronger increase of the electrical output power in
the sub-3 µm region than what standard far-field theory predicts (yellow region). Electrical
output power 4.8 times larger than the far-field one is found.
The photovoltaic conversion efficiency of the near-field flux can be determined from our
experimental data as a function of emitter-cell distance in the last 3 µm. By dividing the
electrical power by the near-field radiation power, we find a near-field conversion efficiency of
the order of 5 % for distances between 0.2 and 2.7 µm (uncertainty is linked to the near-field
radiation data), with a strong increase for the closest distances to a value above 10%. The
distance is measured by the vertical piezoelectric actuator, with an uncertainty close to the
contact up to around 80 nm (see Ext. Data Fig. 6 and SI).
7
Figure 4 Influence of cell diameter, substrate thickness and p-doping level on the maximum
electrical power for the smallest emitter-cell distances. The dark and light parts are respectively the
far-field and near-field contributions. a) Influence of the cell active area diameter. The electrical power
decreases with the cell diameter because less illumination is collected, but the electrical power density
increases. b) (left) Influence of the substrate thickness. The maximum electrical power density is
observed for the device with the thinnest substrate. (right) Influence of the p-doping level. The maximum
power density increases with the doping level.
Parameters that can enhance radiative heat-to-electricity conversion are now analyzed. First,
the effect of the diameter of the active area of the cell is reported in Fig. 4a. As a larger share
of the exchanged power comes from the near field for the smallest cell (dark orange region), it
performs the best, being able to generate about 2 kW.m-2, which is three orders of magnitude
higher than previous reports32 -- 34. Due to the high mobility of electrons in InSb, critical electrical
losses are unlikely. However, parasitic absorption losses are expected for photon energies above
the bandgap. Note that the near-field contribution in the sphere-plane geometry is mostly related
to optical modes that are evanescent in the gap but propagate in the substrate ("frustrated
modes", see SI and Ext. Data Figs. 4-5). By reducing the substrate thickness from 500 µm to
200 µm, part of the parasitic absorption is mitigated and a second passage of these photons
across the junction is possible after reflection by the bottom gold contact layer which acts as a
mirror46. A clear enhancement is observed in the experiment (Fig. 4b (left)). In addition, the
doping level of the p-region of the junction was also shown to be a key parameter for the cell
performances42. An impressive 4-fold enhancement of the electrical power is observed when
increasing the dopant concentration from 1017 to 1018 cm-3, leading to a record 7.5 kW.m-2 value
8
for the tested micron-sized cells, despite the moderate temperature of the emitter (432°C). Let
us analyze the results in terms of efficiency. Near-field efficiency for the cell with a 500 µm-
thick substrate and a 1017 cm-3 p-doping concentration was directly measured to be 5.8% (lower
than that reported in Fig. 3 due to lower temperature in Fig. 4). By thinning down the substrate
of the cell, we find that the near-field efficiency, also measured directly from the output power,
reaches 14.1 % (see also Ext. Data Tab. 1 and SI). This high value is explained first by the
increase in power conversion for photon energies above the bandgap and second by the
reflection of near-field thermal photons below the bandgap, which reduces the parasitic
absorption that does not lead to photoconversion42,47.
In summary, we have demonstrated that near-field thermal energy can be exploited efficiently.
We have measured photovoltaic conversion of thermal photons in the near field with efficiency
beyond 10% and electrical power densities reaching 7.5 kW.m-2, close to the ~1 W.cm-2
threshold typical for powerful energy harvesting devices48 and close to the best far-field
thermophotovoltaic performances despite much lower emitter temperatures35,36. This proof of
principle for efficient near-field thermophotovoltaic conversion paves the way for fabrication
of devices in different fields. Scattering near-field energy to the far field allows collecting only
a weak fraction of it, which may lead to signal-to-noise ratio issues in thermal
nanocharacterization setups4 -- 6. The present approach highlights the possibility of using the
photocurrent as signal, converting the photons directly close to the surface with a superior
sensitivity. For energy harvesting as in solar thermophotovoltaics49 -- 51, it is also required to
upscale the sizes and to avoid cooling the cell. Our architectures are compatible with more
industrial-size implementations involving parallel flat surfaces that were undertaken recently52.
Recent strategies for designing efficient cells with narrow energy bandgap III-V materials
operating at room temperature53 can also be applied for frustrated photon modes and will be
particularly useful for harnessing energy of medium-grade heat sources. For high-grade heat
sources, we have demonstrated the possibility to maintain a temperature difference larger than
1100 K across a small distance (see Ext. Data Fig. 1 and SI Sec. 10), so larger energy gap
materials used in cells operating at room temperature could also be considered with a careful
design for the near field following the highlighted paths. Finally, we remind that further
enhancement of the performances is predicted for materials with polariton resonances above
the energy bandgap26,27.
9
Main references
1.
Shchegrov, A. V, Joulain, K., Carminati, R. & Greffet, J.-J. Near-Field Spectral Effects
due to Electromagnetic Surface Excitations. Phys. Rev. Lett. 85, 1548 -- 1551 (2000).
2.
Joulain, K., Mulet, J.-P., Marquier, F., Carminati, R. & Greffet, J.-J. Surface
electromagnetic waves thermally excited: Radiative heat transfer, coherence properties
and Casimir forces revisited in the near field. Surf. Sci. Rep. 57, 59 -- 112 (2005).
3.
Volokitin, A. I. & Persson, B. N. J. Near-field radiative heat transfer and noncontact
friction. Rev. Mod. Phys. 79, 1291 -- 1329 (2007).
4.
De Wilde, Y. et al. Thermal radiation scanning tunnelling microscopy. Nature 444, 740 --
3 (2006).
5.
Jones, A. C. & Raschke, M. B. Thermal infrared near-field spectroscopy. Nano Lett. 12,
1475 -- 81 (2012).
6. Weng, Q. et al. Imaging of nonlocal hot-electron energy dissipation via shot noise.
Science 360, 775 -- 778 (2018).
7.
Tervo, E., Bagherisereshki, E. & Zhang, Z. Near-field radiative thermoelectric energy
converters: a review. Front. Energy 12, 5 -- 21 (2018).
8.
St-Gelais, R., Bhatt, G. R., Zhu, L., Fan, S. & Lipson, M. Hot Carrier-Based Near-Field
Thermophotovoltaic Energy Conversion. ACS Nano 11, 3001 -- 3009 (2017).
9.
Polder, D. & Van Hove, M. Theory of radiative heat transfer between closely spaced
bodies. Phys. Rev. B 4, 3303 -- 3314 (1971).
10.
Pendry, J. B. Radiative exchange of heat between nanostructures. J. Phys. Condens.
Matter 11, 6621 (1999).
11. Liu, X., Wang, L. & Zhang, Z. M. Near-field thermal radiation: Recent progress and
outlook. Nanoscale Microscale Thermophys. Eng. 19, 98 (2015).
12. Ben-Abdallah, P. & Biehs, S. A. Thermotronics: Towards Nanocircuits to Manage
Radiative Heat Flux. Zeitschrift fur Naturforsch. - Sect. A J. Phys. Sci. 72, 151 -- 162
(2017).
13.
Fiorino, A. et al. A Thermal Diode Based on Nanoscale Thermal Radiation. ACS Nano
12, 5774 -- 5779 (2018).
14. Zhu, L. et al. Near-field photonic cooling through control of the chemical potential of
photons. Nature 566, 239 -- 244 (2019).
15. Cakiroglu, D. et al.
Indium antimonide photovoltaic cells
for near-field
thermophotovoltaics. Sol. Energy Mater. Sol. Cells 203, 110190 (2019).
16. Basu, S., Zhang, Z. & Fu, C. Review of near‐field thermal radiation and its application
10
to energy conversion. Int. J. Energy Res. 33, 1203 -- 1232 (2009).
17. Okanimba Tedah, I. A., Maculewicz, F., Wolf, D. E. & Schmechel, R. Thermoelectrics
versus thermophotovoltaics: Two approaches to convert heat fluxes into electricity. J.
Phys. D. Appl. Phys. 52, 275501 (2019).
18. Narayanaswamy, A., Shen, S. & Chen, G. Near-field radiative heat transfer between a
sphere and a substrate. Phys. Rev. B - Condens. Matter Mater. Phys. 78, 115303 (2008).
19. Rousseau, E. et al. Radiative heat transfer at the nanoscale. Nat. Photonics 3, 514 -- 517
(2009).
20. Lim, M., Lee, S. S. & Lee, B. J. Near-field thermal radiation between doped silicon plates
at nanoscale gaps. Phys. Rev. B - Condens. Matter Mater. Phys. 91, 195136 (2015).
21. Bernardi, M. P., Milovich, D. & Francoeur, M. Radiative heat transfer exceeding the
blackbody limit between macroscale planar surfaces separated by a nanosize vacuum
gap. Nat. Commun. 7, 12900 (2016).
22. Watjen, J. I., Zhao, B. & Zhang, Z. M. Near-field radiative heat transfer between doped-
Si parallel plates separated by a spacing down to 200 nm. Appl. Phys. Lett. 109, (2016).
23.
Pan, J. L., Choy, H. K. H. & Fonstad, C. G. Very large radiative transfer over small
distances from a black body for thermophotovoltaic applications. IEEE Trans. Electron
Devices 47, 241 -- 249 (2000).
24. Whale, M. D. & Cravalho, E. G. Modeling and Performance of Microscale
Thermophotovoltaic Energy Conversion Devices. IEEE Trans. Energy Convers. 17,
130 -- 142 (2002).
25. Mulet, J.-P., Joulain, K., Carminati, R. & Greffet, J.-J. Enhanced radiative heat transfer
at nanometric distances. Microscale Thermophys. Eng. 6, 209 -- 222 (2002).
26. Laroche, M., Carminati, R. & Greffet, J. J. Near-field thermophotovoltaic energy
conversion. J. Appl. Phys. 100, 063704 (2006).
27. Narayanaswamy, A. & Chen, G. Surface modes for near field thermophotovoltaics. Appl.
Phys. Lett. 82, 3544 -- 3546 (2003).
28.
Francoeur, M., Vaillon, R. & Mengüç, M. P. Thermal impacts on the performance of
nanoscale-gap thermophotovoltaic power generators. IEEE Trans. Energy Convers. 26,
(2011).
29.
Jin, S., Lim, M., Lee, S. S. & Lee, B. J. Hyperbolic metamaterial-based near-field
thermophotovoltaic system for hundreds of nanometer vacuum gap. Opt. Express 24,
A635 (2016).
30. Ben-Abdallah, P. & Biehs, S. A. Harvesting the Electromagnetic Energy Confined Close
11
to a Hot Body. Zeitschrift fur Naturforsch. - Sect. A J. Phys. Sci. 74, 689 -- 696 (2019).
31. DiMatteo, R. S. et al. Enhanced photogeneration of carriers in a semiconductor via
coupling across a nonisothermal nanoscale vacuum gap. Appl. Phys. Lett. 79, 1894 -- 1896
(2001).
32.
Inoue, T. et al. One-Chip Near-Field Thermophotovoltaic Device Integrating a Thin-
Film Thermal Emitter and Photovoltaic Cell. Nano Lett. 19, 3948 -- 3952 (2019).
33.
Fiorino, A. et al. Nanogap near-field thermophotovoltaics. Nat. Nanotechnol. 13, 806 --
811 (2018).
34. Bhatt, G. R. et al. Integrated near-field thermo-photovoltaics for on-demand heat
recycling. arXiv: 1911.11137 (2019).
35. Omair, Z. et al. Ultraefficient thermophotovoltaic power conversion by band-edge
spectral filtering. Proc. Natl. Acad. Sci. U. S. A. 116, 15356 -- 15361 (2019).
36. Wernsman, B. et al. Greater Than 20% Radiant Heat Conversion Efficiency of a
Thermophotovoltaic Radiator/Module System Using Reflective Spectral Control. IEEE
Trans. Electron Devices 51, 512 -- 515 (2004).
37. He, R. et al. Achieving high power factor and output power density in p-type half-
Heuslers Nb1-xTixFeSb. Proc. Natl. Acad. Sci. U.S.A. 113, 13576 -- 13581 (2016).
38.
Shen, S., Narayanaswamy, A. & Chen, G. Surface phonon polaritons mediated energy
transfer between nanoscale gaps. Nano Lett. 9, 2909 -- 2913 (2009).
39.
Song, B. et al. Enhancement of near-field radiative heat transfer using polar dielectric
thin films. Nat. Nanotechnol. 10, 253 -- 258 (2015).
40.
Shi, J., Liu, B., Li, P., Ng, L. Y. & Shen, S. Near-field energy extraction with hyperbolic
metamaterials. Nano Lett. 15, 1217 -- 1221 (2015).
41. Menges, F. et al. Thermal radiative near field transport between vanadium dioxide and
silicon oxide across the metal insulator transition. Appl. Phys. Lett. 108, 171904 (2016).
42. Vaillon, R. et al. Micron-sized liquid nitrogen-cooled indium antimonide photovoltaic
cell for near-field thermophotovoltaics. Opt. Express 27, A11 (2019).
43. Gomès, S., Assy, A. & Chapuis, P. O. Scanning thermal microscopy: A review. Phys.
Status Solidi a - Appl. Mater. Sci. 212, 477 -- 494 (2015).
44. Datas, A. & Martí, A. Thermophotovoltaic energy in space applications: Review and
future potential. Sol. Energy Mater. Sol. Cells 161, 285 -- 296 (2017).
45. He, J. & Tritt, T. M. Advances in thermoelectric materials research: Looking back and
moving forward. Science 357, eeak9997 (2017).
46. Bright, T. J., Wang, L. P. & Zhang, Z. M. Performance of near-field thermophotovoltaic
12
cells enhanced with a backside reflector. J. Heat Transfer 136, (2014).
47. Chen, K., Santhanam, P. & Fan, S. Suppressing sub-bandgap phonon-polariton heat
transfer in near-field thermophotovoltaic devices for waste heat recovery. Appl. Phys.
Lett. 107, 091106 (2015).
48.
Shirvanimoghaddam, M. et al. Towards a Green and Self-Powered Internet of Things
Using Piezoelectric Energy Harvesting. IEEE Access 7, 94533 -- 94556 (2019).
49. Bierman, D. M. et al. Enhanced photovoltaic energy conversion using thermally based
spectral shaping. Nat. Energy 1, 16068 (2016).
50. Ungaro, C., Gray, S. K. & Gupta, M. C. Solar thermophotovoltaic system using
nanostructures. Opt. Express 23, A1149 -- A1156 (2015).
51.
Shimizu, M., Kohiyama, A. & Yugami, H. High-efficiency solar-thermophotovoltaic
system equipped with a monolithic planar selective absorber/emitter. J. Photonics
Energy 5, 053099 (2015).
52. DeSutter, J., Tang, L. & Francoeur, M. A near-field radiative heat transfer device. Nat.
Nanotechnol. 14, 751 -- 755 (2019).
53. Huang, W. et al. Enhanced collection efficiencies and performance of interband cascade
structures for narrow bandgap semiconductor thermophotovoltaic devices. J. Appl. Phys.
124, 023101 (2018).
Methods
Fabrication of the emitter
The emitter (Fig. 1b) consists of a commercial doped-Si scanning thermal microscopy probe
(VITA-HE-NANOTA-200 from Bruker) on which a graphite sphere was glued using a 600 nm
particle size Al2O3-based ceramic adhesive (RESBOND 989F) which can withstand
temperatures up to 1920 K. The sphere of diameter 37.5 µm (Carbon powder from Goodfellow)
was heated by the doped-Si tip of the SThM probe. Graphite was selected from an analysis of
various materials (see Ext. Data. Fig. 2a,b). The electrical resistance of the probe and its
dependence on temperature allowed simultaneous temperature measurements and heating by
Joule effect of the sphere up to approximately 1200 K. Above this temperature and until the
melting temperature of silicon (1687 K), the electrical resistance variations are insufficient to
perform accurate temperature measurements, but the emitter can still be heated (see Ext. Data
Fig. 1a).
13
Calibration curve of the emitter temperature
Each emitter had to be calibrated in temperature by measuring the electrical resistance
dependence on temperature of the scanning thermal microscopy probe to which it was glued.
First, the whole system composed of the SThM probe and its half-moon shaped holder was put
in an oven typically used for thermocouples calibration (Fluke 9144). Because of the holder,
the electrical resistance measurement could only be performed from room temperature to 413 K
(140 °C). At higher temperatures, the glue and other components of the holder started to
deteriorate and the probe could be damaged. To retrieve the R(T) curve at these higher
temperatures, a measurement of the resistance according to electrical power was performed. A
current source provided an electrical current to the probe up to 11 mA while the probe voltage
was measured so the resistance could be calculated. This kind of doped-Si probe has a
characteristic R(T) curve with a maximum resistance Rmax for a given temperature TRmax which
depends on the doping level of the low-doped part of the probe, located near the tip (Ext. Data
Fig. 1). At this point, the thermally generated carrier concentration becomes higher than the
doping level and the temperature evolution becomes linear with electrical power54. For the
resistance measurement at T < TRmax made in the oven, the data could be well fitted with a
quadratic model up to a few kelvins before TRmax. For T > TRmax the temperature was calculated
from the electrical power by multiplying with a fitting factor so both temperature scales
matched near TRmax. Put together, the data from the direct measurement in the oven and the
R(P) electrical measurement gave the R(T) curve for the entire temperature range (see Ext..
Data Fig. 1). The temperature coefficient of the electrical resistivity (TCR) in K-1 can be
calculated as α =
1
𝑑𝑅
𝑅
𝑑𝑇
and is useful to determine the temperature TS for which the sensitivity
is maximum. For these SThM probes, TS is around 732 K and corresponds to the maximum
value (in absolute values) of , where a small change in temperature induces a large change in
electrical resistance that can be measured more easily. It can also be noticed that beyond around
1200 K the sensitivity becomes very low and it is therefore complicated to accurately determine
the temperature based on the resistance measurement.
Temperature measurement of the emitter and near-field thermal conductance
Measurement of the emitter temperature T is needed to determine the radiative heat transfer
with the cell characterized by the near-field thermal conductance GNF. The principle is to
14
measure the electrical resistance of the SThM probe R, thermalized with the emitter, and use
calibration data to infer its temperature (see above). A Keithley 6221 signal generator was used
to provide DC current I to the SThM probe, up to 8 mA depending on the required temperature.
A Wheatstone bridge circuit with a 100x voltage amplification was used to accurately measure
the electrical resistance. The bridge was balanced with the emitter out of contact, for a distance
d slightly smaller than 5 m, by adjusting the current supply until the electrical resistance of
the probe matches the one imposed by a programmable resistance (MEATEST M632)
corresponding to the targeted temperature. During the approach, the imbalance of the bridge
was measured according to the displacement of the emitter d, and then the electrical resistance
R and temperature T were calculated. A Keithley 2182A nanovoltmeter was used to have a
direct view of the imbalance of the bridge, while this signal was recorded by means of a NI
cDAQ-9178 data acquisition system with a NI 9239 voltage measurement unit at a 2 kHz
acquisition rate. For each experiment, up to 100 identical approaches were performed and the
resulting curves were averaged to obtain a better signal-to-noise ratio. A sliding average was
also applied to remove the 50 Hz electrical noise.
In order to determine the emitter-cell thermal conductance during an approach, we measured
the electrical resistance variation of the emitter and used the calibration data to calculate its
temperature. The resistance measurements were first averaged before calculating the
temperature and the conductance (SI and Ext. Data Fig. 1b). Variations of total thermal
conductance 𝐺𝑡𝑜𝑡 =
𝑃
𝜃
, where P is the electrical power fed to the probe and 𝜃 = 𝑇 − 𝑇𝑎𝑚𝑏 the
average temperature increase, are small. Thus the radiative thermal conductance in the near
field 𝐺𝑁𝐹 , being exactly the variation Δ𝐺𝑡𝑜𝑡(𝑑) which depends on the emitter-cell distance d,
can be calculated from a logarithmic derivation:
Δ𝐺𝑡𝑜𝑡
𝐺𝑡𝑜𝑡
=
Δ𝑃
𝑃
−
Δ𝜃
𝜃
=
Δ𝑅
𝑅
+
2Δ𝐼
𝐼
−
Δ𝜃
𝜃
. With the
reference temperature taken at the largest distance (Tref), the temperature coefficient (), the
emitter current and the reference current at the largest distance (respectively I and Iref ), the
following equation can be derived:
(1)
where Gtot is measured at the largest distance. Depending on the working temperature, and
(I-Iref) are respectively positive/negative for T < TRmax, and negative/positive for T > TRmax
because the temperature drop during an approach can result in a decrease or an increase of
electrical resistance. In Ext. Data Fig. 1b, the temperature drop close to contact is seen as a
15
resistance increase due to the working temperature (≈ 732 K), which is higher than TRmax.
Finally, it must be noted that GNF is equal to 0 at the largest distances because it represents only
the evanescent wave contribution to the radiative heat transfer, which is considered negligible
at such distances.
Since radiative measurements can be noisy, they require averaging over many approach curves
(see also SI and Ext. Data Fig. 1). At room temperature approximately 100 approach curves
were considered for averaging. At low temperature (Fig. 3a), seven curves were used and
consequently the signal-to-noise ratio became small for distances larger than 3 µm (shaded
region). The value of the background level was estimated by averaging the measured flux over
a distance range of 0.3 µm around the 4 µm position. This sets the main uncertainty on the exact
flux determination.
Design and fabrication of the photovoltaic cells
The photovoltaic cells (Fig. 1c) were specifically designed and fabricated to demonstrate the
near-field enhancement of the electrical power15,42. For the semiconductor material, we chose
indium antimonide (InSb) for its low bandgap energy of 0.23 eV (g = 5.3 µm) at 77 K, thus
allowing the conversion of low-energy infrared radiation (Fig. 1d).
We first analyzed numerically the doping levels and layer thicknesses of the InSb p-n junction
to be able to obtain the largest electrical power generation from the cell. Second, we fabricated
the optimum cells and measured their photovoltaic (PV) response in the far field using different
emitter temperatures. Cells were made with different active area diameters of 20, 40, 80 and
160 µm in order to study the influence of the illuminated area on electrical power generation15.
In order to investigate the impact of cell parameters on conversion performances, two p-doping
levels of 1017 and 1018 cm-3, and two substrate thicknesses of 200 and 500 µm, were selected.
The cells were glued with silver paste on a gold-coated chip carrier and connection from the
front and back contacts of the cells to the chip carrier was made using wire bonding.
In the present work, we added one step to the previously reported fabrication process15.
Passivation is required to avoid accumulation of majority carriers at the side walls of the mesa
and the associated current that lowers performances of the junction. The extra step consisted in
passivating 20 µm of the circular mesa structure (10 µm from inside of circle and 10 µm from
the etched area) using UV-insulated then cured AZ1518 resist. By means of a plasma-enhanced
chemical vapor deposition (PECVD) method at 200 °C, all surfaces, except the cell active area,
were then covered with a dielectric material such as SiO2 or Si3N4, which allows preventing
16
shortcuts between contacts and makes it possible to wire-bond. Top views of the resulting cells
are shown in Ext. Data Fig. 7.
Setup of the experiment
The experimental setup (Fig. 1a) was assembled in a vacuum chamber where the pressure was
decreased to P ≈ 10-6 mbar, using only an ion pump at that threshold to avoid vibrations. The
chip carrier of the cell was glued with silver paste on the cold finger of a liquid-helium cryostat.
The cell position was fixed in the experiment. The emitter was initially placed at approximately
2 mm above the surface of the cell and could be moved in x-, y- and z-axis directions by a
custom piezoelectric positioning system having ranges of 3, 3 and 2.5 mm, respectively.
Positioning of the emitter above the centre of the active area of the cell was done in three steps.
First, the position of the emitter was controlled using piezoelectric positioners and a microscope
camera (Dino-Lite Edge 3.0 AM73115MTF) with a long working distance of 15 cm for a 40x
magnification located outside the vacuum chamber. The emitter was brought in contact with
the chip at an arbitrary location, then retracted at a safe distance of around 100 µm and moved
over the cell (see Ext. Data Fig. 8 for a top view). In a second step, the emitter was brought in
contact with the cell and retracted at a distance d smaller than 5 µm. There the fine positioning
(step 3) was performed by measuring the short-circuit current (ISC) of the cell according to the
emitter displacement along the x and y axes, with the emitter heated at a temperature larger than
730 K. Once the maximum current is reached along one axis (see Ext. Data Fig. 8c), the
maximum current position along the other axis corresponds to the centre of the active area. It
is reminded that parallelization of the emitter and cell surfaces was not required due to the
sphere-plane configuration, chosen on purpose in place of the plane-plane configuration.
Detection of contact
The mechanical contact between the emitter and the cell was detected simultaneously by two
different methods. The temperature of the emitter was measured continuously according to the
piezo displacement. At contact, there is a sudden drop of the temperature due to the contribution
of heat conduction which becomes largely dominant over thermal radiation for the heat transfer
between the emitter and the cell. The other method was based on the measurement of the cell
current according to the displacement of the emitter. When the emitter is in contact, the
performances of the cell harshly decrease due to the temperature increase15 caused by the hot
emitter. A significantly-lower photogenerated current is then measured. These two methods
17
applied simultaneously were useful to check that the contact actually occurs in the active area
of the cell. When some contact took place with the gold surface sides, the temperature drop of
the emitter was still observed but the decrease of the photogenerated current was absent because
the heated area does not contribute to the photovoltaic effect. More details on the analysis of
the emitter-cell distance is provided in SI and Ext. Data Fig. 6.
I-V characteristics measurement according to emitter-cell distance
The current-voltage (I-V) characteristics of the photovoltaic cells were measured using a
Keithley 2400 source measurement unit. A voltage scan was applied to the cell and the resulting
current was measured (Fig. 2). For the dark configuration, we used a cooled radiative shield
that was mounted around the sample to prevent the active area of the cells from receiving
ambient temperature radiation. In order to get the electrical power at the maximum power point
according to the emitter-cell distance, I-V curves must be measured as a function of distance.
For stability purposes, it was chosen to perform continuous approaches of the emitter while
measuring the current produced by the cell for a fixed voltage, and then to repeat this procedure
for different voltages. After performing approaches at voltages ranging from 0 to around 140
mV, we obtained a series of current-distance curves at different cell voltages that could be used
to plot the current-voltage curves at different distances. Then we calculated the photovoltaic
power at each distance and determined the maximum value so as to plot the maximum electrical
power according to distance. We also used another method based on the superposition principle,
where the I-V curve of the cell is considered to keep the same shape but is shifted in short-
circuit current depending on the illumination level. With this method, we measured a reference
I-V curve under ambient illumination, then we measured only the short-circuit current ISC
(current at Vcell = 0) according to distance. The reference I-V curve was then shifted by plotting
the (I-ISC)-V curves at each distance. Then the maximum electrical power could be calculated.
This method was much faster to perform experimentally because it required approach curves at
only one cell voltage, but in this case the maximum power was deduced from the shifted
reference curves and not directly measured. Validation of the superposition principle is
provided in Ext. Data Fig. 9.
Near-field conversion efficiency measurement
The setup allows the simultaneous measurement of the electrical power generated by the cell
and the radiative power exchanged between the emitter and the cell in the near field. In order
18
to identify only the near-field contribution to electrical power generation, the theoretical
evolution of power according to distance was calculated using the analytical expression of the
sphere-disc view factor. We used the electrical power measured at the largest distance in an
approach curve similar to Fig. 3b as the fitting parameter, assuming that the generated power
was only due to the far-field contribution. The difference between the power calculated with
the view factor and the total measured power corresponds to the near-field electrical power.
Note that our definition of the far-field contribution is that related to the view-factor theory, and
that it can depart from an exact calculation of the contribution of the propagative waves in the
selected geometry by means of fluctuational electrodynamics. In case of a slight error on the
identification of the far-field contribution from the experimental data (Fig. 3), our estimation is
a lower bound to the near-field contribution. For the different distances, the ratio between the
near-field electrical power and the near-field exchanged radiative power gave the conversion
efficiency for the near-field contribution only.
Methods reference
54.
Spieser, M., Rawlings, C., Lörtscher, E., Duerig, U. & Knoll, A. W. Comprehensive
modeling of Joule heated cantilever probes. J. Appl. Phys. 121, (2017).
Acknowledgments
Financial support by the French National Research Agency (ANR) under grant No. ANR-16-
CE05-0013 and partial funding by the French "Investment for the Future" program (EquipEx
EXTRA ANR-11-EQPX-0016 and IDEXLYON ANR-16-IDEX-0005) and by the Occitanie
region are acknowledged. C.L. and P.O.C. thank C. Ducat, N. Pouchot and A. Buthod for help
in the design of the setup, P. Mangel and D. Renahy for experimental support, and S. Gomes.
P.O.C thanks S. Volz for equipment handover.
Author contributions
R.V., P-O.C, J-P.P., E.T. and T.T. conceived and supervised the work. J-P.P did the MBE
growth of the InSb material and D.C. fabricated the PV cells. C.L. fabricated the emitter and
performed the experiments. C.L. and R.V. performed the simulations. The manuscript was
written by C.L., P-O.C and R.V. with comments and inputs from all authors.
The authors declare no financial competing interest.
19
Extended Data Figure 1 Determination of the near-field radiative thermal conductance and
impact of emitter temperature. (a) Calibration of the emitter temperature: (Step 1) Electrical resistance
and (Step 2) temperature coefficient of the electrical resistivity ( =
1
𝑅
𝑑𝑅
𝑑𝑇
) of the emitter according to
the oven temperature. (b) Application of the calibration for determining the thermal conductance: (Step
1) Electrical resistance, (Step 2) temperature and (Step 3) near-field radiative thermal conductance as a
function of distance to the contact. The emitter-sample power exchanged is also indicated. The electrical
resistance measurements were averaged over 16 curves. (c) Impact of the emitter temperature on the
near-field thermophotovoltaic (TPV) power output, measured for a photovoltaic cell with diameter 20-
µm at d ≈ 5 µm.
20
Extended Data Figure 2 Near-field radiative power exchanged for different emitters in the
vicinity of semi-infinite planar bodies. (a) Spectral flux between two semi-infinite planar media.
Calculations for different emitters at 732 K and an InSb sample at 77 K with an emitter-sample distance
d = 10 nm. (b) Integrated flux for radiation wavelengths useful for photocurrent generation ( < gap),
as a function of distance. See SI Sec. 1. (c) Near-field thermal conductance as a function of distance, for
two spherical graphite emitter temperatures and an InSb sample at room temperature. Inset: logarithmic
plot. Plain lines are experimental data and dashed lines are predictions from the Proximity
Approximation. The schematic reminds the principle of the Proximity Approximation (use of the infinite
parallel-surfaces heat transfer coefficient and integration as a function of height, see SI Sec. 2). (d)
Power divided by the projection over the surface area for two different geometries (planar or spherical
emitter). The graphite emitter temperature is 732 K and the InSb sample is at 77 K. The fluxes for the
sphere-planar surface geometry are computed by means of the Proximity Approximation.
21
Extended Data Figure 3 Calculations of the contribution of the propagative waves to the radiative
power according to distance (see SI Sec. 4). (a) Schematic of the Monte Carlo computation of the
propagative photon contribution to the transfer between the hot sphere and the cold receiver (see SI).
(b) Local radiative power density deposited by a 40 µm graphite sphere at 732 K on a flat InSb semi-
infinite cylinder of diameter 160 µm at 77 K, for an emitter-cell distance equal to 5 µm. The cylinder
represents the cell. (c) Radiative power density as a function of cell radius. (d) Evolution of the radiative
power calculated using the view factor as a function of distance, compared with the Monte Carlo
simulations. (e) Spectral and directional specular reflectance of an InSb cell extended infinitely in the
lateral directions and consisting of four layers (p-doped, Na = 1017 cm-3, tp = 0.5 m / n-doped, Nd = 1015
cm-3, tn = 2.5 m / n-doped substrate, Nd,sub = 4 1017 cm-3, tsub = 500 m / gold, tbrl = 200 nm). These
data, computed by electromagnetic means, are fed in the Monte Carlo computation of the far-field
contribution.
22
Extended Data Figure 4 Contribution of the frustrated and surface modes to the evanescent
radiative heat transfer. The emitter temperature is 732 K and the InSb semi-infinite planar medium
receiver is at 77 K (a) Real (blue) and imaginary (red) parts of the dielectric function of InSb at 77 K (p
doping level of 1018 cm-3). (b) Evanescent component of the spectral flux between two semi-infinite
planar media separated by a 10 nm vacuum gap. (c) Evanescent flux as a function of the distance between
two semi-infinite planar media. (d) Evanescent conductance as a function of distance, for a 37.5 µm
graphite sphere and an InSb semi-infinite planar medium. See also SI Sec. 5.
23
Extended Data Figure 5 Radiative power absorbed as a function of frequency and depth, for
sphere-cell distances of 105 nm (b,d) and 10.2 nm (c,e) and a substrate thickness of 500 𝛍m. (a)
Spectral absorption coefficient of indium antimonide for different doping levels considered for these
calculations. (b,c) Local spectral radiation power absorbed as a function of depth (z) inside the cell. (d,e)
Local radiation power absorbed as a function of depth. The results are computed with the Proximity
Approximation for both evanescent and propagative contributions, and with the following parameters:
graphite sphere of diameter 40 μm at 773 K, InSb cell at 77 K made of four layers (p-doped, Na = 1017
cm-3, tp = 0.5 m / n-doped, Nd = 1015 cm-3, tn = 2.5 m / n-doped substrate, Nd,sub = 4 1017 cm-3, tsub =
500 m / gold, tbrl = 200 nm). Surface modes decay quickly from the surface at d = 10 nm, and are not
present at 100 nm. See also SI Sec. 5.
24
Extended Data Figure 6 Parameters influencing the determination of the emitter-cell distance at
contact, i.e. the snap-in distance associated to the jump at contact (a,b), the sphere material
roughness (c), and sample vibrations (d-f) (see also SI Sec. 6). (a) Schematic of the cantilever
deflection measurement setup using a quadrant photodiode system. (b) Cantilever deflection as a
function of the displacement during an approach (blue) and a withdrawal motion (red). The emitter is
out of contact when the curve is flat and is in contact when the slope is steep. The data indicate an
uncertainty of few nanometers at maximum. (c) Topographic image of the sphere obtained by scanning
the emitter with an atomic force microscopy tip, before and after spherical shape subtraction. This allows
determining the roughness of the emitter involving an rms roughness of 5 nm and peaks of maximal
height around 30 nm. (d) Schematic of the interferometric vibration measurement setup allowing to
measure the variation of position of the cell with respect to the cantilever basis. (e) Interferometric signal
allowing to calculate the vibration amplitudes. The linear domain is represented by the shaded area. (f)
Histogram of the cold finger positions, indicating a maximal vibration amplitude of about 80 nm.
25
Extended Data Figure 7 Images and characteristics of the cells. (a) Top view of the InSb
photovoltaic cells with different active area diameters. (b) I-V curves at different temperatures. (c) I-V
curves at different temperatures with the current represented in absolute value and using a logarithmic
scale. See SI Sec. 7.
26
Extended Data Figure 8 Variation of the short-circuit current as a function of the position of the
emitter. (a) Optical microscopy image of the setup involving the chip where a thermophotovoltaic cell
(disc where there is no gold) is seen. Red box: View from top of the scanning thermal microscopy probe
on which the spherical emitter is glued, positioned over a cell. (b) Short-circuit current as a function of
the vertical displacement of the emitter and comparison with the view factor. Same data as that of the
inset of Fig. 3b, but in linear scale. (c) Short-circuit current as a function of the lateral displacement of
the emitter and comparison with the view factor. See SI Sec. 8.
27
Extended Data Figure 9 I-V curves of a cell having a 20 m active area diameter in the dark and
under different illumination levels. (a) I-V curves under illumination from different distances and
temperatures. (b) Curves shifted in current: the short-circuit current of each curve is set equal to zero.
All curves superimpose, as expected from the superposition principle (see SI Sec. 9).
28
Extended Data Table 1 Results of the near-field thermophotovoltaic experiments for different
studied configurations (see also SI Sec. 10).
29
SUPPLEMENTARY INFORMATION
Outline
1. Supplementary Methods
1. Selection of the emitter material ........................................................................................... 31
2. Sphere-plane vs. plane-plane configurations ........................................................................ 32
3. Comparison between near-field thermal radiative measurements and the proximity
approximation ...................................................................................................................... 32
4. Contribution of the propagating modes with distance .......................................................... 33
5. Numerical analysis of the near-field modes and the radiative absorption as a function of
depth .................................................................................................................................... 34
6. Estimation of the sphere-cell distance close to contact ........................................................ 35
7. Characterization of the photovoltaic cells ............................................................................. 36
8. Experimental results in the far field showing excellent agreements with the sphere-disc view
factor .................................................................................................................................... 36
9. Validation of the superposition principle .............................................................................. 37
2. Supplementary Discussion
10. Summary of the different studied configurations ................................................................. 38
Supplementary References ............................................................................................................ 39
30
1. Supplementary Methods
1. Selection of the emitter material
In order to choose an efficient emitter material that matches the optical properties of InSb, we
performed radiative heat transfer calculations between two semi-infinite planar media labelled 1
and 3 separated by a vacuum gap labeled 2. The spectral heat flux emitted by one of these media
at temperature T and absorbed by the second one is the sum of the propagative and evanescent
contributions q = qprop + qevan and can be calculated based on the electric properties, here the
complex dielectric function which depends on the angular frequency 𝜔 of the emitter and the
receiver9:
𝑞𝑒𝑣𝑎𝑛(𝜔) =
1
𝜋2
ℏ𝜔
ℏ𝜔
𝑘𝐵𝑇−1
𝑒
∞
∫
𝑘∥=𝑘0
𝑘∥𝑒−2𝐼𝑚(𝑘2⊥)𝑑 ∑
𝑖=𝑇𝐸,𝑇𝑀
𝐼𝑚(𝑟21
𝑖 𝑟23
𝑖 )𝐼𝑚(𝑟23
𝑖 )
2
𝑖 𝑒2𝑖𝑘2⊥𝑑
1−𝑟21
𝑞𝑝𝑟𝑜𝑝(𝜔) =
1
4𝜋2
ℏ𝜔
ℏ𝜔
𝑘𝐵𝑇−1
𝑒
𝑘∥<𝑘0
∫
0
𝑘∥ ∑
𝑖=𝑇𝐸,𝑇𝑀
2
𝑖
(1−𝑟21
𝑖 𝑟23
1−𝑟21
2
𝑖
)(1−𝑟23
)
,
2
𝑖 𝑒2𝑖𝑘2⊥𝑑
,
(S1a)
(S1b)
We remind that ℏ𝜔 is the energy of a photon, 𝑘𝐵is Boltzmann's constant, and the Fresnel
coefficients are defined as
𝑘2⊥−𝑘1⊥
𝑘2⊥+𝑘1⊥
,
𝑇𝐸 =
𝑟21
𝑇𝑀 =
𝑟21
𝑘2⊥𝜀1−𝑘1⊥𝜀2
𝑘2⊥𝜀1+𝑘1⊥𝜀1
,
(S2a)
(S2b)
where 𝑘∥ is the part of the wavevector 𝑘0 = 𝜔/𝑐 which is parallel to a flat interface and 𝑘𝑖⊥ =
2 − 𝑘∥. c is the velocity of light. The total net flux is the difference between that emitted by
√𝜀𝑖𝑘0
the hot body and absorbed by the cold one, and that emitted by the cold body and absorbed by the
hot one.
Different emitter materials were considered at 732 K and coupled with InSb at 77 K. The resulting
spectral fluxes are presented in Ext. Data Fig. 2a. We observed that in near-field conditions at d =
10 nm, graphite is an excellent emitter because the spectral flux exchanged with InSb is enhanced
by more than one order of magnitude compared to the blackbody limit at wavelengths useful for
photocurrent generation ( < gap). The integrated flux over such wavelengths according to
distance is also studied (Ext. Data Fig. 2b). At d > 100 nm, doped silicon is slightly better than
graphite but tends to level off at the lowest distances. As a result, graphite was chosen with the
aim of studying the sub-100 nm regime for thermophotovoltaic conversion. Ideally, a better emitter
material would be supporting surface phonon polaritons at wavelengths close to gap(InSb) at
77 K, in order to enhance by several orders of magnitude the radiative heat transfer and the
electrical power generation in the near field.
31
2. Sphere-plane vs. plane-plane configurations
We chose a sphere-plane instead of a plane-plane configuration to avoid any parallelization issue.
A drawback is that the near-field enhancement is strongly reduced since a lower fraction of the
emitter area is in the near-field zone compared to a planar one at the same distance (Ext. Data Fig.
2d). In order to estimate the radiative flux between a spherical emitter and a semi-infinite planar
medium, we used the proximity approximation (PA, also called Derjaguin approximation)55 for
computing the contribution of the near-field flux. A flux database was calculated for the case of
the plane-plane geometry from 1 nm to 100 µm for the selected temperature. The principle of the
PA is that the spherical shape is approximated as the sum of small planar elements and the sphere-
plane flux or conductance can be calculated by integrating the local plane-plane flux weighted by
the local perimeter over the half sphere (see schematic in the inset of Ext. Data Fig. 2c):
𝐺𝑝𝑙𝑎𝑛𝑒−𝑝𝑙𝑎𝑛𝑒[𝑑(𝑟), 𝑇]2𝜋𝑟𝑑𝑟
,
(S3)
2
− 𝑟2 . We calculated the radiative heat flux between a
𝑅𝑠𝑝ℎ𝑒𝑟𝑒
𝐺𝑠𝑝ℎ𝑒𝑟𝑒−𝑝𝑙𝑎𝑛𝑒(𝑑, 𝑇) = ∫
0
where 𝑑(𝑟) = 𝑑 + 𝑅𝑠𝑝ℎ𝑒𝑟𝑒 − √𝑅𝑠𝑝ℎ𝑒𝑟𝑒
graphite sphere with diameter 40 µm at 732 K and an InSb plane at 77 K and compared it with a
planar emitter having the same surface area (Ext. Data Fig. 2d). As expected, the enhancement due
to the contribution of the evanescent waves at low distances is much weaker with the spherical
emitter (more than one order of magnitude). With a smaller sphere, a larger fraction of the area is
in the near-field zone, so the exchanged power value varies more strongly with distance. However,
the smaller the sphere, the smaller the exchanged power to be measured. Thus we chose a diameter
of 40 µm as a good compromise between near-field enhancement and power detectability.
Consequently, with the sphere-plane geometry, a weaker enhancement of the TPV power in the
near field is expected in comparison with the plane-plane geometry.
3. Comparison between near-field
thermal radiative measurements and
the proximity
approximation
Before making measurements with the cells at low temperature, we validated our experimental
setup by measuring the near-field thermal radiation exchange at room temperature between
different emitters and substrates. Ext. Data Fig. 2c reports on measurements between a graphite
emitter and a flat InSb substrate. In order to obtain a better signal-to-noise ratio, about 100
approach curves were accumulated and averaged in the present cases. The contact point shifts
between the approaches due to some long drift and is therefore tracked before averaging the data
by searching for the distance where the electrical resistance varies abruptly. The effect of the
emitter temperature can be observed: the near-field radiative conductance increases by half a
decade when the emitter temperature is varied from 456 K to 744 K.
A comparison with the prediction of the Proximity Approximation (Eq. (S3)) is also shown in Ext.
Data Fig. 2c. It is observed that there are some little differences, ascribed to the fact that the
Proximity Approximation is more accurate at small distances than at micrometer-scale ones. At
very small distances (in the sub-200 nm regime), where the position is determined less accurately
(see Sec. 6), a lower slope is observed in the logarithmic plot for both the experimental data and
the predictions.
32
4. Contribution of the propagating modes with distance
The far-field radiative flux was estimated based on two different approaches. First, the
macroscopic theory of view factors was used by assuming diffuse isotropic surfaces. The analytic
expression of the sphere-disc view factor depending on the cell and sphere radii Rdisc and Rsphere,
and the sphere-cell distance d, reads as follows56:
𝐹𝑠→𝑑 =
1 −
1
2
(
1
√1 + (
𝑅𝑑𝑖𝑠𝑐
𝑑 + 𝑅𝑠𝑝ℎ𝑒𝑟𝑒
2
)
)
(S4)
Due to the flatness of the surfaces, it could be argued that the surfaces are specular and not diffuse.
A second method was then implemented, using a Monte Carlo ray tracing numerical approach (see
principle on Ext. Data Fig. 3a). For precise calculations, we took into account the properties of
graphite and the materials constituting the InSb cell, such as emissivity and reflectance, as a
function of emission angle. In this method, rays are launched from the emitter into all directions.
A fraction of the total power emitted by the sphere is attributed to each ray depending on the
spectral distribution of the radiative flux at the emitter temperature. Then calculations are made to
check if the rays coming from the sphere are crossing the surface of the cell. If so, the trajectory is
extended by considering the angle of incidence, and the power lost by the ray (absorbed by the
cell) is calculated using the reflectivity of InSb. Then we check if the extended trajectory crosses
the surface of the sphere. Using the same principle as before, the power loss and the reflected
trajectory are calculated. In this method, we considered up to 3 reflections on the sphere for the
same ray, because it was estimated that the number of remaining rays and their energy after such
a number of reflections were negligible. The results of the Monte Carlo method which are shown
are for a spherical emitter with diameter of 40 µm and a finite disc with different diameters as the
receiver. The total flux absorbed by the disc, which is representative of the cell, is provided by a
map of the power density (Ext. Data Fig. 3b) and a power density profile integrated as a function
of cell radius (Ext. Data Fig. 3c). We observe in the figures that the flux decays for radii larger
than 20 µm, so it is not necessary to use cells with a too large radius as most of the surface would
not be much illuminated by the sphere. We could also calculate the evolution of the radiative power
deposited on the cell as a function of distance (Ext. Data Fig. 3d). We can see that for distances
smaller than ≈ 3 µm, variations of the flux are very small so most of the measured enhanced power
will be coming from the contribution of the evanescent waves in the near field. We compared the
Monte Carlo simulations of the specular bodies considering 106 rays with the analytical expression
of the diffuse-emission view factor and found a very good agreement between the results. The
specularity of the surface is therefore not key to the computation of the exchanged power. Taking
into consideration this result and because the Monte Carlo simulations need a significant
computational time, the propagating mode contribution was calculated for a single distance using
the Monte-Carlo method. The flux at other distances was deduced from both this value and the
analytical expression of the view factor. Note that the reflectance of the cell, which is a
multilayered structure, were first calculated by electromagnetic means and then included in the
Monte Carlo approach for the estimation of the propagative-wave contribution (see Ext. Data Fig.
3e).
33
5. Numerical analysis of the near-field modes and radiative absorption as a function of depth
It is key to study the contribution of the different evanescent modes to the heat flux, i.e. frustrated
and surface modes. Frustrated modes undergo total internal reflection at an interface, they are
propagative in the medium and evanescent in vacuum. Surface modes are evanescent both in
vacuum and in the medium. The dielectric function of InSb (Ext. Data Fig. 4a) allows for resonant
surface waves (phonon-polaritons) around the wavelength = 55 µm, which is by far larger than
the wavelength corresponding to the bandgap of InSb at 77 K (5.3 µm) and Wien (3.96 µm at 732
K).
We consider a semi-infinite medium of InSb as the receiver. We first analyze the plane-plane
configuration. In Ext. Data Fig. 4b, computed for d = 10 nm (a very small distance with respect to
possible applications), we observe a peak in the spectral flux around 55 µm due to the phonon-
polariton of InSb. The distance dependence shows that the evanescent component of the radiative
heat transfer is dominated by the frustrated modes for approximately d > 40 nm (Ext. Data Fig.
4c). For the sphere-plane configuration (Ext. Data Fig. 4d), this translates into a domination of the
frustrated modes for distances d > 10 nm. Knowing the uncertainty on distance close to the contact
(see Sec. 6), this study indicates that, in our case, the near-field radiative heat transfer is fully
dominated by the frustrated modes. The surface modes do not contribute to the electrical power
generation by the cell in our experiment.
Let us now move to actual geometries and materials. The dependence of the dielectric function on
doping concentration is reminded in Ext. Data Fig. 5a. The absorption coefficient of InSb is
calculated4 at 77 K for the p-region (Na = 1017 cm-3) in the standard configuration used in the core
paper, the n-region (Nd = 1015 cm-3), and the n-doped substrate (Nd,sub = 4 1017 cm-3). Some
differences can be seen below the bandgap, where the doping has a strong influence. The
penetration depth just above the bandgap is large, and the substrate is even more transparent. The
spectral heat flux absorbed as a function of position in the cell is reported in Ext. Data Fig. 5b,c. It
is computed with the Proximity Approximation (see Secs. 2-3) and the methods described in Ref.
57. Note that in contrast to other sections, the PA is also applied here for the propagative
contribution, which is a stronger approximation but allows obtaining an estimation of the
absorption as a function of depth. At a distance of ~100 nm (Ext. Data Fig. 5b), which is close to
the minimum one achievable in our experiment (see Sec. 6), all the power absorbed in the junction
(first 3 μm from the top of the cell, see Fig. 1 for the whole architecture) is located above the
bandgap and there is no contribution of the polariton as suggested by Ext. Data Fig. 4d. Some
interference patterns are observed in the junction. The power absorbed decays slowly as a function
of depth, which calls for the introduction of a mirror as close as possible from the junction to allow
for a second passage of the photons that have not been absorbed. We remind that the electron-hole
pair generation rates is a direct translation of the absorbed power in the junction above the bandgap.
Let us now analyze smaller distances (d ~ 10 nm), which may be more difficult to implement in
practical applications. Now the contribution of surface modes, which decay quickly from the
surface, can be observed (Ext. Data Fig. 4c,d). Exploiting these modes requires positioning the np
contact region close to the top surface and induces addition constrains. While this distance regime
was not tested experimentally, our design is compatible with it.
34
6. Estimation of the sphere-cell distance close to contact
a) Effect of the attraction forces
The distance when the sphere is few nanometers close to contact can also be modified by the
attraction forces between the sphere and the surface. These forces are expected to bend the SThM
probe cantilever and thus bring the sphere into contact. In order to quantify the distance where the
snap-in occurs, we performed cantilever deflection measurements as a function of distance at room
temperature for different emitter temperatures. The deflection was measured with a photodiode
system in an NTMDT atomic force microscope (AFM) equipped with a moderate-vacuum
chamber (10-1 mbar). In this experiment, a laser was focused on the edge of a SThM probe with a
sphere glued on the tip, and the reflection of the laser was observed with a quadrant photodiode.
The cantilever deflection z could be measured by looking at the x displacement of the reflected
laser on the photodiode (Ext. Data Fig. 6a). In out-of-contact position, the deflection of the probe
is constant because there is no interaction between the probe and the sample. In the approach curve
close to contact (Ext. Data Fig. 6b), the cantilever bends slightly due to the attraction forces. We
estimated this bending to be around 2-3 nm. Thus it corresponds to the distance range not
achievable with this experimental configuration. It can also be noticed that the adhesion forces,
taking place when the probe is withdrawn, lead to a much larger deflection, greater than 20 nm.
At low distances, the effect of surface roughness can become important. We performed roughness
measurements both on the sphere and the cell using atomic force microscopy to acquire
topographic images. By using the same principle as before, the deflection of an AFM probe was
measured when the surface of the sample was scanned is xy directions. The root mean square
roughness RRMS and more importantly the maximum peak height and valley depth were determined
from these measurements. For the sphere, an additional data processing was necessary as the
surface is not flat: a spherical shape with the corresponding radius of curvature was subtracted
from the measured data to obtain a flat topographic image (Ext. Data Fig. 6c). We found that the
cells were almost perfectly flat with RRMS = 0.2 nm whereas the graphite spheres had RRMS = 5.2
nm with +29.9 nm and -28.7 nm as peak height and valley depth, respectively. These values mean
that when we considered the sphere to be in contact with the cell for d = 0, the effective distance
is most probably d ≈ 30 nm between the surface of the cell and the mean spherical shape. This
indicates that the study is restricted to distances larger than 30 nm.
b) Effect of roughness
c) Vibrations of the cryostat
In order to cool the cells down to 77 K, we used a liquid helium (LHe) cryostat with a cold finger
located inside a vacuum chamber (LHe was used for practical reasons but liquid nitrogen can also
be used). The cold finger is 15 cm long inside the vacuum chamber so the continuous flow of LHe
in the finger induces mechanical oscillations. In our case it is very important to know the amplitude
of these vibrations because it corresponds to the minimum achievable emitter-cell distance. We
used an Attocube LDM1300 interferometric module based on an IR laser with a 1310 nm
wavelength fed through an optical fiber in order to measure the amplitude of the vibrations. The
fiber was attached to the setup on the z-piezo where the emitter is usually located during the
experiments, and was placed close to the surface of the cold finger (see Ext. Data Fig. 6d). Then
we moved the fiber at a constant speed over a 5 µm range by doing a series of approach/withdrawal
motions while the periodic interferometric signal was acquired. When the fiber moves by a
35
distance corresponding to the wavelength of the laser, the optical path is modified by 2 so the
period of the interferometric signal corresponds to a /2 displacement. We can deduce the signal
for half a period corresponding to a /4 = 327.5 nm displacement and find a local linear fit for the
data (Ext. Data Fig. 6e). Knowing the displacement and the moving speed of the fiber, we can
easily establish a relation between the intensity of the signal and the displacement, and calculate
the vibrations amplitude around the average position given by the linear fit by looking at the
histogram of the positions of the cold finger (Ext. Data Fig. 6f). We measured oscillations of the
cold finger around the mean position of ± 83 nm. In the experiment performed at low temperature,
the impact of the roughness of the sphere combined with the vibrations of the cryostat induce
therefore a strong distance uncertainty in the sub-100 nm regime.
7. Characterization of the photovoltaic cells
The fabricated photovoltaic cells (see Methods) are shown in Ext. Data Fig. 7a. Because the cells
are made of InSb, a very low bandgap energy material, they need to be cooled to operate properly15.
In order to verify this requirement, we performed I-V curve measurements in the dark for different
cell temperatures from 30 K to 250 K (Ext. Data Fig. 7b). In this configuration, thermal radiation
illuminating the cell comes only from the ambient-temperature environment. Above 110 K, the
I-V characteristics are linear. They do not correspond to a diode behavior anymore but are those
of a passive resistive device. In this case the thermally-generated carrier concentration is high and
the p-n junction effect does not exist anymore. When temperature decreases the exponential shape
of the curve progressively appears. The reverse bias current rises while the forward bias current
decreases. We observe in the semi-logarithmic scale (Ext. Data Fig. 7c) that for T ≤ 90 K, the open
circuit voltage (VOC) becomes positive and keeps increasing when the cell temperature is
decreasing. The presence of a positive VOC means that the cell is generating power due to the 300 K
ambient illumination. Thus the cell needs to be cooled to work properly. For our experiments, we
chose a working temperature of 77 K as it corresponds to the commonly-used boiling point of
liquid nitrogen. In addition, the illumination level provided by the emitter is high so the generated
current is large enough to be detected at this temperature and does not require a lower temperature
cooling to reduce the dark current. Note that in the main text (Fig. 2), it can be seen that the I-V
characteristics in the dark exhibits a slightly-positive value for the short-circuit current. This is due
to a slight offset of the electronics (2-3 nA), which was not subtracted in the results presented in
the article.
The I-V characteristics indicate low series resistance in the devices. This is especially possible for
micron-sized photovoltaic cells. However, the series resistance losses raise with current, which
would become huge in large area photovoltaic cells. Smart designs of the front electrode have been
recently discussed in this respect57,58 and could help in upscaling.
8. Experimental results in the far field showing excellent agreements with the sphere-disc view
factor
The experimental setup allows the measurement of the short-circuit current of the cell for emitter-
cell distances up to 2 mm. In order to analyze the contribution of the far field to the photocurrent
36
generation, we measured the short-circuit current (ISC) of a cell having a 20 µm diameter active
area with the emitter moving from the contact up to d > 150 µm, where the current starts to level
off due to the low illumination. In this case, we used the z-piezo positioner in slip-stick mode with
80 nm steps. Then we compared the measured data to the evolution predicted by the analytic
expression of the sphere-disc view factor given in Eq. (S4). Ext Data Fig. 8b (same data as that in
the inset of Fig. 3b, different experiment than that reported in the other panels of Fig. 3) shows
that the evolution of the short-circuit current matches well the prediction of the view factor from
a few micrometers to more than 150 µm. This analysis shows that the far-field thermophotovoltaic
conversion efficiency does not depend on distance. Below 2-3 µm, the measured data and the view
factor prediction are not in agreement because the evanescent waves are contributing to the
radiative heat transfer in addition to the propagative wave contribution predicted by the view factor
(insets, Ext. Data Fig. 8b and Fig. 3b).
We also measured the evolution of ISC as a function of the lateral displacement of the emitter along
the x or y axis, as it is the main parameter used to accurately position the emitter above the center
of the active area of the cell (Ext. Data Fig. 8c). In this example the hot emitter was placed at d ≈
10 µm from the cell surface and was moved laterally from approximately -120 µm to +120 µm
relative to the center of the cell, while the short-circuit current was measured. As expected we
observed a maximum for the current when the emitter is above the center (see optical-microscopy
view from top in Ext. Data Fig. 8a), which first decreases rapidly and then more smoothly as the
emitter goes away from the center. The measurements are shown in Ext. Data Fig. 8c., together
with the theoretical model using the view factor, with a normalization applied at the maximum
value. It can be noted that the measurements are slightly non-symmetrical. This issue is mainly
due to the fact that the motion of the x and y positioners is not always perfectly smooth and linear,
and can vary slightly over large displacements (larger than 100 µm). Most importantly, performing
this kind of measurement along both x and y directions for a lower displacement range
(approximately the size of the active area) provides a precise positioning of the emitter above the
center of the cell. This is performed at vertical distances lower than 5 µm.
9. Validation of the superposition principle
For photovoltaic cells, the superposition principle tells that the measured photocurrent is equal to
the sum of the current generated in dark conditions and the short-circuit current under
illumination59 in low-injection conditions60. So theory suggests that the shape of the I-V curve
remains the same and is just shifted in ISC depending on illumination. This principle is very
interesting experimentally because the entire I-V curve could be retrieved by measuring only the
curve in the dark and then measuring ISC as a function of illumination. In order to verify the
superposition principle, we started by measuring the I-V curve of a cell having a 20 µm active area
diameter under dark conditions with a cooled radiative shield over the active area of the cell to
block the ambient radiation coming from the environment. Then we used the emitter to provide
different levels of illumination to the cell by either changing the emitter-cell distance or the emitter
temperature (Ext. Data Fig. 9a). As expected we observed that when illumination increases the I-
V curves are lowered into the photogeneration quadrant, corresponding to larger ISC and VOC. It is
worth noticing that the fill factors range from 0.69 to 0.75 at the highest illumination, which is
remarkable for a cell with a low bandgap. Then we shifted each curves in current only so their ISC
37
were all set equal to 0 (Ext. Data Fig. 9b). We observe that all the curves are well superimposed,
so the superposition principle is valid for this kind of TPV cell. The fact that the superposition
principle is valid is a proof that the temperature of the cell remains the same (77 K). Since the
shape of the I-V curve strongly depends on temperature (see Ext. Data Fig. 7b,c), the curves would
not be superimposed if the cell temperature was modified.
2. Supplementary Discussion
10. Summary of the different studied configurations
In the present work, we studied different parameters such as the emitter temperature, thickness of
the cell substrate, diameter of the active area and p-doping level of the top layer. Ext. Data Tab. 1
sums up the main results obtained for each configuration, using either the superposition principle
(i.e "Sup. pr.") or full measurements by scanning the whole voltage range (i.e "≠ V."). The best
enhancement factor of 5.9 and electrical power density of 7.5 kW.m-2 were obtained with the
sample having the largest p-doping concentration.
Although the melting point of InSb is 800 K, it was possible to make experiments with the emitter
heated beyond this temperature (900 K) without degrading the cell at contact, most probably
because the thermal contact resistance between the emitter and the cell limits the cell heating at
contact. When the emitter is too hot, however, the cell can be locally heated at T > 800 K in the
contact region. An attempt made with an emitter at 1200 K led to a degradation of the cell. As a
consequence, we know that the threshold for the contact between the emitter and the cell is for an
emitter temperature between 900 K and 1200 K. This issue does not prevent possible
measurements with very hot emitters close to the cell, however the exact distance separating them
cannot be measured easily with our method (see Ext. Data. Fig. 1c).
Finally, it must be reminded that the thermal conductance measurements require averaging over
many curves. At room temperature, this is straightforward and about 100 curves allow obtaining a
satisfying signal-to-noise ratio in a decent time (few hours). At low temperature, the permanent
cooling of the cold finger, which does not reach quickly the stationary regime in regions far from
the sample, induces a displacement, in particular laterally, of the sample that requires manual
realignment of the sphere in front of the cell. This means that the time to perform experiments is
much longer. As a result, averaging was performed on a smaller number of data than when done
at room temperature and the near-field thermal conductance appears noisier. The long acquisition
time is also the reason why the near-field efficiency was not always determined in the experiments
reported in Tab. S1. It must also be noted that the estimation of the far-field contribution has some
influence on the determination of the near-field efficiency, which can lead to some uncertainty. As
a consequence, the near-field efficiencies provided in Tab. S1 should be considered with an
uncertainty of ~15 %.
38
Supplementary References
55. Derjaguin, B. V, Abrikosova, I. I. & Lifshitz, E. M. Direct measurement of molecular
attraction between solids separated by a narrow gap. Q. Rev. Chem. Soc. 10, 295 -- 329
(1956).
56.
Feingold, A. & Gupta, K. G. New analytical approach to the evaluation of configuration
factors in radiation from spheres and infinitely long cylinders. J. Heat Transfer 92, 69
(1970).
57. Karalis, A. & Joannopoulos, J. D. Transparent and 'opaque' conducting electrodes for ultra-
thin highly-efficient near-field thermophotovoltaic cells. Sci. Rep. 7, 14046 (2017).
58. Datas, A. & Vaillon, R. Thermionic-enhanced near-field thermophotovoltaics. Nano Energy
61, 10 -- 17 (2019).
59. Tarr, N. G. & Pulfrey, D. L. The Superposition Principle for Homojunction Solar Cells.
IEEE Trans. Electron Devices 27, 771 -- 776 (1980).
60. Blandre, E., Chapuis, P. O., & Vaillon, R. High-injection effects in near-field
thermophotovoltaic devices. Sci. Rep. 7, 15860 (2017).
Corresponding author
Rodolphe Vaillon ([email protected]).
39
|
1902.06727 | 2 | 1902 | 2019-08-29T14:17:33 | Single-step fabrication of surface waveguides in fused silica with few-cycle laser pulses | [
"physics.app-ph",
"physics.optics"
] | Direct laser writing of surface waveguides with ultrashort pulses is a crucial achievement towards all-laser manufacturing of photonic integrated circuits sensitive to their environment. In this Letter, few-cycle laser pulses (with a sub-10 fs duration) are used to produce subsurface waveguides in a non-doped, non-coated fused silica substrate. The fabrication technique relies on laser-induced microdensification below the threshold for nanopore formation. The optical losses of the fabricated waveguides are governed by the optical properties of the superstrate. We have measured losses ranging from less than 0.1~dB/mm (air superstrate) up to 2.8~dB/mm when immersion oil is applied on top of the waveguide. | physics.app-ph | physics |
Single-step fabrication of surface waveguides in
fused silica with few-cycle laser pulses
Federico J. Furch1, W. Dieter Engel1, Tobias Witting1, Armando
Perez-Leija1, Marc J. J. Vrakking1, and Alexandre
Mermillod-Blondin1*
1Max-Born-Institut fur Nichtlineare Optik und
Kurzzeitspektroskopie, Max-Born-Strasse, D-12489 Berlin,
*Corresponding author: [email protected]
Germany
August 30, 2019
Abstract
Direct laser writing of surface waveguides with ultrashort pulses is
a crucial achievement towards all-laser manufacturing of photonic inte-
grated circuits sensitive to their environment. In this Letter, few-cycle
laser pulses (with a sub-10 fs duration) are used to produce subsurface
waveguides in a non-doped, non-coated fused silica substrate. The fab-
rication technique relies on laser-induced microdensification below the
threshold for nanopore formation. The optical losses of the fabricated
waveguides are governed by the optical properties of the superstrate. We
have measured losses ranging from less than 0.1 dB/mm (air superstrate)
up to 2.8 dB/mm when immersion oil is applied on top of the waveguide.
Among the several fabrication techniques that may be used to produce
waveguides in glass substrates [12], fs-laser direct writing [8] is especially at-
tractive because it does not require a cleanroom environment, is cost-effective,
and offers a high throughput [12]. Fs-laser writing enables full 3d waveguide
fabrication in crystals [7], polymers [2] as well as glasses [10]. Although re-
search efforts have been mostly concentrated on volume microprocessing, the
direct fabrication of surface waveguides would greatly extend the domain of
applications of laser-written photonic integrated chips. The propagation of an
optical field in a surface waveguide implies the presence of an evanescent wave
which can be exploited for refractive index sensing [19, 20], plasmonic excitation
sensing [29] or Fourier-transform spectrometry [22]. However, using fs-laser di-
rect writing on or near the surface of the host material is challenging. Attempts
in non-optimized glasses have resulted in surface swelling [6], cracking [19] and
1
ablation [33]. In pure fused silica, the direct fabrication of near-surface waveg-
uides with a good refractive index contrast has even been considered impossible
[5]. The main reason is that fs-laser direct writing relies on pulse-to-pulse heat
accumulation to induce controlled, localized heating of the substrate. After cool-
ing, high-refractive index regions with light-guiding capabilities appear. In this
scheme, the magnitude of the stress load produced upon thermal relaxation is a
major limitation. Recently, several routes have been explored to enable surface
waveguide photoinscription. One strategy consists in using toughened glasses as
a host substrate [19, 20]. Another approach is to enhance the photosensitivity of
the glass by doping with silver ions [1]. A third method relies on suppressing the
air/dielectric interface by bonding a thin glass on top of the sample [5]. Because
the bond is ensured by weak van der Waals attractive forces (only manifesting
when the sample and the cover glass are in close contact), such a method might
only be applicable to planar substrates. In this Letter, we describe the direct
fabrication of surface waveguides in fused silica with the help of few-cycle laser
pulses. Our method does not rely on pulse-to-pulse heat accumulation, but in-
stead is based on the type 1 laser-matter interaction regime [25] triggered in
a grazing incidence irradiation scheme. Modifications induced in the type 1
regime exhibit a smooth, uniform and positive refractive index change ∆n, in
contrast to type 2 modifications which are characterized by an important bire-
fringence due to the presence of periodic nanogratings in the irradiated region.
Furthermore, we demonstrate the possibility to control the propagation losses
by playing on the refractive index of the superstrate which opens a route for the
laser-assisted fabrication of non-hermitian microoptical systems [9].
The experimental setup is depicted in Fig. 1. Few-cycle pulses (sub-10 fs
duration, central wavelength at 800 nm) from a high repetition rate (400 kHz),
non-collinear optical parametric amplifier [11] are focused on the surface of a
fused silica sample with a grazing incidence. The samples are parallelepipedic
and polished to optical quality on all sides. In order to preserve the temporal
structure of the laser pulse in the focal plane, we use a gold-coated reflective
objective (numerical aperture 0.5) [26]. Such an irradiation scheme provides
two focii F1 and F2 formed by the upper and lower halves of the laser beam,
respectively. The presence of a planar air/glass interface induces wavefront
distortions and leads to the formation of an aberrated focus F2 [18] which is
shifted and spread along the propagation axis. The corresponding laser-induced
modifications are shown in Fig. 1(b). An analogous distortion happens in the
time domain with consequences on the temporal profile of the irradiation [31].
Because the spatio-temporal distortions imparted on the lower half of the beam
increase with the amount of propagation in the sample, the irradiation char-
acteristics vary when translating the sample along the propagation axis.
In
order to limit the irradiation to the part of the beam traveling through air only,
the lower half of the laser beam was blocked. Furthermore, the entrance of the
beam diameter was slightly reduced with the help of an iris, providing an overall
transmission of 0.17 for the focusing unit. The fabrication of a single line-shaped
microstructure is then straightforward. It suffices to irradiate the surface of the
substrate continuously from edge to edge by translating the sample with the
2
Figure 1: (a): Cross-sectional view of the experimental setup used for laser
microprocessing at grazing incidence. The part of the laser beam represented
in bright red propagates through air only and forms the focus F1. The beam
blocker can be put in the beam path to prevent the formation of the aberrated
focus F2. (b): Phase-contrast microscopy (PCM) picture of the sample's top
surface showing the refractive index distribution resulting from F1 and F2.
help of a stepper motor. In what follows, the laser pulse energy was constant
with a value of E = 530 nJ after the microscope objective and the speed of the
stepper motor was 60 µm s−1. The laser-induced microstructures had a length
of 10 mm (i.e. the width of the sample). For higher pulse energies, intense laser
light scattering occurred and the irradiated region exhibited a mix of negative
and positive refractive index changes (not shown). These features are indicative
of the type 2 interaction regime characterized by the formation of nanopores
[4, 25].
In Fig. 2(a) we show the laser footprint on the exit facet after sample pol-
ishing, using optical transmission microscopy (numerical aperture 0.9). The
laser-induced microstructure has a width w ≈ 2 µm and a height h ≈ 6 µm.
Diagnostics of the top surface with an atomic force microscope (AFM) are pre-
sented in Fig. 2(b), and reveal the presence of a shallow surface topography
variation (< −10 nm, about 430 nm FWHM) on top of the laser-induced mi-
crostructure. The negative sign of the topology variation is indicative of a
volume reduction (and hence a density increase) in the irradiated area, and is
the opposite to what happens when microprocessing is performed in the type 2
regime, with longer (35 fs) and more energetic (about 2 µJ) laser pulses where
surface swelling as high as 250 nm was measured [6]. An inversion in the sign of
the surface topography is consistent with recent observations reporting a volume
reduction of glass cantilevers irradiated in the type 1 regime and a net volume
increase of cantilevers irradiated in the type 2 regime [4]. We emphasize that the
absence of material re-deposition in the AFM pictures hints towards a purely
non-ablative process. The phase shift distribution ∆φ across the laser-induced
microstructure presented in Fig. 2(c) was measured by spatial light interference
microscopy [35]. As expected from a local density increase, ∆φ is positive in the
irradiated volume [32]. The corresponding spatial average of the laser-induced
3
refractive index change ∆n = λc∆φ
≈ 0.006, where λc =550 nm is the cen-
2πh
tral wavelength of the illumination source (an halogen light bulb in our case),
h = 6 µm is defined in Fig. 2(a) and ∆φ = 0.43 rad is the phase shift measured
at the center of the microstructure. We emphasize that the magnitude of ∆n
obtained exceeds the value of 10−4
− 10−3 usually measured in the bulk for type
1 modifications [25].
Laser-induced microstructure (end facet)
Air
h ~ 6
m
Substrate
(fused silica)
(a)
w ~ 2
m
10
m
Height [nm]
Phase shift [rad.]
AFM
QPCM
430 nm
(b)
2
m
(c)
2
m
]
s
t
i
n
u
.
b
r
a
[
y
t
i
s
n
e
t
n
I
(d)
D1
Main Band
D2
3
Wavenumber [cm-1]
Figure 2: Characterization of the laser-induced optical structures. (a): Side
view of the sample acquired with an optical transmission microscope. The
sample is illuminated with an halogen lamp. (b): Surface topography (top view)
measured with an atomic force microscope (AFM).(c): Phase shift across the
microstructure, measured with a spatial light interference microscope (SLIM).
(d): Micro-Raman investigations of the irradiated zone (purple line) and of the
pristine sample (grey line). The spectra were normalized with respect to the ω3
band.
To further confirm our conclusion on the laser-induced compaction and re-
fractive index increase, we have examined the signature of the fs laser-induced
structural modifications in micro-Raman spectra [25, 30] measured in the pris-
4
tine and irradiated material, as shown in Fig. 2(d). The excitation wavelength
was 442 nm and the microscope objective used for excitation and collection of the
Raman signal has a numerical aperture of 0.8 which provides a depth resolution
of 0.85 µm. The spectra were normalized with respect to the amplitude of the
ω3 band [28]. They reveal an increase of the D2 peak (centered at 600 cm−1) in
the irradiated volume, confirming a local laser-induced compaction [3, 4]. These
micro-Raman measurements were carried out on the top surface of the sample
and might not necessarily correspond to the maximum of compaction, presum-
ably located in the center of the laser-induced microstructure (i.e. ≈ 3 µm away
from the surface).
In order to examine the ability of the microstructures to guide light at op-
tical frequencies, the fundamental mode of a CW He-Ne beam (λ = 633 nm)
was focused onto the entrance facet of the laser-induced microstructures with a
microscope objective (numerical aperture 0.42). A second microscope objective
(Olympus MPlan, 100x, numerical aperture 0.9) used in combination with a
tube lens (focal length of 200 mm) formed a 111-fold magnified image of the
end facet on a camera sensor. Figure 3(a) shows the obtained output intensity
distribution. It demonstrates that these laser-induced microstructures support
optical waveguiding at 633 nm. A modal analysis in the directions parallel (p−)
and perpendicular (s−) to the surface of the sample (see Fig. 3(b)) provides a
mode field diameter (MFD, defined as the 1/e2 decay of the maximum mode
intensity), of 4.3 and 6.0 µm in the p− and s− directions, respectively. Out
of the center region, the guided mode intensity decays exponentially [16] with
decay constants of ≈ 1.1 µm and 0.4 µm in glass and in air, respectively. We
checked that the optical transfer function of the microscope objective did not
significantly influence these values by applying a deconvolution algorithm to the
curves shown in Fig. 3(b). The point spread function used for the deconvolution
was estimated numerically based on the model of Gibson and Lanni [13, 23].
The influence of the input polarization was examined by placing the waveg-
uide between a polarizer and an analyzer. The transmission of the waveguide
was measured as a function of the relative angle between the polarizer and the
analyzer [see Fig. 3(b)]. The polarization is maintained for input fields with
a linear polarization in the s- and p- directions (see thick transparent lines
in Fig. 3(b) left). However, an input field with a linear input polarization in
another direction becomes elliptically polarized upon propagation in the waveg-
uide, which indicates that s- and p- polarized fields have different propagation
constants.
Having confirmed the waveguiding capabilities of the laser-induced microstruc-
tures and their polarization-maintaining properties, we now examine the pos-
sibility to control the propagation losses by varying the refractive index of the
superstrate ns. By applying the optimum end-fire coupling method [15], losses
< 0.1 dB mm−1 were measured for ns = 1.00 (in air). The losses obtained when
applying a drop of immersion oil on top of the waveguide are shown in Fig. 4.
The diameter of the oil droplet was controlled by using a graduated microsy-
ringe.
A first order exponential fit of the experimental data indicates that the
5
Figure 3: (a): Near-field intensity profile at the exit of the laser-induced mi-
crostructure at a wavelength of 633 nm. (b): Mode-field analysis. The regions
in gray correspond to the size of the waveguide core deduced from Fig. 2(a).
The dotted lines are first-order exponential fits of the experimental data. MFD:
mode field diameter defined as the 1/e2 decay of the maximum intensity. (c):
Transmitted intensity as a function of the analyzer angle for linear input po-
larizations in the p- and s- directions (left) and for an arbitrary linear input
polarization (right). The dark thin lines represent the polarization of the beam
after propagation in the waveguide and the thick, semi transparent lines repre-
sent the polarization of the input beam.
6
]
s
t
i
n
u
.
b
r
a
[
y
t
i
s
n
e
t
n
i
d
e
t
t
i
m
s
n
a
r
T
Oil droplet
Waveguide
v
ii
iii
iv
Oil droplet diameter [mm]
Figure 4: Evolution of the optical transmission at 633 nm as a droplet of im-
mersion oil (noil = 1.516) with a variable diameter is placed on top of the
laser-induced surface waveguide. The dotted line represents a fit of the exper-
imental data using a single exponential decay. The presence of oil induces an
attenuation of 2.83+0.48
−0.66 dB mm−1.
superstrate-induced leakage is as strong as 2.83+0.48
−0.66 dB mm−1. The uncertain-
ties were obtained from fits of the lower and upper bounds of the experimental
values. In the so-called ray-optic approach of guided mode theory [34], the light
propagating through a waveguide is described as a sum of oblique rays experi-
encing total internal reflection at the boundary of the waveguide. Changing ns
changes the conditions for total internal reflection. When ns increases, the min-
imum angle for total internal reflection decreases and the steepest rays escape
the waveguide [27, 20].
The large attenuation coefficient deduced from the numerical fit indicates
that the optical structures are sensitive to the refractive index of their envi-
ronment and can thus be employed as refractive index sensors, with significant
potential for lab-on-chip applications. These waveguides may for instance con-
stitute the sensitive part of optical biosensors microsystems that are used for
label-free bio-sensing [29]. The optical structures presented in this Letter may
also be used as efficient photon/plasmon couplers to interface photonic and
plasmonic architectures [14], or as the backbone for the fabrication of compact
stationary-wave integrated Fourier-transform spectrometers [22]. Furthermore,
the ability to control the losses in integrated waveguide configurations opens the
door to new perspectives to manufacture non-Hermitian non-resonant photonic
systems in connection with exceptional points singularities [24].
In this Letter we have presented a method to inscribe waveguides directly
on the surface of a pure fused silica substrate. For the results presented here,
few-cycle pulses from a high repetition rate non-collinear optical parametric
amplifier have been utilized. The extent to which longer pulses can be utilized
will be the subject of future investigations. We note that stretching the pulse by
linear dispersion is not a viable option. The highly structured ultra broadband
spectrum quickly leads to a multi-pulse structure in the time domain, which
7
could potentially change the dynamics of the ionization process. We also note
that new trends in non-linear pulse compression have the potential to bring
few-cycle pulse capabilities to lasers typically used in laser material processing
[21].
The results presented in this Letter represent the first demonstration of
waveguides that are directly photoinduced on the surface of fused silica without
the need for pre- (e.g. deposition of a photosentive material or applicaton of
a cover slip) or post-processing of the target substrate. The optical structures
produced correspond to type 1 modifications, support waveguiding at optical
frequencies, possesses polarization-maintaining properties and exhibit a core re-
fractive index change of ∆n ≈ +0.006 on average. AFM measurements and
Micro-Raman investigations indicate that laser-induced microcompaction is at
the origin of the observed refractive index change. The waveguides are sensitive
to their environment, extending the capability of the direct laser write method
to the rapid prototyping of compact optical non-Hermitian microsystems tak-
ing advantage of all the well-known benefits (small size and weight, low power
consumption, improved reliability and vibration sensitivity [17]) of integrated
optical devices.
Funding Information
Deutsche Forschungsgemeinschaft, Grants ME4427/1-1 and ME4427/1-2.
Acknowledgments
The authors thank J. Tomm and S. Schwirzke-Schaaf for their assistance with
the micro-Raman measurements.
Supplemental Documents
References
[1] Alain Abou Khalil, Jean-Philippe B´erub´e, Sylvain Danto, Jean-Charles
Desmoulin, Thierry Cardinal, Yannick Petit, R´eal Vall´ee, and Lionel Can-
ioni. Direct laser writing of a new type of waveguides in silver containing
glasses. Scientific Reports, 7(1):11124, September 2017.
[2] Alexandra Baum, Patricia J. Scully, Maria Basanta, C. L. Paul Thomas,
Peter R. Fielden, Nicholas J. Goddard, Walter Perrie, and Paul R. Chalker.
Photochemistry of refractive index structures in poly(methyl methacrylate)
by femtosecond laser irradiation. Opt. Lett., 32(2):190 -- 192, Jan 2007.
[3] Y. Bellouard, E. Barthel, A. A. Said, M. Dugan, and P. Bado. Scanning
thermal microscopy and raman analysis of bulk fused silica exposed to low-
8
energy femtosecond laser pulses. Opt. Express, 16(24):19520 -- 19534, Nov
2008.
[4] Yves Bellouard, Audrey Champion, Benjamin McMillen, Sebabrata
Mukherjee, Robert R. Thomson, Charles P´epin, Philippe Gillet, and
Ya Cheng. Stress-state manipulation in fused silica via femtosecond laser
irradiation. Optica, 3(12):1285 -- 1293, Dec 2016.
[5] Jean-Philippe B´erub´e and R´eal Vall´ee. Femtosecond laser direct inscription
of surface skimming waveguides in bulk glass. Opt. Lett., 41(13):3074 -- 3077,
Jul 2016.
[6] V. R. Bhardwaj, P. B. Corkum, D. M. Rayner, C. Hnatovsky, E. Simova,
and R. S. Taylor. Stress in femtosecond-laser-written waveguides in fused
silica. Opt. Lett., 29(12):1312 -- 1314, Jun 2004.
[7] Feng Chen and J. R. V´azquez de Aldana. Optical waveguides in crystalline
dielectric materials produced by femtosecond-laser micromachining. Laser
& Photonics Reviews, 8(2):251 -- 275, 2014.
[8] K. M. Davis, K. Miura, N. Sugimoto, and K. Hirao. Writing waveguides in
glass with a femtosecond laser. Opt. Lett., 21(21):1729 -- 1731, 1996.
[9] Ramy El-Ganainy, Mercedeh Khajavikhan, Demetrios N. Christodoulides,
and Sahin K. Ozdemir. The dawn of non-hermitian optics. Communications
Physics, 2(1):37, March 2019.
[10] C. Florea and K. A. Winick. Fabrication and characterization of photonic
devices directly written in glass using femtosecond laser pulses. Journal of
Lightwave Technology, 21(1):246 -- 253, Jan 2003.
[11] Federico J. Furch, Achut Giree, Felipe Morales, Alexandria Anderson,
Yicheng Wang, Claus Peter Schulz, and Marc J. J. Vrakking. Close to
transform-limited, few-cycle 12 µj pulses at 400 khz for applications in
ultrafast spectroscopy. Opt. Express, 24(17):19293 -- 19310, Aug 2016.
[12] Andrea Chiappini Giancarlo C. Righini. Glass optical waveguides: a review
of fabrication techniques. Optical Engineering, 53:53 -- 53 -- 15, 2014.
[13] Sarah Frisken Gibson and Frederick Lanni. Diffraction by a circular aper-
ture as a model for three-dimensional optical microscopy. J. Opt. Soc. Am.
A, 6(9):1357 -- 1367, Sep 1989.
[14] Xin Guo, Min Qiu, Jiming Bao, Benjamin J. Wiley, Qing Yang, Xining
Zhang, Yaoguang Ma, Huakang Yu, and Limin Tong. Direct coupling of
plasmonic and photonic nanowires for hybrid nanophotonic components
and circuits. Nano Letters, 9(12):4515 -- 4519, 2009. PMID: 19995088.
[15] M. Haruna, Y. Segawa, and H. Nishihara. Nondestructive and simple
method of optical-waveguide loss measurement with optimisation of end-
fire coupling. Electronics Letters, 28(17):1612 -- 1613, 1992.
9
[16] Jonathan Hu and Curtis R. Menyuk. Understanding leaky modes: slab
waveguide revisited. Adv. Opt. Photon., 1(1):58 -- 106, Jan 2009.
[17] Robert G. Hunsperger. Integrated Optics Theory and Technology 6th Edi-
tion. Springer, 2009.
[18] N. Huot, R. Stoian, A. Mermillod-Blondin, C. Mauclair, and E. Audouard.
Analysis of the effects of spherical aberration on ultrafast laser-induced
refractive index variation in glass. Opt. Express, 15(19):12395 -- 12408, SEP
17 2007.
[19] Jerome Lapointe, Mathieu Gagn´e, Ming-Jun Li, and Raman Kashyap.
Making smart phones smarter with photonics. Opt. Express, 22(13):15473 --
15483, Jun 2014.
[20] Jerome Lapointe, Francois Parent, Elton Soares de Lima Filho, S´ebastien
Loranger, and Raman Kashyap. Toward the integration of optical sen-
sors in smartphone screens using femtosecond laser writing. Opt. Lett.,
40(23):5654 -- 5657, Dec 2015.
[21] L. Lavenu, M. Natile, F. Guichard, Y. Zaouter, X. Delen, M. Hanna,
E. Mottay, and P. Georges. Nonlinear pulse compression based on a gas-
filled multipass cell. Opt. Lett., 43(10):2252 -- 2255, May 2018.
[22] Etienne Le Coarer, Sylvain Blaize, Pierre Benech, Ilan Stefanon, Alain
Morand, Gilles Lerondel, Gregory Leblond, Pierre Kern, Jean Marc Fedeli,
and Pascal Royer. Wavelength-scale stationary-wave integrated fourier-
transform spectrometry. Nat Photon, 1(8):473 -- 478, August 2007.
[23] Jizhou Li, Feng Xue, and Thierry Blu. Fast and accurate three-dimensional
point spread function computation for fluorescence microscopy. J. Opt. Soc.
Am. A, 34(6):1029 -- 1034, Jun 2017.
[24] Mohammad-Ali Miri and Andrea Al`u. Exceptional points in optics and
photonics. Science, 363(6422), 2019.
[25] K. Mishchik, C. D'Amico, P. K. Velpula, C. Mauclair, A. Boukenter,
Y. Ouerdane, and R. Stoian. Ultrafast laser induced electronic and
structural modifications in bulk fused silica. Journal of Applied Physics,
114(13):133502, 2013.
[26] B. Piglosiewicz, D. Sadiq, M. Mascheck, S. Schmidt, M. Silies, P. Vasa, and
C. Lienau. Ultrasmall bullets of light -- focusing few-cycle light pulses to
the diffraction limit. Opt. Express, 19(15):14451 -- 14463, Jul 2011.
[27] F. Rehouma, D. Persegol, and A. Kevorkian. Optical waveguides for evanes-
cent field sensing. Applied Physics Letters, 65(12):1477 -- 1479, 1994.
10
[28] R. Saavedra, M. Le´on, P. Martin, D. Jim´enez-Rey, R. Vila, S. Girard,
A. Boukenter, and Y. Ouerdane. Raman measurements in silica glasses
irradiated with energetic ions. AIP Conference Proceedings, 1624(1):118 --
124, 2014.
[29] B Sep´ulveda, J S´anchez del R´ıo, M Moreno, F J Blanco, K Mayora,
C Dom´ınguez, and L M Lechuga. Optical biosensor microsystems based
on the integration of highly sensitive machzehnder interferometer devices.
Journal of Optics A: Pure and Applied Optics, 8(7):S561, 2006.
[30] N. S. Shcheblanov, M. E. Povarnitsyn, K. N. Mishchik, and A. Tanguy. Ra-
man spectroscopy of femtosecond multipulse irradiation of vitreous silica:
Experiment and simulation. Phys. Rev. B, 97:054106, Feb 2018.
[31] Bangshan Sun, Patrick S. Salter, and Martin J. Booth. Effects of sample
dispersion on ultrafast laser focusing. J. Opt. Soc. Am. B, 32(7):1272 -- 1280,
Jul 2015.
[32] C.Z. Tan, J. Arndt, and H.S. Xie. Optical properties of densified silica
glasses. Physica B: Condensed Matter, 252(12):28 -- 33, 1998.
[33] Gustavo A. Torchia, Pablo F. Meil´an, Airan Rodenas, Daniel Jaque, Cruz
Mendez, and Luis Roso. Femtosecond laser written surface waveguides
fabricated in nd:yag ceramics. Opt. Express, 15(20):13266 -- 13271, Oct 2007.
[34] R. Ulrich and R. J. Martin. Geometrical optics in thin film light guides.
Appl. Opt., 10(9):2077 -- 2085, Sep 1971.
[35] Zhuo Wang, Larry Millet, Mustafa Mir, Huafeng Ding, Sakulsuk Unaruno-
tai, John Rogers, Martha U. Gillette, and Gabriel Popescu. Spatial light
interference microscopy (slim). Opt. Express, 19(2):1016 -- 1026, Jan 2011.
11
|
1809.05603 | 1 | 1809 | 2018-09-12T19:11:47 | Towards On-Chip MEMS-Based Optical Autocorrelator | [
"physics.app-ph",
"physics.optics"
] | We propose a compact MEMS-based optical autocorrelator based on a micromachined Michelson interferometer in silicon and the two-photon absorption non-linearity in a photodetector. The miniaturized autocorrelator has a scanning range of 1.2 ps and operates in the wavelength range of 1100-2000 nm. The device measures the interferometric autocorrelation due to its collinear nature, from which the intensity autocorrelation can be calculated. The field autocorrelation can also be measured, from which the optical pulse spectrum can be calculated. A theoretical model based on Gaussian beam propagation is developed to study the effect of optical beam divergence, pulse dispersion, tilt angle between the interferometer mirrors, and amplitude mismatch between the interfering pulses. This model explains many of the effects observed in experimental measurements due to the use of a MEMS interferometer. The experimental results of autocorrelation signals for several pulses in the order of 100 fs are compared to a commercial autocorrelator and a good match is found. | physics.app-ph | physics | This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at
http://dx.doi.org/10.1109/JLT.2018.2867473
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
1
Towards On-Chip MEMS-Based Optical
Autocorrelator
Ahmed M. Othman, Student Member, IEEE, Hussein E. Kotb, Member, IEEE, Yasser M. Sabry,
Member, IEEE, Osama Terra, and Diaa A. Khalil, Senior Member, IEEE
on
the
based
can be
Abstract -- We propose a compact MEMS-based optical
autocorrelator
a micromachined Michelson
interferometer in silicon and the two-photon absorption non-
linearity in a photodetector. The miniaturized autocorrelator has
a scanning range of 1.2 ps and operates in the wavelength range of
1100-2000 nm. The device measures
interferometric
autocorrelation due to its collinear nature, from which the
intensity autocorrelation
field
autocorrelation can also be measured, from which the optical pulse
spectrum can be calculated. A theoretical model based on
Gaussian beam propagation is developed to study the effect of
optical beam divergence, pulse dispersion, tilt angle between the
interferometer mirrors, and amplitude mismatch between the
interfering pulses. This model explains many of the effects
observed in experimental measurements due to the use of a MEMS
interferometer. The experimental results of autocorrelation
signals for several pulses in the order of 100 fs are compared to a
commercial autocorrelator and a good match is found.
calculated. The
Index
Terms -- Autocorrelator,
integrated,
interferometric autocorrelation, micro-optical bench, ultrashort
pulse measurement.
dispersion,
I. INTRODUCTION
O
PTICAL autocorrelators are used for measuring ultrashort
pulses that have a width in the order of tens of picoseconds
or less. These pulses cannot be measured directly using
conventional photodetectors due to the slow response time of
the latter. Thus, autocorrelators are useful in the development
of ultrashort pulsed sources such as mode-locked lasers,
supercontinuum laser sources and optical frequency combs.
These sources have a wide range of applications from
spectroscopy and optical communication to applications in the
biomedical domain [1] -- [3]. However, autocorrelators are
usually bulky and expensive due to the use of many components
and mechanical moving parts that require precise alignment.
Many efforts have been exerted recently to design a compact
autocorrelator based on silicon photonics. But the reported
devices are either incapable of measuring sub-ps pulses [4] -- [6],
A. M. Othman is with the Faculty of Engineering, Ain Shams University,
Cairo 11517, Egypt (e-mail: [email protected]).
H. E. Kotb
is with
Telecommunication
[email protected]).
the Transmission Department, National
(e-mail:
11768,
Egypt
Cairo
Institute,
Y. M. Sabry and D. A. Khalil are with the Faculty of Engineering, Ain
Shams University, Cairo 11517, Egypt. They are also with Si-Ware Systems,
have a very limited wavelength range of operation [7], or only
capable of measuring pulses having a time-bandwidth product
greater than 100 [8]. In addition, some work has been reported
based on CdS or CdTe nanowires [9] -- [10] but the use of
alignment-sensitive optical components for coupling light into
the nanowires is still a challenge. Another technique that
requires high spatial coherence
few-cycle pulses
measurement using an angular tunable bi-mirror for non-
collinear autocorrelation is reported in ref. [11].
for
using
fabricated
interferometer
In this work, a MEMS-based autocorrelator that uses a
Michelson
silicon
micromachining technology is reported. The device uses the
two-photon absorption (TPA) non-linearity in a silicon
detector, allowing the potential of integration into a single chip.
The rest of this paper is organized as follows. Section II reviews
the background of optical autocorrelation and describes the
MEMS device. The experimental results are presented in
section III. Section IV discusses the non-ideal effects that can
be present in the interferometer such as beam divergence,
silicon dispersion, non-vertical mirror surfaces and amplitude
mismatch between the interfering pulses. Finally, the work is
concluded in section V.
II. THEORETICAL BACKGROUND AND DEVICE DESCRIPTION
typical collinear autocorrelator uses a Michelson
A
interferometer as shown in Fig. 1, where the input pulse is split
into two pulses using a beam splitter. One of the two
interferometer arms has a moving mirror to allow scanning the
delay (cid:1) between the two interfering beams. The output electric
field from the interferometer in the time domain can then be
written as [12]:
(cid:12) Re(cid:16)(cid:17)(cid:2)(cid:18)(cid:6)(cid:7)(cid:9)+(cid:2)(cid:18)(cid:6)(cid:7)−(cid:1)(cid:9)(cid:21)(cid:22)(cid:23)(cid:24)(cid:25)(cid:26)(cid:27)(cid:21)(cid:23)(cid:24)(cid:25)(cid:5)(cid:28)
(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7);(cid:1)(cid:9)= (cid:11)
where (cid:2)(cid:18)(cid:6)(cid:7)(cid:9) is the temporal pulse shape and (cid:29)(cid:30) is the optical
linear
angular frequency of
photodetector at the output of the interferometer, the output
the pulse. Using a slow
(1)
Cairo 11361, Egypt (e-mail: [email protected]; diaa.khalil@si-
ware.com).
O. Terra is with the Primary Length Standard and Laser Technology
Laboratory, National Institute of Standard, Giza 12211, Egypt (e-mail:
[email protected]; [email protected]).
Copyright (c) 2015 IEEE. Personal use of this material is permitted.
However, permission to use this material for any other purposes must be
obtained from the IEEE by sending a request to [email protected].
Copyright (c) 2018 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at
http://dx.doi.org/10.1109/JLT.2018.2867473
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
2
current from the detector yields the field autocorrelation signal
Fig. 1. Schematic of a Michelson interferometer.
given by:
(cid:31) !"#$(cid:6)(cid:1)(cid:9)=
%(cid:11)&' ()(cid:2)(cid:18)(cid:6)(cid:7)(cid:9))(cid:12)+*(cid:2)(cid:18)(cid:6)(cid:7)(cid:9)(cid:2)(cid:18)∗(cid:6)(cid:7)−(cid:1)(cid:9)(cid:21)(cid:23)(cid:24)(cid:25)(cid:26)+c.c../0(cid:7)1 (2)
where (cid:31) !"#$(cid:6)(cid:1)(cid:9) is the detector current versus delay, %(cid:11) is a
constant and c.c. denotes the complex conjugate. The Fourier
transform of the field autocorrelation signal yields the optical
spectrum of the input pulse [13].
the detector versus delay yields
Adding a non-linear element before the detector, the output
of
interferometric
autocorrelation signal. The output current from the detector in
this case is written as [11]:
the
(3)
(cid:31)!2(cid:5)"3 (cid:6)(cid:1)(cid:9)=2%(cid:12)' ()(cid:2)(cid:18)(cid:6)(cid:7)(cid:9))5+2)(cid:2)(cid:18)(cid:6)(cid:7)(cid:9))(cid:12))(cid:2)(cid:18)(cid:6)(cid:7)−(cid:1)(cid:9))(cid:12)/0(cid:7)+
%(cid:12)&2(cid:21)(cid:23)(cid:24)(cid:25)(cid:26)'(cid:2)(cid:18)(cid:6)(cid:7)(cid:9)(cid:2)(cid:18)∗(cid:6)(cid:7)−(cid:1)(cid:9)6)(cid:2)(cid:18)(cid:6)(cid:7)(cid:9))(cid:12)+)(cid:2)(cid:18)(cid:6)(cid:7)−(cid:1)(cid:9))(cid:12)70(cid:7)+
(cid:21)(cid:23)(cid:12)(cid:24)(cid:25)(cid:26)'*(cid:2)(cid:18)(cid:6)(cid:7)(cid:9)(cid:2)(cid:18)∗(cid:6)(cid:7)−(cid:1)(cid:9).(cid:12)0(cid:7)+c.c.1
where (cid:31)!2(cid:5)"3 (cid:6)(cid:1)(cid:9) is the detector current for the interferometric
autocorrelation signal, and %(cid:12) is a constant. The exponential
terms in this equation are fast-varying terms, which can be
suppressed by averaging the interferometric autocorrelation
signal over many fringes, yielding the intensity autocorrelation
signal:
(cid:31)!2(cid:5)"2(cid:6)(cid:1)(cid:9)=2%(cid:12)' ()(cid:2)(cid:18)(cid:6)(cid:7)(cid:9))5+2)(cid:2)(cid:18)(cid:6)(cid:7)(cid:9))(cid:12))(cid:2)(cid:18)(cid:6)(cid:7)−(cid:1)(cid:9))(cid:12)/0(cid:7)
which can be used to get the pulse width of the input pulse. It is
worth mentioning that the autocorrelation signal, defined by (3)
or (4), is ideally an even function of (cid:1) regardless of the
symmetry of the actual pulse. Furthermore, the ratio between
the maximum value and the background level for the
interferometric and intensity autocorrelation signals is 8 to 1
and 3 to 1, respectively [11].
the detector
achieved using step coverage of vertical surfaces [20]. Light is
propagating in-plane with respect to the chip substrate in a
micro-optical bench arrangement. The output beam from the
MEMS interferometer is then focused on a silicon detector that
is used outside its linear absorption wavelength range to exploit
the TPA process as a source for the nonlinearity necessary for
obtaining information about the pulse width. The output current
from
interferometric
autocorrelation signal can be constructed. The silicon detector
can also be replaced in the same setup by an InGaAs
photodetector, allowing the field autocorrelation signal to be
measured. The Fourier transform of the field autocorrelation
signal yields the pulse optical spectrum. Therefore, the
presented device can be used as an optical autocorrelator or a
spectrometer for full characterization of the pulse in the time
and wavelength domains.
is measured and
the
Fig. 2. SEM image for a fabricated MEMS-based (a) Michelson interferometer
and (b) comb-drive actuator.
III. EXPERIMENTAL RESULTS
the proposed autocorrelator and
A. Device and Measurement Setup
A schematic of
the
measurement setup is shown in Fig. 3. The input pulse under
test is fed to the MEMS interferometer by the means of a GRIN
lensed fiber that partially collimates light to 8(cid:30) of about 10 9:
to decrease the divergence losses inside the interferometer. The
output beam from the interferometer is then tightly focused on
a silicon avalanche photodetector (Thorlabs APD130A2) using
a microscope objective lens with a negligible dispersion effect
on the measured pulse width. The silicon photodetector
generates a current proportional to the square of the input
optical intensity by the TPA process; allowing it to replace the
second harmonic generation (SHG) crystal typically used in
scanning autocorrelators [21]. Since the MEMS interferometer
is also fabricated in silicon, the proposed autocorrelator has the
potential of integration into a single chip by combining the
photodetector onto the same die. For the TPA to be the
dominant absorption mechanism, the input pulse wavelength
should be outside the linear absorption range of silicon (400 nm
-- 1100 nm). In addition, the input pulse wavelength should be
within double the wavelength range of the linear absorption of
silicon. Combining these two conditions determines the
possible wavelength range of the device to be from 1100 nm to
about 2000 nm. To measure the pulse autocorrelation signal,
electronic circuits drive the MEMS-based comb-drive actuator
and measure the current from the photodetector. The measured
(4)
The proposed device uses a MEMS interferometer, as shown
in Fig. 2(a). The MEMS interferometer is composed of a silicon
beam splitter, a fixed mirror and a moving mirror driven by a
comb-drive actuator as shown in Fig. 2(b). All the components
of the MEMS chip are fabricated at the same time in a self-
aligned manner, which is crucial for the device operation. The
self-alignment is enabled by the photolithographic accuracy and
subsequent etching [14] -- [19]. The mirror metallization is
Copyright (c) 2018 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at
http://dx.doi.org/10.1109/JLT.2018.2867473
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
3
signals are then processed to obtain the pulse autocorrelation
signal versus the temporal delay.
Fig. 3. Schematic of the proposed autocorrelator and the experimental setup for
autocorrelation measurement.
B. Measurement Results
To test the autocorrelator, a femtosecond mode-locked fiber
laser is fed to the autocorrelator input to measure its pulse
width. The mode-locked laser has a central wavelength and a
pulse repetition rate of 1560 nm and 16 MHz, respectively. The
gain medium of the laser is an erbium doped fiber that is
pumped by a laser diode at 980 nm. The output pulse shape and
width from the source can be changed by changing the pump
laser diode output power. The pulse
interferometric
autocorrelation was measured at 3 values of pump laser diode
power; namely 180 mW, 215 mW and 240 mW as shown in
Fig. 4. The fringe-averaged intensity autocorrelation was
calculated from the interferometric autocorrelation and the
result was compared to the intensity autocorrelation measured
using a commercial autocorrelator (A. P. E. pulseCheck). The
measured autocorrelation full width half maximum (FWHM)
using the proposed device is 178 fs, 183 fs and 166 fs
corresponding to pulse widths of 126 fs, 130 fs and 117 fs,
respectively, assuming a Gaussian pulse shape for the
deconvolution factor. Table 1 lists the measured pulse width at
different values of mode-locked laser pump power for the
proposed MEMS-based autocorrelator and the commercial
device, where the results indicate a good match.
The proposed device is also used to measure the pulse optical
spectrum only by replacing the silicon detector by an InGaAs
detector, which works as a linear detector. The measured
autocorrelation signal in this case is the field autocorrelation.
The Fourier transform of the field autocorrelation yields the
pulse optical power spectrum. The proposed device may be
potentially used to measure the pulse width and optical
spectrum simultaneously by integrating an on-chip beam
splitter in the output of the interferometer, where one path can
be directed to a silicon detector and the other path to an InGaAs
one. The measured field autocorrelation signal using an InGaAs
detector with our device is shown in Fig. 5 along with the
corresponding calculated pulse optical
spectrum. The
calculated optical spectrum is compared to the optical spectrum
measured using an optical spectrum analyzer (OSA) and a good
match is found given the higher resolution of the OSA
measurement.
Fig. 4.
Measured interferometric autocorrelation (left) and intensity
autocorrelation (right) using the proposed autocorrelator for the mode-locked
laser pump power of (a) 180 mW, (b) 215 mW and (c) 240 mW. The intensity
AC is compared to that of a reference commercial autocorrelator. A Gaussian
pulse fitting for the measured autocorrelation is also plotted as the dash-dotted
curve. The inset of the top-left figure shows a zoom in of the interferometric
AC.
TABLE I
MEASURED PULSE WIDTH USING THE PROPOSED DEVICE AND USING A
REFERENCE COMMERCIAL DEVICE
Pump Power
Proposed Device Pulse
Reference Device Pulse
(mW)
180
215
240
Width (fs)
Width (fs)
126
130
117
103
116
108
Fig. 5. Measured field autocorrelation signal (left) and the corresponding pulse
optical spectrum (right). The optical spectrum is compared to the measurement
of an OSA.
C. Device Limitations
The autocorrelator scanning range is limited by the moving
mirror travel range, which is about 200 µm in our case. This
corresponds to a scanning range of about 1.2 ps. The
autocorrelator sensitivity is defined as the minimum detected
product of the peak and average power [6]. It was measured by
adding a variable optical attenuator after the mode-locked fiber
laser. The attenuation was increased until the minimum
detectable pulse was reached. The sensitivity was then
calculated to be about 8 W2. The relatively low sensitivity of
Copyright (c) 2018 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at
http://dx.doi.org/10.1109/JLT.2018.2867473
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
4
the device is due to the insertion loss of the MEMS-based
interferometer. The minimum measurable pulse width is limited
by the silicon dispersion, due to the relatively high silicon group
velocity dispersion (GVD), which is equal to 1108 fs2/mm at
1560 nm. The silicon propagation distance for MEMS-based
the minimum
Assuming an acceptable error of 10%,
measurable pulse width can be estimated to be equal to 58 fs
Michelson interferometers ranges from 500 μm to 1000 μm.
and 82 fs for a silicon propagation distance of 500 μm and
1000 μm, respectively.
IV. SIMULATION RESULTS
in Fig. 6(b). Since the divergence of Gaussian beams is
wavelength-dependent, the input pulse electric field in time
domain (cid:2)(cid:18)(cid:4)#@"A(cid:6)(cid:7)(cid:9) is transformed to the frequency B domain
(cid:2)(cid:18)(cid:4)#@"C(cid:6)B(cid:9) =D.E.(cid:16)(cid:2)(cid:18)(cid:4)#@"A(cid:6)(cid:7)(cid:9)(cid:28)
using the Fourier transform.
(5)
the measured
The measured autocorrelation signals have some discrepancies
from the ideal theoretical case due to different effects in the
MEMS interferometer as will be explained in this section.
Asymmetry could be observed in the measured interferometric
autocorrelation. Also,
and
interferometric autocorrelation signals were found to be not
following the theoretical ratio between the maximum value and
the background level. For the sake of comparison with the
reference
intensity
autocorrelations measured by the MEMS-based autocorrelator
shown in Fig. 4 were scaled and down-shifted. The unscaled
intensity autocorrelation measured at a pump power of 215 mW
is shown in Fig. 6(a).
autocorrelator measurements,
intensity
the
Propagation of the beam in the MEMS-based Michelson
interferometer has some effects on the input light beam such as
divergence and dispersion. Moreover, the interferometer is not
ideal as the mirrors may be slightly tilted from the ideal
position, and the amplitude of the two interfering signals may
be different. A simulation model is developed as discussed in
the following subsections to study these effects and their impact
on the measured autocorrelation signals.
A. Divergence Effect
The input light beam to the interferometer is not perfectly
collimated and, hence, suffers from beam divergence. Due to
the unequal distance travelled by the two interfering beams at
non-zero optical path difference values, the two beams have
different values of width and a different phase profile.
A simulation model based on the Gaussian beam propagation
is developed to study this effect. The two beams are assumed to
be initially identical Gaussian beams with a beam waist radius
of 8(cid:30). The profiles of the beams are then calculated after
propagation in the interferometer. The total field at each point
on the detector head is calculated. Then, the autocorrelation
signal is calculated for both the field autocorrelation (without
non-linearity), and the interferometric autocorrelation, by
changing the temporal delay (cid:6)(cid:1)(cid:9) between the two pulses and
calculating the detector current at each delay value. The
intensity autocorrelation (fringe-averaged) is also calculated
from the interferometric autocorrelation.
Two temporal pulse shapes (cid:2)(cid:18)(cid:4)#@"A(cid:6)(cid:7)(cid:9) are chosen in our
simulation model. The first one has a symmetric intensity
profile to have side lobes like the practically measured pulse
(Fig. 6(b)). The input pulse width is 240 fs. The second pulse
has an asymmetric pulse shape, with the intensity profile shown
(6)
Fig. 6. (a) Intensity autocorrelation measured at pump power 215 mW before
and after shifting and scaling. (b) The pulse intensity profile for the asymmetric
pulse used in simulation.
delay is calculated in the frequency (or wavelength) domain as
traveled by the second beam and c is the speed of light in air.
The total field at the detector head (cid:2)$"(cid:5)C (cid:6)B,G,H;(cid:1)(cid:9) at each
the sum of the two interfering Gaussian electric fields (cid:2)(cid:11) and
(cid:2)(cid:12) such that:
(cid:2)$"(cid:5)C (cid:6)B,G,H;(cid:1)(cid:9)=(cid:2)(cid:11)(cid:6)B,G,H;I(cid:11)(cid:9)+(cid:2)(cid:12)(cid:6)B,G,H;I(cid:12)(cid:9)
where x and y are the transverse space coordinates, I(cid:11) is the
distance traveled by the first beam, I(cid:12) =I(cid:11)+(cid:1) J is the distance
The fields (cid:2)(cid:11) and (cid:2)(cid:12) are defined as [22]:
(cid:2)!(cid:6)B,G,H;I!(cid:9)=(cid:2)(cid:18)(cid:4)#@"C(cid:6)B(cid:9) 8(cid:30)8(cid:6)I!,B(cid:9)expM− G(cid:12)+H(cid:12)
8(cid:12)(cid:6)I!,B(cid:9)N
∙ expP−QM R(cid:6)B(cid:9)I! + S(cid:6) (cid:9)(cid:17)TUVWU(cid:27)
(cid:12)X(cid:6)YZ, (cid:9) −[(cid:6)I!, B(cid:9)N\
where ] = 1 and 2, 8(cid:30) is the input beam waist radius, 8(cid:6)I!,B(cid:9)
is the beam radius at distance I! from the beam waist, ^(cid:6)I!,B(cid:9)
the radius of curvature of the beam wavefront at I!, [(cid:6)I!,B(cid:9) is
the Gouy phase [22] at I! and R(cid:6)B(cid:9) =2_B/J is the propagation
(7)
constant.
Due to the non-linearity of the detector (as TPA is considered
the main mechanism for non-linearity in calculating the
interferometric autocorrelation), the output current from the
detector is dependent on both the interfering pulses temporal
profile and their average value. Therefore, the detector output
current should be calculated at each time point of the interfering
pulses at each value of delay. Thus, the total field at the detector
is calculated in the time domain by calculating the inverse
Fourier transform of (cid:2)$"(cid:5)C (cid:6)B,G,H;(cid:1)(cid:9):
(cid:2)$"(cid:5)A (cid:6)(cid:7),G,H;(cid:1)(cid:9)=(cid:31).D.E.&(cid:2)$"(cid:5)C (cid:6)B,G,H;(cid:1)(cid:9)1.
The detector output current (cid:31)!2(cid:5)"3 (cid:6)(cid:1)(cid:9) is calculated for the
interferometric autocorrelation case as a function of delay
(8)
Copyright (c) 2018 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at
http://dx.doi.org/10.1109/JLT.2018.2867473
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
5
(9)
(10)
assuming a non-linear detector that employs TPA. It is given
as:
detector is calculated as:
(cid:31)!2(cid:5)"3 (cid:6)(cid:1)(cid:9) =a''')(cid:2)$"(cid:5)A (cid:6)(cid:7),G,H;(cid:1)(cid:9))50(cid:7) 0G 0H
Also, the field autocorrelation (cid:31) !"#$(cid:6)(cid:1)(cid:9) assuming a linear
(cid:31) !"#$(cid:6)(cid:1)(cid:9)=b''')(cid:2)$"(cid:5)A (cid:6)(cid:7),G,H;(cid:1)(cid:9))(cid:12)0(cid:7) 0G 0H
where b and a are constants that depend on the responsivity of
two-photon absorption coefficient,
the detector and
respectively. Their values are not
the
autocorrelation signals are normalized to their maximum value.
Fig. 7 shows
interferometric and field
autocorrelation signals at 8(cid:30) values of 5 µm, 10 µm and 20 µm,
as well as for the asymmetric pulse case at a 8(cid:30) of 10 µm. The
8(cid:30) and a slight asymmetry can be observed in the resulting
amplitude of the autocorrelation side lobes varies by changing
the simulated
important as
interferometric and
for both
the
field
its
autocorrelation
autocorrelation signals.
Fig. 7. Simulated interferometric autocorrelation (left) and field autocorrelation
(right) signals for 8(cid:30) values of (a) 5 µm, (b) 10 µm, and (c) 20 µm; and (d) for
the asymmetric pulse at a 8(cid:30) of 10 µm.
B. Dispersion Effect
The fabricated interferometer uses a silicon beam splitter.
Silicon has a relatively large GVD of 1108 fs2/mm at 1560 nm,
as calculated from the Sellmeier coefficients of silicon [23].
The fabricated interferometer structure is compensated for
dispersion in the sense that both interferometer arms have the
same silicon propagation distance [24]. However,
this
dispersion compensation may not be perfect and some silicon
propagation distance mismatch may be present in the fabricated
interferometer. To simulate this effect, the second interfering
term of
beam (cid:2)(cid:12)(cid:6)B,G,H;I(cid:12)(cid:9)
exp (cid:6)−Qc@!(cid:6)B(cid:9) R(cid:6)B(cid:9) Δ0(cid:9), where c@! is the silicon refractive
function of frequency and Δ0 is the silicon propagation distance
index calculated using the Sellmeier coefficients of silicon as a
mismatch. The simulated interferometric as well as field
is given an extra phase
autocorrelation signals are shown in Fig. 8(a), 8(b) and 8(c) for
shows the interferometric and field autocorrelation signals for
Δ0 values of 50 µm, 200 µm, 600 µm, respectively. Fig. 8(d)
the asymmetric pulse at Δ0 of 200 µm. It is observed that
introducing a dispersion mismatch between the arms affects the
interference peak-to-background ratio for the interferometric
autocorrelation and the interference visibility for the field
autocorrelation. A clear asymmetry can also be observed for the
interferometric autocorrelation of the asymmetric pulse, but not
for the field autocorrelation. The intensity autocorrelation is
shown
intensity
autocorrelation is much less affected by the mismatch, and
shows no asymmetry for an asymmetric pulse shape.
in Fig. 9(a), which shows
that
the
Fig. 8. Simulated interferometric autocorrelation (left) and field autocorrelation
(right) signals for a silicon propagation distance mismatch Δ0 of (a) 50 µm, (b)
200 µm, and (c) 600 µm; and (d) for the asymmetric pulse at a Δ0 of 200 µm.
C. Surfaces Tilt Angle Effect
The fabricated MEMS interferometer beam splitter and
reflecting mirrors are not perfectly vertical due to fabrication
technology limitations and, hence, the interfering beams may
have slightly shifted centers at the detector [25]. This effect is
introduced to the simulation model by making one of the beams
propagate along a slightly tilted axis from the I direction. The
simulated interferometric and field autocorrelation signals are
shown in Fig. 10(a) and 10(b) for tilt angles of 0.10 and 0.50
respectively, while Fig. 10(c) shows them for the asymmetric
pulse case at a tilt angle of 0.50. The tilt angle affects the pulse
side lobe shape in addition to affecting the interference
visibility. The field interference visibility becomes less than
unity and
to
background ratio is no longer 8 to 1. Also, the interferometric
autocorrelation becomes slightly asymmetric. However, the
intensity autocorrelation is much less affected by the tilt angle
and it shows no noticeable change in shape by changing the tilt
angle to from 0.10 to 0.50 as shown in Fig. 11.
D. Unequal Amplitude of Interfering Beams Effect
interferometric autocorrelation peak
the
The two interfering beams travel different paths after
splitting and hence may be subject to unequal losses. This may
also happen due to the different spot sizes at the detector
Copyright (c) 2018 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at
http://dx.doi.org/10.1109/JLT.2018.2867473
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
6
resulting from different propagation distance, which may cause
the larger spot to be truncated due to the limited size of the
detector. To study this effect, a simple model is developed for
the interferometer, where the two beams are assumed to have
unequal amplitude. The amplitude ratio between the electric
field of the two beams was varied from 1 to 0.2.
their shape. However, for the interferometric autocorrelation,
the amplitude ratio also affects the shape of the side lobes of the
autocorrelation signal in addition to its peak to background
ratio. In addition, for the asymmetric input pulse shape, a clear
asymmetry is observed.
Fig. 9. (a) Simulated intensity autocorrelation signals for a silicon propagation
distance mismatch Δ0 of 50 µm, 200 µm and 600 µm. (b) The intensity
autocorrelation signal for the asymmetric pulse at Δ0 of 200 µm compared to
an ideal interferometer.3
Simulated
Fig. 10.
interferometric autocorrelation (left) and field
autocorrelation (right) signals for surface tilt angle values of (a) 0.10 and (b)
0.50; and (c) for the asymmetric pulse at a tilt angle of 0.50
Fig. 11. (a) Simulated intensity autocorrelation signals for surface tilt angle
values of (a) 0.10 and (b) 0.50. (b) The intensity autocorrelation signal for the
asymmetric pulse at a tilt angle of 0.50 compared to an ideal interferometer.
The simulated results for
interferometric and field
autocorrelation signals are shown Fig. 12(a), 12(b) and 12(c)
for amplitude ratios of 1, 0.5 and 0.25, respectively. Fig. 12(d)
shows the interferometric and field autocorrelation signals for
the asymmetric input pulse. Fig. 13(a) shows the intensity
autocorrelation signals for the symmetric pulse at different
values of pulse amplitude ratio, while Fig. 13(b) shows the
intensity autocorrelation signal for the asymmetric pulse at a
pulse amplitude
ideal
interferometer case. The amplitude ratio is shown to have only
a scaling effect on the field and intensity autocorrelation signals
(a change in peak to background ratio) and no effect at all on
the
ratio of 0.2 compared
to
Simulated
Fig. 12.
interferometric autocorrelation (left) and field
autocorrelation (right) signals for interfering pulses amplitude ratio of (a) 1, (b)
0.5, (c) 0.25; and (d) for the asymmetric pulse at a pulse amplitude ratio of 0.2
The asymmetry in the interferometric autocorrelation can be
attributed to the second integral in (3), which for a small
delayed pulse becomes a cross correlation between
)(cid:2)(cid:18)(cid:6)(cid:7)(cid:9))(cid:12)(cid:2)(cid:18)(cid:6)(cid:7)(cid:9) and (cid:2)(cid:18)(cid:6)(cid:7)(cid:9). Hence, it is asymmetric if (cid:2)(cid:18)(cid:6)(cid:7)(cid:9) is
not symmetric. The autocorrelation signals in Fig. 12(d) and
Fig. 13(b) show a good agreement with the measurement results
in Fig. 4 that indicates that the measured pulse was an
asymmetric pulse, which is a known case in mode-locked fiber
lasers [26], [27].
V. CONCLUSION
A compact MEMS-based optical autocorrelator has been
presented, which can measure both the pulse width and the
optical power spectral density. The device operates in the
wavelength range of 1100-2000 nm and has a scanning range
of 1.2 ps. Different femtosecond pulses were measured and
showed to have a good agreement with the measurements done
by a commercial autocorrelator. A simulation model has been
presented to study the effects of the light divergence, the beam
splitter dispersion, the surfaces tilt angle and the amplitude
mismatch of interfering beams that may exist in our system due
to the use of a MEMS-based interferometer instead of a
conventional one. Simulation results show impairments in the
interferometric autocorrelation signals similar to those in the
measurement results. Also, the intensity autocorrelation signals
illustrated in both the experimental and simulation results are
shown to have higher immunity to the light divergence, the
beam splitter dispersion, and the surfaces tilt angle than the
interferometric autocorrelation signals.
Copyright (c) 2018 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at
http://dx.doi.org/10.1109/JLT.2018.2867473
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
7
structures," US Patent Application 15/047,032, 2016.
[21] Y. Takagi, T. Kobayashi, K. Yoshihara, and S. Imamura, "Multiple- and
in
single-shot autocorrelator based on
semiconductors," Opt. Lett., vol. 17, no. 9, p. 658, 1992.
two-photon conductivity
[22] B. E. A. Saleh and M. C. Teich, "Beam Optics," in Fundamentals of
Photonics, John Wiley & Sons, Inc., 2001, pp. 80 -- 107.
[23] B. Tatian, "Fitting refractive-index data with the Sellmeier dispersion
formula," Appl. Opt., vol. 23, no. 24, pp. 4477 -- 4485, Dec. 1984.
[24] D. A. Khalil, B. Mortada, M. Nabil, M. Medhat, and B. A. Saadany,
"Compensated MEMS FTIR Spectrometer Architecture," Patent
US20110058180 A1, 2011.
[25] B. Saadany et al., "Micro-optical bench device with highly/selectively-
controlled optical surfaces," US Patent Application 15/047,205, 2016.
[26] M. Erkintalo, C. Aguergaray, A. Runge, and N. G. R. Broderick,
"Environmentally stable all-PM all-fiber giant chirp oscillator," Opt.
Express, vol. 20, no. 20, pp. 22669 -- 22674, Sep. 2012.
[27] Z. Zhang, B. Öktem, and F. Ö. Ilday, "All-fiber-integrated soliton-
similariton laser with in-line fiber filter," Opt. Lett., vol. 37, no. 17, pp.
3489 -- 3491, Sep. 2012.
Fig. 13. (a) Simulated intensity autocorrelation signals for different values of
interfering pulses amplitude ratio. (b) The intensity autocorrelation signal for
the asymmetric pulse at a pulse amplitude ratio of 0.2 compared to an ideal
interferometer.
REFERENCES
[1] S. Kawanishi, "Ultrahigh-speed optical
time-division-multiplexed
transmission technology based on optical signal processing," IEEE J.
Quantum Electron., vol. 34, no. 11, pp. 2064 -- 2079, Nov. 1998.
J. Serbin, T. Bauer, C. Fallnich, A. Kasenbacher, and W. H. Arnold,
"Femtosecond lasers as novel tool in dental surgery," Appl. Surf. Sci., vol.
197 -- 198, pp. 737 -- 740, 2002.
[2]
[3] P. Hannaford, Femtosecond laser spectroscopy. 2005.
[4] F. R. Laughton, J. H. Marsh, D. A. Barrow, and E. L. Portnoi, "The two-
photon absorption semiconductor waveguide autocorrelator," IEEE J.
Quantum Electron., vol. 30, no. 3, pp. 838 -- 845, Mar. 1994.
[5] S. Kinugasa, N. Ishikura, H. Ito, N. Yazawa, and T. Baba, "One-chip
integration of optical correlator based on slow-light devices," Opt.
Express, vol. 23, no. 16, pp. 20767 -- 20773, Aug. 2015.
[6] K. Kondo and T. Baba, "On-chip autocorrelator using counter-
propagating slow light in a photonic crystal with two-photon absorption
photodiodes," Optica, vol. 4, no. 9, pp. 1109 -- 1112, Sep. 2017.
[7] C. Monat et al., "Integrated optical auto-correlator based on third-
harmonic generation in a silicon photonic crystal waveguide," Nat.
Commun., vol. 5, p. 3246, Feb. 2014.
[8] A. Pasquazi et al., "Sub-picosecond phase-sensitive optical pulse
characterization on a chip," Nat. Photonics, vol. 5, p. 618, Aug. 2011.
[9] H. Yu et al., "Single Nanowire Optical Correlator," Nano Lett., vol. 14,
no. 6, pp. 3487 -- 3490, Jun. 2014.
[10] C. Xin et al., "Single CdTe Nanowire Optical Correlator for Femtojoule
Pulses," Nano Lett., vol. 16, no. 8, pp. 4807 -- 4810, Aug. 2016.
[11] A. Treffer, J. Brunne, M. Bock, S. König, U. Wallrabe, and R. Grunwald,
"Adaptive non-collinear autocorrelation of few-cycle pulses with an
angular tunable bi-mirror," Appl. Phys. Lett., vol. 108, no. 5, p. 51103,
2016.
[12] A. Weiner, "Ultrafast Optics," Wiley Publishing, 2009, pp. 90 -- 106.
[13] M. Erfan, Y. M. Sabry, M. Ragheb, D. A. M. Khalil, and Society of Photo-
optical Instrumentation Engineers, Optical gas sensing with miniaturized
MEMS FTIR spectrometers. .
[14] B. Mortada et al., "High-throughput deeply-etched scanning Michelson
interferometer on-chip," in 2014 International Conference on Optical
MEMS and Nanophotonics, 2014, pp. 161 -- 162.
[15] Y. M. Sabry, H. Omran, and D. Khalil, "Intrinsic improvement of
diffraction-limited resolution
in optical MEMS fourier-transform
spectrometers," in 2014 31st National Radio Science Conference (NRSC),
2014, pp. 326 -- 333.
[16] Y. M. Sabry, D. Khalil, and T. Bourouina, "Monolithic silicon-
micromachined free-space optical interferometers onchip," Laser Photon.
Rev., vol. 9, no. 1, pp. 1 -- 24, 2015.
[17] Y. M. Sabry, D. Khalil, B. Saadany, and T. Bourouina, "Curved Silicon
Micromirror for Linear Displacement-to-Angle Conversion With
Uniform Spot Size," IEEE J. Sel. Top. Quantum Electron., vol. 21, no. 4,
pp. 165 -- 173, Jul. 2015.
[18] M. Erfan, Y. M. Sabry, M. Sakr, B. Mortada, M. Medhat, and D. Khalil,
"On-Chip Micro -- Electro -- Mechanical System Fourier Transform Infrared
(MEMS FT-IR) Spectrometer-Based Gas Sensing," Appl. Spectrosc., vol.
70, no. 5, pp. 897 -- 904, 2016.
[19] B. Mortada, M. Erfan, M. Medhat, Y. M. Sabry, B. Saadany, and D.
Khalil, "Wideband Optical MEMS Interferometer Enabled by Multimode
Interference Waveguides," J. Light. Technol., vol. 34, no. 9, pp. 2145 --
2151, May 2016.
[20] M. Medhat et al., "Selective step coverage for micro-fabricated
Copyright (c) 2018 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].
|
1906.06235 | 1 | 1906 | 2019-06-14T14:56:34 | Nonstoichiometric titanium dioxide nanotubes with enhanced catalytical activity under visible light | [
"physics.app-ph",
"cond-mat.mes-hall",
"cond-mat.mtrl-sci",
"physics.chem-ph"
] | The catalytic activity of nanotubular titanium dioxide films formed during the oxidation of acetone to carbon dioxide under the action of visible light with a wavelength of 450 nm was found to be approximately 2 times higher compared to standard titanium dioxide (Degussa P25). The nanotubular films were grown by the anodization of titanium foil using an original technique. Diffuse reflectance spectra of the films are attributed to enhanced activity in the visible spectrum by the nonstoichiometry of titanium dioxide near the interface between the nanotubular film and the titanium foil substrate. | physics.app-ph | physics | Received: 6 March 2018
Accepted: 15 June 2018
Published: xx xx xxxx
OPEN
Nonstoichiometric titanium
dioxide nanotubes with enhanced
catalytical activity under visible
light
A. A. Valeeva
A. A. Saraev3, I. A. Weinstein2 & A. A. Rempel1,2
1,2, E. A. Kozlova3, A. S. Vokhmintsev2, R. V. Kamalov2, I. B. Dorosheva1,2,
The catalytic activity of nanotubular titanium dioxide films formed during the oxidation of acetone
to carbon dioxide under the action of visible light with a wavelength of 450 nm was found to be
approximately 2 times higher compared to standard titanium dioxide (Degussa P25). The nanotubular
films were grown by the anodization of titanium foil using an original technique. Diffuse reflectance
spectra of the films are attributed to enhanced activity in the visible spectrum by the nonstoichiometry
of titanium dioxide near the interface between the nanotubular film and the titanium foil substrate.
Nanostructural modifications of titanium dioxide are of great research interest, particularly in the development
of promising functional media for renewable energy sources (i.e., solar cells and the photochemical decomposi-
tion of water)1 -- 4, selective organic synthesis with atomic precision for direct C -- H functionalization of a variety
of organic molecules5, the manufacture of efficient photoelectric transducers and memristor memory cells6, the
development of photocatalysts for the removal of organic impurities7 -- 9, and other applications10 -- 12. The optical
gap width of stoichiometric TiO2 (>3.1 eV) determines its photocatalytic activity under the action of near -UV
radiation, which constitutes only a small percent of the solar spectrum13. To shift the spectral response to the
visible region and enhance the catalytic activity, it is reasonable to decrease the optical gap width of the material,
in particular, from the generation of structural vacancies in the oxygen sublattice12,14.
The most promising possibility for generating oxygen vacancies in titanium dioxide is the growth of nano-
films from titanium foil, as has been achieved previously6,15,16 or by sonoelectrochemical method17. In this case,
deficiencies in oxygen can appear in the vicinity of the nonstoichiometric contact between titanium dioxide and
titanium. Some indications of this phenomenon are apparent in the results of the photocatalytic activity of nano-
tubular films. Indeed, an enhancement of activity by 18% in comparison to Degussa P25 under the action of UV
light was registered17. In16, a much higher activity level of nanotubes on Ti foil compared to than on Degussa P25
was also observed under visible light.
Therefore, the aim of the present work is to consider the growth of the nanotubular titanium dioxide film with
nonstoichiometry near the interface with titanium foil in detail, and to study the photocatalytic activity of such
films under visible light by a reliable technique -- the oxidation of acetone to CO2.
Results and Discussion
Figures 1 -- 4 show SEM images of the surface and a side view of the grown films as a function of variation of the
anodization time. A morphological analysis of the resulting films made it possible to deduce the dependences of
geometric parameters, such as the effective surface and solid fraction of the nanotubular titanium dioxide layer
as a function of anodization time.
According to18, the effective surface Seff and roughness factor H are described by equations (1) and (2)
S
eff
= ⋅
H S
,
film
(1)
1Institute of Solid State Chemistry, Ural Branch of the RAS, Yekaterinburg, Russia. 2NANOTECH Centre, Ural Federal
University, Yekaterinburg, Russia. 3Boreskov Institute of Catalysis, Siberian Branch of the RAS, Novosibirsk, Russia.
Correspondence and requests for materials should be addressed to A.A.V. (email: [email protected])
1
SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1www.nature.com/scientificreports Figure 1. SEM images of the sample anodized for 15 min: (a) Surface of the sample; (b) side view of the sample.
H
=
8 LR
π
2
2R
−
,
+
2
y)
(2)
,D
where Sfilm is the area of anodization on which the nanotubular film was grown, =
in L is the
2
nanotube (NT) length, w is the thickness of the NT wall, Din is the inner diameter of the NT, and y denotes the
voids between nanotubes.
=R
1
w ,
+
R
R
1
2
1
1
2
2
The solid fraction SF of the nanotubular film was calculated using equation (3), as given by18
3(4R
SF
π=
4
(
2
D
out
−
2
D
in
)
2
−
,
2
R
+
y
2
)
3(4
R
1
2
(3)
where Dout is the outer diameter of a nanotube. According to an analysis of the morphology of the films grown at
different anodization times, from 15 minutes to 6 hours, nanotube lengths range from 500 nm to 2 μm, the value
of the inner diameter Din of nanotubes varies from 30 to 40 nm, and the outer diameter Dout varies from 45 to
55 nm. It should be noted that at a certain time points, the growth of the oxide layer slows down considerably
because the current density decreases; hence, dissolution of the oxide layer becomes more intense than oxidation.
In the present study, this time point was 2 hours. After this time of anodization (tA = 120 min), an effective surface
for the grown array of nanotubes reaches a maximum (Seff = 415 cm2) at this anodization time, and a maximum
length of the nanotubes of about L = 2 μm was also achieved. According to equitation (3), the solid fraction SF of
nanotubes at an anodization time of 120 min reaches a maximum value approximately 0.41.
Therefore, an optimal anodization time to grow ordered nanotubular TiO2 layers with maximum values of the
L and Seff parameters under the conditions indicated above is 120 min, which is necessary for enhancing catalytic
activity.
Figure 5 depicts the XRD pattern of the nanotubular titanium dioxide film on titanium foil that was grown by
anodization for 120 min. At small 2θ angles (from 18 to 33°), the intensity of the signal increased, which manifests
as a diffuse halo with a maximum marked by an arrow. In addition, Fig. 5 displays the XRD pattern of the initial
titanium foil on which the nanotubular titanium dioxide layer was grown by anodization. The intense diffuse halo
and the absence of diffraction peaks clearly demonstrates that the nanotubular titanium dioxide layer is amor-
phous. This result agrees satisfactorily with independent studies19 -- 21.
Figure 6 shows diffuse reflectance spectra DRS of the nanotubular TiO2 layer and the Degussa P25 TiO2 nan-
opowder. In the visible spectrum at approximately 450 nm, there is a broad reduction in diffuse reflection, which
is related to the nonstoichiometry of amorphous titanium dioxide. The presence of nonstoichiometry is supported
2
www.nature.com/scientificreports/SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1 Figure 2. SEM images of the sample anodized for 60 min: (a) Surface of the sample; (b) Side view of the sample.
by thermodynamic calculations and experimental studies using energy dispersive X-ray spectroscopy19. Atomic
defects represented by vacancies in the amorphous network lead to the formation of energy levels in the band gap of
titanium dioxide, which in its turn results in the absorption of visible light with a wavelength near 450 nm, shown by
an arrow in Fig. 6. This absorption exerts a pronounced effect on the catalytic activity of the film in the visible region
including the wavelength at a 450 nm, which was used in the present work for measurements of catalytic activity.
In addition, the analysis of DRS of the nanotubular TiO2 layer that was anodized for 120 min allowed us to
determine the optical gap width by the following technique. It was assumed that the relationship between the
band gap width of a semiconductor Eg and the optical absorption coefficient α has a Tauc function form21 -- 23, as
given by equation (4)
(
hv
1/α
)
n
=
A hv
(
−
E
g
),
(4)
where h is the Planck constant, ν is the frequency of electromagnetic waves, and A is a constant. In the case of
nanotubular TiO2 layers, the exponent n = 2 for indirect allowed transitions, and n = ½ for direct ones. To find
the band gap width, the diffuse reflectance spectrum was converted into a Kubelka-Munk function24,25, as given
by equation (5).
(5)
where R∞ is the diffuse reflection of the oxide layer relative to a white body, and it is known that F(R∞) is propor-
tional to the coefficient λ25. In this situation, equation (4) takes the form of equation (6).
F R
(
)
∞
=
(1
2
)
∞
R
−
R
2
∞
(
h F R
(
ν
n
1/
))
=
A h
(
ν
−
E
g
)
∞
(6)
Figure 7 shows the Kubelka-Munk function for indirect allowed transitions and the approximation of the
dependence by linear functions for a nanotubular TiO2 layer (top) and the Degussa P25 TiO2 nanopowder (bot-
tom). The extrapolation with a linear function to zero absorption in the region of high photon energies shows that
the maximum band gap width for a nanotubular TiO2 layer anodized for 120 min is Eg = 3.3 eV, while Eg = 3.0 eV
for the Degussa P25 TiO2 nanopowder. An analysis of the Kubelka-Munk function in the region of low photon
energies using a procedure reported in26 revealed that the nanotubular layer contains not only a titanium dioxide
phase with a wide band gap, but also the nonstoichiometric phases have an essentially narrower optical gap width,
which results in the absorption of visible light, in particular with the wavelength of 450 nm (see Fig. 7, top).
3
www.nature.com/scientificreports/SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1 Figure 3. SEM images of the sample anodized for 120 min: (a) Surface of the sample; (b) Side view of the
sample.
A comparison of Kubelka-Munk functions for the nanotubular amorphous film and the Degussa P25 titanium
dioxide nanopowder corroborates the model of indirect transitions in the semiconductor nanotubular film. Thus,
the amorphous structure of the nanotubular titanium dioxide does not produce a switch from an indirect transi-
tion (crystalline titanium dioxide) to direct one (amorphous titanium dioxide).
The Table 1 lists mean data for the catalytic activity of three different nanotubular TiO2 films prepared by ano-
dization under similar conditions for 120 min. The length of the nanotubes synthesized by this method within this
time vary from 1.4 to 2 μm, and are on average 1.7 μm; however, all nanotubular layers have a similar morphology.
The typical morphology is shown on Fig. 3.
The weight (m) of nanotubes linearly depends on their length (L) according to the formula given by equation (7)
ρ=m
SF S
L,
film
(7)
where ρ is the density of the material. According to XRD data, the material in the present case is amorphous tita-
nium dioxide with a density estimated at 3.6 g/cm3 27. Therefore, the weight of nanotubes array on the area of
Sfilm = 3.8 cm2 is equal to 1.1 mg at a length of 2 μm.
The average catalytic activity measured for titanium dioxide nanotubes is approximately 4.2 ± 0.8 μmol/
(min·g), which is more than 2 times higher compared to standard titanium dioxide (Degussa P25) (see Table 1).
An analysis of the diffuse reflectance spectra of the films suggests that such a pronounced enhancement of activity
are result from more efficient absorption of visible light of the films due to narrowing of the optical gap because of
the nonstoichiometry of the film near the interface with the metal substrate. It should be noted that enhancement
of the activity by 110% cannot be explained only by the 35% greater specific surface area of the nanotubular TiO2
(74 ± 8 m2/g) in comparison to standard P25 TiO2 (55 ± 5 m2/g).
A numerical comparison of the catalytic activity of different substrates (see Table 1) shows that those grown in
this work are pure titanium dioxide nanotubes, which were produced more effectively than those manufactured
from better-known but similar materials, and some more complicated materials. Only a few more complicated
and expensive materials (e.g. 10% UO2(NO3)2/TiO2) are more active for oxidizing acetone, and (Au@Ag)@Au/
TiO2 and Au@Ag/TiO2 are more active for oxidizing a 2-propanol substrate (see Table 1 and references therein).
A scheme that considers the expected band diagram of photooxidation is shown on Fig. 8. Impurity levels are
electron acceptors. Energies of levels are shown on the scheme. The optical gap width of the stoichiometric TiO2
is approximately 3.3 eV, and for nonstoichiometric titanium oxide (TiO2−x) is approximately 2.8 eV. It is observed
that more efficient absorption of visible light owing to an oxygen vacancy, leads to a pronounced enhancement of
activity and the optical gap width decreases.
4
www.nature.com/scientificreports/SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1 Figure 4. SEM images of the sample anodized for 360 min: (a) Surface of the sample; (b) Side view of the
sample.
Figure 5. XRD pattern of the nanotubular TiO2 film with the diffuse reflection maximum marked by an arrow.
The XRD pattern of the initial titanium foil is displayed for comparison.
A Ti2p core-level spectrum of the surface of the film is presented on Fig. 9. The spectrum is described by
three doublets Ti2p3/2 -Ti2p1/2, corresponding to Ti in the different oxidation forms. The intense doublet Ti2p3/2-
Ti2p1/2 with a Ti2p3/2 peak binding energy at 459.0 eV corresponds to Ti4+ in the structure of TiO2. The second
doublet Ti2p3/2 -Ti2p1/2 with a Ti2p3/2 peak binding energy at 457.1 eV corresponds to Ti3+ in the defective struc-
ture of nonstoichiometric TiO2−x. The low intense doublet Ti2p3/2 -Ti2p1/2 with a Ti2p3/2 peak binding energy
at 453.6 eV corresponds to Ti° in the structure of titanium foil substrate. In the literature, the values of binding
energy of Ti2p3/2 in TiO2 are given in range of 458.7 -- 459.2 eV, while the Ti3+ is in the range of 456.2 -- 457.4 eV28 -- 30.
Therefore, XPS spectra of the films allow us to identify that the oxygen deficiency and the presence of Ti3+ con-
tribute to nonstoichiometry in the titanium dioxide nanotubes31.
5
www.nature.com/scientificreports/SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1 Figure 6. DRS of a nanotubular TiO2 layer and nanostructured TiO2 (Degussa P25). An arrow indicates a
broad lowering of diffuse reflection in the visible spectrum near 450 nm.
Figure 7. The Kubelka-Munk absorption curve for an indirect allowed transition (n = 2) for a nanotubular
TiO2 layer and Degussa P25 TiO2 nanopowder. Approximations of the slope by linear functions in different
regions of the absorption curve are shown. The numbers indicate energies of the optical gap width that were
obtained by extrapolation to zero absorption. An arrow (450 nm) marks the energy corresponding to the
wavelength of the excitation light.
Conclusions
In the present work, the nanotubular films with a nonstoichiometric layer in the vicinity of titanium foil were
grown by the anodization of titanium foil using an original technique. The highest activity of approximately
4.2 ± 0.8 μmol/(min·g), which is nearly twofold greater compared to standard titanium dioxide (Degussa P25),
was observed for nanotubular films prepared by anodization for a period of 120 min. Such a pronounced enhance-
ment of activity may be attributed to a more efficient absorption of visible light by the films due to narrowing of
the optical gap because of difference in the nonstoichiometry of titanium dioxide near the interface between
6
www.nature.com/scientificreports/SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1 Photocatalyst
Organic compounds mineralization
Substrate
T, °C Light source
Cut-off filter
W (СО2)/due,
μmol g−1 min−1
W(СО2)/due,
μmol min−1
Reference
TiO2 nanotubes
Degussa P25 TiO2
Kronos vlp7000
CQDs/Bi2WO6
Bi2WO6
10% UO2(NO3)2/TiO2
5% UO2(NO3)2/TiO2
(Au@Ag)@Au/TiO2
Au@Ag/TiO2
BiO(ClBr)0.375I0.25
BiO(ClBr)0.5
Dye degradation
WO3@TiO2-nanotubes@WO3
C-TiO2 nanotubes
TiO2 nanotubes
ZnFe2O4-TiO2 nanotubes
GO-Ag-TiO2 nanotubes
TiO2 nanotubes
TiO2 nanotubes
TiO2 nanotubes
TiO2 nanotubes
GR-TiO2 nanotubes
40
450 nm LED, 4.5 mW/cm2
acetone
2-propanol
RhB
AOII
MB
25
40
25
25
25
20
25
30
25
Orange G
25
AO7
20
MB
Alachlor (herbicide) 25
500 W Xe lamp, 400 mW/cm2
1000 W Xe lamp, 17 mW/cm2
>420 nm
500 W Xe lamp, 30 mW/cm2
>440 nm
500 W Xe lamp, 30.5 mW/cm2
400 nm < λ < 800 nm
Xe lamp with the intensity of 250 W at 420 nm
500 W tungsten-halogen lamp >420 nm
1000 W Xe-Hg lamp
500 W tungsten-halogen lamp
500 W tungsten-halogen lamp
Ultraviolet lamp (λ = 365 nm)
Ultraviolet lamp (λ = 254 nm)
Ultraviolet lamp (λ = 325 nm)
Ultraviolet lamp (λ = 325 nm)
Xe lamp
>400 nm
--
--
4.2 ± 0.8
2.0 ± 0.2
3.9 ± 0.4
0.4
0.2
7.5
3.0
13
7.0
0.08
0.04
0.3
0.4
0.1
--
--
--
2.4
--
--
0.005 for
1.1 mg
0.018 for 9.0 mg
0.035 for 9.0 mg
0.008
0.004
0.113
0.045
0.002
0.001
0.008
0.003
0.017
0.023
0.0008
0.0006
0.002
4∙10−6
--
0.0006
0.0007
0.1
Present study
32
33
34
35
36
37
38
39
40
16
17
41
42
43
Table 1. A numerical comparison of previously published catalytic activities with data obtained from the
present study.
nanotubular film and the titanium foil substrate. Oxygen deficiency in the titanium dioxide nanotubes are proven
by XPS measurements.
Methods
Nanotubular titanium dioxide films were synthesized on a Digma setup consisting of an electrochemical cell,
thermostat and power supply6. Anodic oxidation was carried out using a substrate made of 100 μm thick titanium
foil. The surface of the titanium foil was preliminarily treated with acetone and an aqueous solution of acids
mixed in a volumetric ratio of HNO3:HF:H2O = 6:1:18, and was then washed with distilled water. Primary anodi-
zation was performed in a 1% solution of hydrofluoric acid in ethylene glycol for 10 min at a temperature of 20 °C.
This anodization was followed by etching of the titanium substrate in a solution of acids (HNO3:HF:H2O = 1:1:20)
for 3 min, washing in distilled water and acetone, and finally drying at room temperature. Secondary anodization
was carried out potentiostatically at a voltage of 20 V and anodization times of tA = 15, 30, 60, 120, 180 or 360 min.
A solution of ethylene glycol and ammonium fluoride with a concentration equal to 1 wt. % was employed as the
electrolyte. All the chemical reagents used in the study were of analytical grade. The area of primary and second-
ary anodization for every sample was approximately 3.8 cm2.
XRD analysis of the films and substrate was performed using CuKα1,2 radiation on a Shimadzu XRD-7000
(Japan) diffractometer with the Bragg-Brentano recording geometry. XRD patterns were measured using the
step-scan mode at Δ(2θ) = 0.02° in the 2θ angular range from 10 to 100° with a long exposure time at each step.
The surface and side views of the synthesized samples were examined on a SIGMA VP (Carl Zeiss) scanning elec-
tron microscope (SEM) under high vacuum using an InLens detector. The obtained SEM images were processed
with Clinker C7 software (SIAMS).
DRS measurements of the films and standard were recorded on a Lambda 35 (Perkin Elmer, USA) UV-VIS
spectrophotometer with a RSA-PE-20 (Labsphere) diffuse reflectance attachment.
The XPS measurements were performed on a photoelectron spectrometer (SPECS Surface Nano Analysis
GmbH, Germany) equipped with a PHOIBOS-150 hemispherical electron energy analyzer, a FOCUS-500 X-ray
monochromator, and a XR-50 M X-ray source with a double Al/Ag anode. The core-level spectra were obtained
using monochromatic Al Kα radiation (hν = 1486.74 eV) and a fixed analyser pass energy of 20 eV under
ultra-high-vacuum conditions. The charge correction was performed by setting the Ti2p3/2 peak at 459.0 eV, cor-
responding to titanium in TiO2. For detailed analysis, the spectra were fitted into several peaks after background
subtraction using the Shirley method. The fitting procedure was performed using CasaXPS software. The line
shapes were approximated by the product of Gaussian and Lorentz functions.
Photocatalytic activity of the films and the standard was measured in a flow reactor, and reactants were
identified using an FTIR spectrometer. Standards -- (titanium dioxide Degussa P25 and Kronos vlp7000) were
7
www.nature.com/scientificreports/SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1 Figure 8. A scheme considering the expected band diagram of photooxidation.
Figure 9. Ti2p core-level spectra of the surface of the film. There are Ti3+ ions with Ti2p3/2 binding energy at
457.1 eV, in addition to Ti4+ (Ebin = 459.0 eV).
uniformly deposited on a square glass slides (S = 9 cm2) from aqueous suspensions, so that the surface density
was 1 mg/cm2. The study was performed using a light-emitting diode with a maximum wavelength of 450 nm.
The other reaction condition were as follows: T = 40 °C, relative humidity of 20%, an acetone concentration of 800
ppm, and a flow rate of 60 mL/min. Oxidation was performed using the oxygen in air.
The concentration of acetone was found by integrating the area of the absorption band at 1160 -- 1265 cm−1, and
the concentration of carbon dioxide was found by integrating the area of the absorption band at 2200 -- 2450 cm−1.
The rate of photocatalytic oxidation was calculated from the accumulation of carbon dioxide, which was a sole
product of the reaction:
C H O 4 O
2
+
3
6
→
3 CO
2
+
3 H O
2
.
As CO2 is the sole product of acetone photooxidation, the rate of acetone decomposition was one third of the
rate of carbon dioxide formation rate.
8
www.nature.com/scientificreports/SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1 References
1. Higashimoto, S., Suetsugu, N., Azuma, M., Ohue, H. & Sakata, Y. Efficient and selective oxidation of benzylic alcohol by O2 into
corresponding aldehydes on a TiO2 photocatalyst under visible light irradiation: effect of phenyl-ring substitution on the
photocatalytic activity. J. of Catalysis 276, 76 -- 83 (2010).
2. Higashimoto, S. et al. Selective photocatalytic oxidation of benzyl alcohol and its derivatives into corresponding aldehydes by
molecular oxygen on titanium dioxide under visible light irradiation. J. of Catalysis 266, 279 -- 285 (2009).
3. Kozlova, E. A. & Parmon, V. N. Heterogeneous Photocatalysts for the Hydrogen production from Aqueous Solutions of Electron
4. Recatalá, D. et al. Photogeneration of hydrogen from water by hybrid molybdenum sulfide clusters immobilized on titania. Chem.
Donors. Russ. Chem. Rev. 86, 870 -- 906 (2017).
Sus. Chem 8, 148 -- 157 (2015).
5. Utepova, I. A., Trestova, M. A., Chupakhin, O. N., Charushin, V. N. & Rempel, A. A. Aerobic oxidative C-H/C-H coupling of
azaaromatics with indoles and pyrroles in the presence of TiO2 as a photocatalyst. Green Chemistry 17, 4401 -- 4410 (2015).
6. Vokhmintsev, A. S., Weinstein, I. A., Kamalov, R. V. & Dorosheva, I. B. Memristive effect in a nanotubular layer of anodized titanium
dioxide. Bulletin of the Russian Academy of Sciences: Physics 78, 932 -- 935 (2014).
7. Rempel, A. A. et al. Synthesis and solar light catalytic properties of titania-cadmium sulfide hybrid nanostructures. Catalysis
Communications 68, 61 -- 65 (2015).
8. Vorontsov, A. V. et al. Kinet. Catal. 51, 801 -- 808 (2010).
9. Kozlova, E. A. & Vorontsov, A. V. Influence of mesoporous and platinum modified titanium dioxide preparation methods on
photocatalytic activity in liquid and gas phase. Appl. Catal. B 77, 35 -- 45 (2007).
10. Kitano, S., Hashimoto, K. & Kominami, H. Photocatalytic mineralization of volatile organic compounds over commercial
titanium(IV) oxide modified with rhodium(III) ion under visible light irradiation and correlation between physical properties and
photocatalytic activity. Catalysis Today 164, 404 -- 409 (2011).
11. Roy, P., Berger, S. & Schmuki, P. TiO2 nanotubes: synthesis and applications. Angewandte Chemie - International Edition 50,
12. Rempel, A. A. Hybrid nanoparticles based on sulfides, oxides, and carbides. Russian Chemical Bulletin 4, 857 -- 868 (2013).
13. Carp, O., Huisman, C. L. & Reller, A. Photoinduced reactivity of titanium dioxide. Progress in Solid State Chemistry 32, 33 -- 177
2904 -- 2939 (2011).
(2004).
14. Zainullina, V. M., Zhukov, V. P. & Korotin, M. A. Influence of oxygen nonstoichiometry and doping with 2p-, 3p-, 6p- and
3d-elements on electronic structure, optical properties and photocatalytic activity of rutile and anatase: ab inito approaches. Journal
of Photochemistry and Photobiology C: Photochemistry Reviews 22, 58 -- 83 (2015).
15. Zhang, M. et al. Temperature-dependent differences in wettability and photocatalysis of TiO2 nanotube arrays thin films. Applied
Surface Science 356, 546 -- 552 (2015).
16. Li, L., Ya, J., Hu, F., Liu, Z. & Lei, E. Photocatalytic decompositions of gaseous HCHO and methylene blue with highly ordered TiO2
nanotube arrays. Journal of Energy Chemistry 25, 740 -- 746 (2016).
17. Tekin, D. Photocatalytic degradation of textile dyestuffs using TiO2 nanotubes prepared by sonoelectrochemical method. Applied
18. Alivov, Y. & Fan, Z. A. TiO2 nanostructure transformation: from ordered nanotubes to nanoparticles. Nanotechnology 20, 1 -- 6
Surface Science 318, 132 -- 136 (2014).
(2009).
19. Chen, C.-C., Say, W. C., Hsieh, S.-J. & Diau, E. W.-G. A mechanism for the formation of annealed compact oxide layers at the
interface between anodic titania nanotube arrays and Ti foil. Applied Physics A: Materials Science and Processing 95, 889 -- 898 (2009).
20. Macak, J. M. et al. TiO2 nanotubes: Self-organized electrochemical formation, properties and applications. Current Opinion in Solid
State and Materials Science 11, 3 -- 18 (2007).
Electrochimica Acta 190, 744 -- 752 (2016).
21. Sopha, H. et al. Self-organized anodic TiO2 nanotube layers: Influence of the Ti substrate on nanotube growth and dimensions.
22. Tauc, J., Grigorovici, R. & Vancu, A. Optical Properties and Electronic Structure of Amorphous Germanium. Phys. Status Solidi 15,
627 -- 637 (1966).
(1963).
23. Tauc, J. (ed. Abeles, F.), Optical Properties of Solids, North-Holland, Amsterdam, 277 (1972).
24. Kubelka, P. & Munk, F. Ein Beitrag Zur Optik Der Farbanstriche. Techn. Phys 12, 593 -- 601 (1931).
25. Kortum, G., Braun, W. & Herzog, G. Principles and Techniques of Diffuse‐Reflectance Spectroscopy. Angew. Chem 2, 333 -- 341
26. Serpone, N., Lawless, D. & Khairutdinov, R. Size effects on the photophysical properties of colloidal anatase TiO2 particles-size
quantization or direct transitions in this indirect semiconductor. J. Phys. Chem 99, 16646 -- 16654 (1995).
27. Prasai, B., Cai, B., Underwood, M. K., Lewis, J. P. & Drabold, D. A. Properties of amorphous and crystalline titanium dioxide from
first principles. J. of Material Science 47, 7515 -- 7521 (2012).
28. Luan, Z. et al. Incorporation of Titanium into Mesoporous Silica Molecular Sieve SBA-15. Chem. Mater. 11, 3680 -- 3686 (1999).
29. Hasegawa, Y. & Ayame, A. Investigation of oxidation states of titanium in titanium silicalite-1 by X-ray photoelectron spectroscopy.
30. Finetti, P. et al. Core and Valence Band Photoemission Spectroscopy of Well-Ordered Ultrathin TiOx Films on Pt (111). J. Phys.
31. Yu, X., Fan, X., An, L., Li, Z. & Liu, J. Facile synthesis of Ti3+-TiO2 mesocrystals for efficient visible-light photocatalysis. Journal of
32. Qian, X. et al. Carbon quantum dots decorated Bi2WO6 nanocomposite with enhanced photocatalytic oxidation activity for VOCs.
Catal. Today 71, 177 -- 187 (2001).
Chem. C 111, 869 -- 876 (2007).
Physics and Chemistry of Solids 119, 94 -- 99 (2018).
Appl. Catal. B 193, 16 -- 21 (2016).
33. Kolinko, P. A., Filippov, T. N., Kozlov, D. V. & Parmon, V. N. Ethanol vapor photocatalytic oxidation with uranyl modified titania
under visible light: Comparison with silica and alumina. J. Photochem. Photobiol. A 250, 72 -- 77 (2012).
34. Kamimura, S. et al. (Au@Ag)@Au double shell nanoparticles loaded on rutile TiO2 for photocatalytic decomposition of 2-propanol
under visible light irradiation. Appl. Catal. B 180, 255 -- 262 (2016).
35. Liu, G. et al. Band-structure-controlled BiO(ClBr)(1−x)/2Ix solid solutions for visible-light photocatalysis. J. Mater. Chem. A 3,
36. Lu, B., Li, X., Wang, T., Xie, E. & Xu, Z. WO3 nanoparticles decorated on both sidewalls of highly porous TiO2 nanotubes to improve
UV and visible-light photocatalysis. J. Mater. Chem. A 1, 3900 -- 3906 (2013).
37. Pan, J. et al. Preparation of carbon quantum dots/ TiO2 nanotubes composites and their visible light catalytic applications. J. Mater.
Chem. A 2, 18082 -- 18086 (2014).
38. Alsawat, M., Altalhi, T., Shapter, J. G. & Losic, D. Influence of dimensions, inter-distance and crystallinity of titania nanotubes
(TNTs) on their photocatalytic activity. Catal. Sci. Technol. 4, 2091 -- 2098 (2014).
39. Wang, M. et al. A facile hydrothermal deposition of ZnFe2O4 nanoparticles on TiO2nanotube arrays for enhanced visible light
photocatalytic activity. J. Mater. Chem. A 1, 12082 -- 12087 (2013).
40. Sim, L. C., Leong, K. H., Ibrahim, S. & Saravanan, P. Graphene oxide and Ag engulfed TiO2 nanotube arrays for enhanced electron
mobility and visible-light-driven photocatalytic performance. J. Mater. Chem. A 2, 5315 -- 5322 (2014).
8123 -- 8132 (2015).
9
www.nature.com/scientificreports/SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1 41. Hahn, R., Stark, M., Killian, M. S. & Schmuki, P. Photocatalytic properties of in situ doped TiO2-nanotubes grown by rapid
breakdown anodization. Catal. Sci. Technol. 3, 1765 -- 1770 (2013).
42. Macak, J. M., Zlamal, M., Krysa, J. & Schmuki, P. Self-Organized TiO2 Nanotube Layers as Highly Efficient Photocatalysts. Small 3,
300 -- 304 (2007).
43. Zheng, D. et al. Preparation of graphene/TiO2 nanotube array photoelectrodes and their photocatalytic activity for the degradation
of alachlor. Catal. Sci. Technol. 6, 1892 -- 1902 (2016).
Acknowledgements
This work was partially financially supported by the Russian Foundation for Basic Research (project No. 17-03-
00702) and by Act 211 Government of the Russian Federation, contract No. 02.A03.21.0006. A.S.V. and I.A.W.
thank Minobrnauki initiative research project № 16.5186.2017/8.9 for support.
Author Contributions
A.A.R. and I.A.W. designed the research. A.A.V., I.B.D., R.V.K. prepared the samples. A.A.V., E.A.K., A.S.V.,
A.A.S. performed the experiments. All authors wrote, read and approved the final manuscript.
Additional Information
Competing Interests: The authors declare no competing interests.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. The images or other third party material in this
article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article's Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
© The Author(s) 2018
10
www.nature.com/scientificreports/SCIENTIFIC RepoRts (2018) 8:9607 DOI:10.1038/s41598-018-28045-1 |
1903.06021 | 1 | 1903 | 2019-03-14T14:18:48 | Cryogenic characterization of 28nm FD-SOI ring oscillators with energy efficiency optimization | [
"physics.app-ph",
"cond-mat.mes-hall"
] | Extensive electrical characterization of ring oscillators (ROs) made in high-$\kappa$ metal gate 28nm Fully-Depleted Silicon-on- Insulator (FD-SOI) technology is presented for a set of temperatures between 296 and 4.3K. First, delay per stage ($\tau_P$), static current ($I_{STAT}$), and dynamic current ($I_{DYN}$) are analyzed for the case of the increase of threshold voltage ($V_{TH}$) observed at low temperature. Then, the same analysis is performed by compensating $V_{TH}$ to a constant, temperature independent value through forward body-biasing (FBB). Energy efficiency optimization is proposed for different supply voltages ($V_{DD}$) in order to find an optimal operating point combining both high RO frequencies and low power dissipation. We show that the Energy-Delay product ($EDP$) can be significantly reduced at low temperature by applying a forward body bias voltage ($V_{FBB}$). We demonstrate that outstanding performance of RO in terms of speed ($\tau_P$=37ps) and static power (7nA/stage) can be achieved at 4.3K with $V_{DD}$ reduced down to 0.325V. | physics.app-ph | physics | © 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including
reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or
reuse of any copyrighted component of this work in other works. DOI: 10.1109/TED.2018.2859636
1
Cryogenic characterization of 28nm FD-SOI ring
oscillators with energy efficiency optimization
H. Bohuslavskyi, S. Barraud, V. Barral, M. Cassé, L. Le Guevel, L. Hutin, B. Bertrand, A. Crippa, X. Jehl,
G. Pillonnet, L. Jansen, F. Arnaud, P. Galy, R. Maurand, S. De Franceschi, M. Sanquer, and M. Vinet
the same analysis
Abstract -- Extensive electrical characterization of ring oscillators
(ROs) made in high-κ metal gate 28nm Fully-Depleted Silicon-on-
Insulator (FD-SOI) technology
is presented for a set of
temperatures between 296 and 4.3K. First, delay per stage (τP),
static current (ISTAT), and dynamic current (IDYN) are analyzed for
the case of the increase of threshold voltage (VTH) observed at low
temperature. Then,
is performed by
compensating VTH to a constant, temperature independent value
through
(FBB). Energy efficiency
optimization is proposed for different supply voltages (VDD) in
order to find an optimal operating point combining both high RO
frequencies and low power dissipation. We show that the Energy-
Delay product (EDP) can be significantly reduced at low
temperature by applying a forward body bias voltage (VFBB). We
demonstrate that outstanding performance of RO in terms of
speed (τp=37ps) and static power (7nA/stage) can be achieved at
4.3K with VDD reduced down to 0.325V.
forward body-biasing
Index Terms -- Cryogenic
oscillator, back-biasing, quantum computing, ultra-low-power.
electronics, 28nm FD-SOI,
ring
I.
INTRODUCTION
S
ince the famous proposal for quantum computing with
quantum dots [1], significant progress in Si spin qubits have
been reported [2-5]. The first two-qubit logic gate in
isotopically enriched Si was demonstrated [3] and a foundry-
compatible CMOS SOI platform was used to demonstrate a
hole spin qubit functionality [4-5]. These major achievements
on the basic building block of a quantum computer are paving
the way to the implementation of a large number of qubits
individually
tunable nearest-neighbor
couplings. However, scaling up these systems in complex
quantum
extremely
challenging. In addition to fundamental advances in physical
understanding and material development needed for spin
qubits, a consistent engineering work should be made to
propose a means of controlling, interacting and reading out a
large number of qubits in parallel.
controlled with
architectures
computing
can be
In recent years, hardware interfaces based on advanced
CMOS technologies that operate at cryogenic temperature so as
to ensure proximity to qubits have been proposed and discussed
[6-9]. Bulk Si MOSFET [10-13] and other basic circuits such
as Ring Oscillators (ROs)
[11] and an FPGA (Field-
Programmable Gate Array) [14] were characterized down to
4K. Despite a significant reduction of subthreshold swing (SS)
and improvement of carrier mobility at low temperature, some
limitations have been raised on bulk Si technologies. Non-ideal
kink behavior and hysteresis in the characteristics are induced
by the bulk current generated by impact ionization at the drain
combined with increased resistivity from the freeze-out of
charge carriers [15]. Moreover, the increase of drive current
from the enhanced carrier mobility at low temperature is
partially mitigated by the increase of threshold voltage which
can be hardly compensated by back-biasing in bulk
technologies.
In this work, an alternative to bulk Si CMOS technologies
is proposed in order to provide more flexibility to designers for
optimizing both high-performance and low power cryogenic
electronics. Undoped thin-planar 28nm FD-SOI devices offer
an excellent short-channel electrostatic control, low leakage
current and immunity to random dopant fluctuations. More
flexibility is brought to the circuit through an extremely
effective back-biasing allowing to switch dynamically between
high performance mode (Forward Body Biasing) and ultra-low
leakage mode (Reverse Body Biasing) [16]. We thus propose to
study the low temperature characterization of 28nm FD-SOI
ring oscillators (RO) down to 4.3K. The RO performance in
terms of delay per stage (τP), dynamic (IDYN), and static (ISTAT)
currents is studied from 296 down to 4.3K for VDD ranging
between 0.325 and 1.2V. Cryogenic effects on the RO
performance are investigated with and without forward back-
biasing (FBB) in order to compensate the shift of threshold
voltage (VTH) at
low temperature. In addition, an energy
efficiency optimization using FBB is proposed. We show that
VDD can be reduced down to 0.325V while maintaining an ultra-
low Energy-Delay product (EDP).
is given
The paper is organized as follows: a brief review of 28nm
FD-SOI
temperature
characterization of FD-SOI ring oscillators and their energy
efficiency optimization are discussed in Section III. Finally, the
main conclusions are drawn in Section IV.
II. Low
Section
in
II. BRIEF REVIEW OF 28NM FD-SOI TECHNOLOGY
The GO1 FD-SOI transistors are fabricated with a gate-first
high-κ metal gate using STMicroelectronics
technology
[17-18]. They are processed on 300mm (100) SOI wafers with
a buried oxide (BOX) thickness of 25nm. The equivalent oxide
H. Bohuslavskyi, S. Barraud, V. Barral, M. Cassé, L. Le Guevel, L. Hutin, G.
Pillonnet, B. Bertrand and M. Vinet are with CEA, LETI, Minatec Campus, F-
38054 Grenoble, France. X. Jehl, L. Jansen, A. Crippa, R. Maurand, S. De
Franceschi, and M. Sanquer are with Univ. Grenoble Alpes, CEA, INAC-
PHELIQS, F-38054 Grenoble, France. F. Arnaud and P. Galy are with
STMicroelectronics, 850 rue J. Monnet, 38920 Crolles, France.
E-mail: [email protected]
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
thickness is 1.55nm for n-MOS and 1.7nm for p-MOS. Low-
threshold-voltage (LVT) transistors are used with the flip-well
architecture: N-well (resp. P-well) with p-type (resp. n-type)
back-plane doping for n-MOS (resp. p-MOS).
a)
VENB
in Fig. 1. The RO consists of 101 identical stages together with
an enabling two-way AND gate. The output is fed to a
frequency divider to lower its frequency in the sub-MHz
regime. The parameters and the measurement protocols used in
this work are summarized in Table I.
III. RESULTS AND DISCUSSION
VN-WELL
VP-WELL
VDD
2
LVT Ring Oscillator
LG = 34nm
WNMOS = 420nm
WPMOS = 600nm
IV-1
IV-2
IV-101
VCC
/1024
OUT
FD
NMOS
S
D
PMOS
S
D
VP-WELL
GND
c)
VN-WELL
N+
BOX
BOX
P+
p-type
back-plane
n-type
back-plane
SiO2
SiO2
P-well
SiO2
N-well
b)
PMOS
VIN
VDD
VOUT
NMOS
GND
Fig. 1. (a) Schematic layout of 101-stages ring oscillator with 1024-frequency
divider (FD). (b) Single inverter stage composed of n- and p-MOS. (c)
Illustration of LVT transistors in the flip-well configuration. VCC is the supply
voltage of the clock divider polarized with +1V. VENB controls the AND gate
that enables the oscillations (VENB=VDD was kept during all the measurements).
For more details on the measurement protocol, see Table I.
RING OSCILLATOR AND MOSFET PARAMETERS, UNITS, DESCRIPTION, AND
TABLE I
MEASUREMENT PROTOCOL
Description
RO frequency
RO delay per
Measurement protocol
VENB=VDD, VCC = 1V and
measure OUT.
= 1/(f´2´101´1024).
A 34nm gate length (LG) is considered for both n- and p-MOS
transistors of width WNMOS=420nm and WPMOS=600nm. The
schematic layout of RO in the flip-well configuration is shown
A. RO performance at room temperature
In contrast to usual bulk technology, ultra-thin body and
buried oxide FD-SOI enables an extended body-bias range from
-3V (RBB) up to +3V (FBB), thanks to the thin buried oxide
providing either high-performance or low-power transistors
[20]. In order to compensate the increase of VTH at low-
temperature this work is focused on LVT transistors which
enable a strong improvement in the switching speed thanks to
FBB (body factor of 85mV/V) at the cost of a higher leakage
current. When no input is provided on the AND logic gate, ISTAT
and then static power consumption occur due to the flow of
leakage from supply to ground. For the room temperature data
in Fig. 2, the static current of 101-stages RO is plotted as a
function of the delay per stage for different VDD and FBB
voltages (VFBB). Let us note that the delay of CMOS invertor
can be approximated by tP = CLOAD×VDD/IEFF [21] with IEFF the
effective current (defined in Table I) and CLOAD the load
capacitance including the inversion capacitance, the parasitic
capacitances and the wiring capacitance of back-end-of-line.
Hence, lowering the supply voltage VDD to achieve lower static
power dissipation (i.e.
the RO
performance giving higher delay per stage. This frequency
reduction can be balanced by using FBB voltage. The efficiency
of forward back biasing for tP reduction and improved
performance is well observed in Fig. 2. For the RO operating at
VDD=1V, a tp reduction of 33% is achieved (at VFBB=2.8V) at
the expense of a significant enhancement of static current (i.e.
static power dissipation). Thus, a compromise in terms of power
dissipation and speed should be carefully considered.
Maintaining the back biasing efficiency at very low temperature
is crucial to ensure accuracy and speed in the control and
readout of qubits while using comparatively less power
dissipation to perform the calculations.
lower ISTAT) degrades
1000
)
e
g
a
t
s
/
A
n
(
T
A
T
S
I
,
t
n
e
r
r
u
c
c
i
t
a
t
S
100
10
1
0.1
VDD = 0.7V
T = 296K
VN-WELL = -VP-WELL = 2.8V
VDD = 0.7V
VN-WELL = -VP-WELL = 1.6V
1.2V
VDD = 0.7V
VN-WELL = VP-WELL = 0V
VDD-step 50mV
10
30
20
Delay/stage, τP (ps)
25
15
35
40
Fig. 2. Static current vs delay per stage for a set of VDD from 0.7 to 1.2V
(DVDD=50mV). Three different FBB voltages are considered. ISTAT is
significantly increased at high VFBB illustrating that ROs are well optimized at
room temperature with VFBB=0V.
Parameter Unit
Hz
ps
f
tP
IDYN
ISTAT
µA/
stage
nA/
stage
stage
per stage in
Dynamic current
oscillating state
Static current per
stage in non-
oscillating state
IEFF
µA
Effective drive
current of FD-SOI
transistor [19]
G(IEFF)
G(tP)
EPT
EDP
%
%
fJ
fJ´ps
IEFF enhancement
temperature
at low
tP enhancement at
low temperature
Energy per
transition
Energy Delay
product
VENB=VDD, VCC = GND and
measure OUT.
VENB = VCC = GND and
measure OUT.
IEFF = 1/(1/ IEFF-N + 1/ IEFF-P)
IEFF-N(P) = (IH+ IL)/2 where
IH = IDS (VGS = VDD,
IL = IDS (VGS = VDD/2,
VD = VDD/2)
VD = VDD)
where VGS, VDS, and IDS are
MOSFET gate voltage, drain
voltage and drain current.
= [IEFF (T) - IEFF (296K)] /
IEFF (T)
= [tP (T) - tP (296K)] / tP (T)
=101 ´ tP ´ (IDYN - ISTAT)
´VDD
= tP ´ EPT
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
3
a)
45
40
)
s
p
(
35
P
τ
,
e
g
a
t
s
/
y
a
l
e
D
30
25
20
15
10
b)
VDD = 0.8V
VN-WELL = 0V
VP-WELL = 0V
VDD = 1V
VDD = 1.2V
)
/
m
µ
A
µ
(
S
D
I
,
t
n
e
r
r
u
c
n
i
a
r
D
100
0
Temperature, T (K)
200
300
1200
1000
800
600
400
200
0
T↓
)
A
(
S
D
I
10-6
10-7
10-8
10-9
-0.4
0
VGS (V)
0.4
c)
10-4
VN-WELL = 0V
VP-WELL = 0V
)
A
(
F
F
E
I
,
t
n
e
r
r
u
c
e
v
i
t
c
e
f
f
E
10-5
10-6
10-7
10-8
VDD = 1.2V
VDD = 1V
PMOS
NMOS
0.5
-1
Gate voltage, VGS (V)
-0.5
0
1
10-9
VDD = 0.8V
100
0
Temperature, T (K)
200
300
Fig. 3. (a) Delay per stage vs temperature for VDD = 0.8, 1, and 1.2V showing the RO slowing down due to the increase of VTH at low temperature. (b) IDS -VGS
curves recorded at VDD = 1V plotted in linear scale. The insert shows the subthreshold regime where current is plotted in logarithmic scale. Green (resp. blue) color
corresponds to room temperature (resp. 4.3K). (c) Effective current vs temperature for different VDD. The procedure used to calculate IEFF is described in Table I.
B. RO performance without FBB down to 4.3K
Operation of FD-SOI transistors at cryogenic temperature
was already reported in [18,22,23]. Since the scattering of
charge carriers with phonons is sufficiently weak and can be
neglected at liquid helium temperature, electron and hole
mobilities are enhanced and should lead to a smaller tp at lower
temperature for a given VDD. However, despite a significant
increase of drive current expected at low temperature [18], the
RO slows down as it can be seen in Fig. 3a. This increase of
delay per stage is explained by the VTH-shift at lower
temperatures (Fig. 3b) for both n- and p-MOS transistors.
Consequently, the effective current is strongly reduced at lower
temperatures due to a lower overdrive current for the gate
voltage. This IEFF reduction becomes especially important at
low VDD. Hence, without FBB (VN-WELL=VP-WELL=0V) the
enhanced carrier mobility at low temperature does not improve
the effective current which becomes strongly limited by the VTH
increase. The IEFF reduction shown on Fig. 3c is then directly
responsible for the observed increase of tP. The degradation of
RO performance becomes even more important at low VDD.
Therefore, without forward-back-biasing used to compensate
the VTH-shift it may be difficult to work with an optimized
cryogenic digital control electronics combining the demands of
both high-performance and low-power consumption.
C. RO performance by applying FBB down to 4.3K
In order to preserve the benefit of higher carrier mobility and
thus, higher driving current at low temperatures, the VTH-shift
should be compensated. The ability to adjust VTH through back-
biasing was already successfully demonstrated down to 4.3K
using 28nm FD-SOI transistors [18]. In this work, the threshold
voltages of n- and p-MOSFET as well as the body-factors
(DVTH/DVFBB) were systematically extracted over a wide range
in Fig. 4c. By monitoring
temperature range (Fig. 4a). By reducing
of VFBB from 0 up to 3V. Then, by adjusting VN-WELL and
VP-WELL for each temperature, the VTH-shift was compensated
(in order to keep the VTH measured at 300K) and thus, VTH_NMOS
and VTH_PMOS were kept constant in temperature as shown in
Fig. 4b. In contrast to previous results (without FBB), a
significant speed up for different VDD is now observed when
threshold voltages were settled to VTH(T)=VTH(296K) within
the full
the
temperature from room temperature down to 4.3K, the delay per
stage is decreased by 38% at VDD=0.8V, 33% at VDD=1V, and
31% at VDD=1.2V. The tP reduction can be explained by the
increase of IEFF shown
the
enhancement of IEFF and tP at lower temperatures (down to
4.3K) we observe a good correlation between the relative
variations of IEFF increase and the tP reduction for VDD=0.8V as
shown in Fig. 5. Hence, the load capacitance seems to be
weakly dependent on temperature since the IEFF increase is
mainly responsible for the speed up of RO at low temperature.
Besides the low temperature behavior of tP, other critical
parameters such as dynamic current (IDYN) and static current
(ISTAT) were monitored for the evaluation of the power
dissipation in oscillating and sleeping mode. As shown in
Fig. 6, IDYN naturally increases as the temperature is reduced
due to the IEFF enhancement, which leads to the increase of
active power dissipation. Indeed, by applying FBB for the VTH
compensation, the effective current is improved due to higher
carrier mobility at low temperature and then IDYN is enhanced.
In comparison to the room temperature case, the IDYN
enhancement normalized by VDD is about 34% at 4.3K. At the
same time a significant reduction in static current (Fig. 6b) and
thus, in static power dissipation is observed thanks to the
decrease of subthreshold slope at low temperature [18].
)
A
(
S
D
I
10-6
10-7
10-8
10-9
b)
1200
)
/
m
µ
A
µ
(
S
D
I
,
t
n
e
r
r
u
c
n
i
a
r
D
1000
800
600
400
200
0
)
s
p
(
P
τ
,
e
g
a
t
s
/
y
a
l
e
D
25
20
15
10
VDD = 0.8V
VDD = 1V
VDD = 1.2V
100
0
Temperature, T (K)
200
300
10-4
)
A
(
F
F
E
I
,
t
n
e
r
r
u
c
e
v
i
t
c
e
f
f
E
10-5
10-6
T↓
-0.3 0
VGS (V)
0.3
T↓
PMOS
NMOS
0.5
-1
Gate voltage, VGS (V)
-0.5
0
1
VDD = 1V
VDD = 0.8V
100
0
Temperature, T (K)
200
300
Fig. 4. (a) Delay per stage vs temperature for VDD = 0.8V, 1V, and 1.2V in the case of compensated VTH. The RO speeds up at low temperature due to the carrier
mobility enhancement. (b) IDS-VGS curves recorded at VDD=1V plotted in linear scale. The insert shows the subthreshold behavior with current plotted in logarithmic
scale. VTH_NMOS and VTH_PMOS within the whole temperature range are constant and can be found in Table II. Green (resp. blue) color corresponds to room temperature
(resp. 4.3K) (c) Effective current vs temperature for different VDD. As it can be seen in Fig. 4b, the effective current increases at any stage of cooling down.
As compared to room temperature, the calculated static power
dissipation PSTAT=ISTAT×VDD at 4.3K is reduced by a factor of
1600 (VDD=0.8V), 100 (VDD=1V), and 6.5 (VDD=1.2V). The
RO performance in terms of tP, IDYN, and ISTAT measured at 4.3
and 296K with and without FBB are summarized in Table II.
VTH(T)=VTH(296K)
VDD = 1.2V
)
e
g
a
t
s
/
A
µ
(
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
a)
c)
VTH(T)=VTH(293K)
4
30
VTH(T)=VTH(296K)
VDD = 1.2V
D. Energy efficiency optimization down to 4.3K.
Wiring up large qubit arrays today is a key issue across all qubit
platforms. Indeed, a large number of qubits must be read out
periodically and rapidly processed to check whether errors
occur along the way and to correct them. Furthermore, the
energy efficiency of a cryogenic platform for the classical
control of a scalable quantum computer is essential. Low power
dissipation is required, down to few watts at 4K [24] to enable
N
Y
D
I
,
t
n
e
r
r
u
c
c
i
m
a
n
y
D
VTH(T)=VTH(296K)
)
e
g
a
t
s
/
A
n
(
VDD = 0.8V
G(τP)
G(IEFF)
50
0
100
200
Temperature, T (K)
150
250
300
Fig. 5. Relative enhancement of IEFF and tp at VDD = 0.8V for temperature
between 296 and 4.3K. FBB is applied to compensate the increase of VTH at low
temperatures. Note that both relative gains are well correlated within the whole
temperature range.
)
%
(
n
i
a
G
60
50
40
30
20
10
0
a)
1.6
1.4
1.2
1.0
0.8
0.6
0.4
b)
1
10-1
10-2
T
A
T
S
I
,
t
n
e
r
r
u
c
c
i
t
a
t
10-3S
VDD = 1V
VDD = 0.8V
0
50
100
200
Temperature, T (K)
150
250
300
VDD = 1.2V
VDD = 1V
VDD = 0.8V
0
50
VTH(T)=VTH(296K)
150
250
300
100
200
Temperature, T (K)
Fig. 6. (a) Dynamic current and (b) static current as a function of temperature
for different VDD in case of compensated VTH. Note that ISTAT continuously
decreases and reaches 2.3 pA/stage for VDD=0.8V at 4.3K.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
5
COMPARISON OF RING OSCILLATOR PERFORMANCE AT 296K AND 4.3K
BETWEEN UNCOMPENSATED AND COMPENSATED VTH
TABLE II
VTH_NMOS
(V)
VTH_PMOS
VN-WELL
(V)
VP-WELL
T (K)
VDD (V)
296
0.8
1
1.2
0.8
4.3
1
1.2
(V)
0.224
-0.376
0.216
-0.363
0.208
-0.352
0.391
-0.549
0.224
-0.376
0.384
-0.531
0.216
-0.363
0.375
-0.518
0.208
-0.352
(V)
0
0
0
0
0
0
0
0
1.65
-2.75
0
0
1.65
-2.65
0
0
1.62
-2.62
tP
(ps)
IDYN
(nA/stage)
ISTAT
(nA/stage)
27
18
14
42
18
22
13
15
10
497
885
1308
363
717
812
1182
1281
1669
0.67
1.29
1.9
0.002
0.002
0.007
0.01
0.037
0.29
the energy dissipation. The
The set of VN-WELL and VP-WELL used for different VDD is chosen to compensate
the corresponding VTH-shift at 4.3K.
the operation of thousands of cryogenic fault-tolerant loops in
existing refrigerators. The adoption of FD-SOI technology to
reduce power consumption while keeping high speed of
operation can be a promising solution in future implementations
of large-scale quantum computers. Then, ensuring that control
signals produce minimal dissipation will be essential at the
lowest temperature stage of the refrigerator. We thus propose to
deal with the energy consumption per transition and more
specifically with the Energy-Delay metric to translate the more
and more stringent constraint on the speed, while not
disregarding
temperature
dependence of Energy per Transition (EPT) for different VDD is
shown in Fig. 7a. As in the previous section, FBB is applied to
correct the VTH-shift at low temperatures. One can notice that
despite an increase of IDYN (shown in Fig. 6a), the significant
decrease of delay per stage at low temperature (Fig. 4a) results
in EPT becoming smaller and smaller during the cooling down
to 4.3K. Lowering VDD at liquid helium temperature also leads
to a significant reduction of EPT at the expense of performance
with a lower RO frequency. To achieve the best possible
performance, the Energy-Delay metric [21, 25] defined in
Table I can be used. Smaller Energy-Delay values imply a
lower energy consumption at the same level of performance
corresponding to a more energy-efficient design. At RT, the
lowest EDP is obtained for VDD=1V as shown on Fig. 7b.
However, as the RO is cooled down, we see that the optimal
VDD can be lowered. At 4.3K, the minimal EDP is now obtained
at VDD=0.8V. The VDD dependence on the Energy-Delay
Product is shown in Fig. 8a at 4.3 and 296K. At 4.3K, FBB is
first used to keep the same VTH as that obtained at room
temperature. Then, higher VFBB was applied in order to
maximize the energy efficiency. As expected, EDP is minimal
for VDD=1V at 296K. However, a strong increase is observed at
low VDD due to the rise of tP as shown on Fig. 3a. At 4.3K, the
VTH-compensation allows to reduce tP (see Fig. 4a) and leads
to the decrease of EDP with a minimal value close to
,
)
J
f
(
T
P
E
2.4
2.2
2.0
1.8
1.6
1.4
1.2
1.0E
n
o
i
t
i
s
n
a
r
t
r
e
p
y
g
r
e
n
b)
)
s
p
×
J
f
(
P
D
E
,
t
c
u
d
o
r
p
y
a
l
e
D
y
g
r
e
n
E
32
28
24
20
VDD=0.8V. However, once again, the EDP remains strongly
enhanced at low VDD. The best FBB configuration minimizing
EPT while keeping high RO speed is obtained for VN-WELL=4V
and VP-WELL=−5.8V. This FBB configuration was chosen such
as no minimum is observed in EDP even at VDD as low as
0.325V. The combination of LVT transistors with high
forward-back-biasing allows to further reduce VTH and then to
obtain lower tP for a given VDD. In Fig. 8b, we show the static
current versus tp for different VDD in order to highlight the
advantage of FBB on LVT FD-SOI transistors. At 4.3K, low
delay operation can be achieved without suffering from
excessive increase of static power. However, if high FBB
voltages (VN-WELL = 4V and VP-WELL= −5.8V) are applied to
reduce VTH close to 0V, ISTAT is significantly increased.
Nevertheless, excellent RO performance with tP=37ps and
ISTAT=7nA/stage are demonstrated at VDD = 0.325V. This static
current can be further reduced by a factor 10 if VP-WELL is
lowered from -5.8 to −4.8V while keeping VN-WELL=4V. In
addition, as shown in Fig. 8c, ultra-low IDYN current of
92nA/stage (at VDD=0.325V) can be of great importance if low
power dissipation in oscillating mode at moderate speed
(tP=71ps) is required. It should be noted that the RO
performance for different FBB configurations can be further
optimized depending on targeted applications.
a)
VTH(T)=VTH(296K)
VDD = 0.8V
VDD = 1V
VDD = 1.2V
0
50
100
200
Temperature, T (K)
150
250
300
VDD = 0.8V
VDD = 1V
VDD = 1.2V
VTH(T)=VTH(296K)
0
50
200
100
Temperature, T (K)
150
Fig. 7. (a) Energy per transition vs temperature for three different VDD. Note
that the decrease of EPT for all VDD at cryogenic temperature is mainly due to
an important tP reduction. (b) Energy-Delay product vs temperature for
VDD=0.8, 1, and 1.2V. The lowest EDP indicates the optimal VDD.
250
300
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
6
a)
80
60
40
20
)
s
p
×
J
f
(
P
D
E
,
t
c
u
d
o
r
p
y
a
l
e
D
y
g
r
e
n
0E
b)
102
101
100
10-1
10-2
10-3
10-4
)
e
g
a
t
s
/
A
n
(
T
A
T
S
I
,
t
n
e
r
r
u
c
c
i
t
a
t
S
0.4
1.0
Supply Voltage, VDD (V)
0.8
0.6
)
e
g
a
t
s
/
A
n
(
T
A
T
S
I
,
t
n
e
r
r
u
c
c
i
t
a
t
S
0.325V
c)
102
101
100
10-1
10-2
10-3
10-4
Dynamic current, IDYN (µA/stage)
VDD=0.45V
1
VDD = 1V
0.45V
1.5
0.5
0
0.325V
0.425V
VDD = 1V
0.7V
VDD=0.55V
40
20
10
Delay/stage, τP (ps)
30
296K
VN-WELL = 0V
VP-WELL = 0V
4.3K
VTH = VTH(296K)
4.3K
VN-WELL = 4V
VP-WELL = -5V
4.3K
VN-WELL = 4V
VP-WELL = -5.8V
Fig. 8. (a) Comparison of Energy-Delay product vs VDD at room temperature and 4.3K with different forward back-biasing conditions. Note that for the case of
very high VFBB (VN-WELL=4V, VP-WELL=−5.8V) no minimum is observed down to VDD=0.325V. (b) Static current vs Delay/stage metric measured at RT and 4.3K
for different VDD (initial and final values for the studied VDD interval are given). (c) Static current vs dynamic current for different VDD at 296K and 4.3K. Note that
depending on FBB voltage, IDYN between 30nA/stage and 1.76µA/stage can be achieved. Initial and final values for the studied VDD interval are given.
(VDD from 0.325 to 1.2V). It is demonstrated that by properly
balancing the energy consumption and delay, the maximum
benefit in terms of speed and energy consumption can be
derived at a very low supply voltage of VDD=0.325V only. At
4.3K and high FBB voltage (VN-WELL=4V and VP-WELL=-5.8V),
a very small energy per transition of 0.186fJ with tP=37ps and
EDP=6.9fJ×ps
at VDD=0.325V. These
achievements prove that the 28nm FD-SOI platform provides
significant opportunities towards optimizing highly-efficient
and ultra-low-power cryogenic circuits for large-scale quantum
computing. Based on the promising results in this work, other
applications such as low temperature sensors, low power
neuromorphic circuits [26], or space electronics can be
envisioned.
achieved
are
ACKNOWLEDGMENT
This work is partly funded by French Public Authorities (NANO
2017) and Equipex FD-SOI. We also acknowledge financial
support from the EU under Project MOS-QUITO (No.688539).
The authors would like to thank ST-Crolles characterization
team, T. Poiroux, and A. Toffoli for fruitful discussions and
help in establishing the measurement protocol for ring
oscillators.
REFERENCES
[1] D. Loss and D. P. Divincenzo, "Quantum computation with quantum
1998.
vol. 57,
Phys.
Rev.
120,
no.
A,
dots",
doi:10.1103/PhysRevA.57.120
ENERGY EFFICIENCY OPTIMIZATION FOR ULTRA-LOW SUPPLY VOLTAGE UNDER
FORWARD BACK-BIASING AT 4.3K
TABLE III
VN-WELL (V)
VP-WELL (V)
IDYN
(nA/stage)
VDD (V)
0.325
0.5
4
-5.8
4
-4.8
4
-5.8
4
-4.8
tP
(ps)
37
71
17
21
170
92
518
426
ISTAT
(nA/stage)
7
0.7
30
1.6
The data reported in Table III are given to illustrate the
tremendous versatility of 28nm FD-SOI technology for both
ultra-low power and high performance applications for highly
reduced VDD at 4.3K. This highlights the possible trade-offs
between the delay and the static/dynamic power dissipation. For
instance, in case of VN-WELL=4V and VN-WELL= -4.8V, we obtain
highly improved delay of 31ps with IDYN=271nA/stage and
ISTAT=1.56nA/stage if VDD is increased to 0.425V.
IV. CONCLUSION
for
the
first
This paper describes,
time, electrical
characterization of 28nm FD-SOI ring oscillators down to 4.3K.
The unique capability of back-biasing in the development of
fast power-efficient peripheral circuitry is shown through the
analysis of delay per stage, static and dynamic current of the
ROs. Also, the trade-off between energy consumption and
delay per stage is discussed for a large range of supply voltage
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
[2] L. R. Schreiber and H. Bluhm, "Silicon comes back", Nature Nanotech.,
7
vol. 9, pp. 966-968, 2014. doi: 10.1038/nnano.2014.249
[3] M. Veldhorst, C. H. Yang, J. C. C. Hwang, W. Huang, J. P. Dehollain, J.
T. Muhonen, S. Simmons, A. Laucht, F. E. Hudson, K. M. Itoh, A.
Morello, and A. S. Dzurak, "A two-qubit logic gate in silicon", Nature
vol. 526, pp. 410-414, 2015. doi: 10.1038/nature15263
[4] R. Maurand, X. Jehl, D. Kotekar-Patil, A. Corna, H. Bohuslavskyi, R.
Laviéville, L. Hutin, S. Barraud, M. Vinet, M. Sanquer, and S. De
Franceschi, "A CMOS silicon spin qubit", Nature Com., vol. 7, no. 13575,
2016. doi: 10.1038/ncomms13575
[5] L. Hutin, R. Maurand, D. Kotekar-Patil, A. Corna, H. Bohuslavskyi, X.
Jehl, S. Barraud, S. De Franceschi, M. Sanquer, and M. Vinet, "Si CMOS
platform for quantum information processing," 2016 IEEE Symposium on
VLSI
doi: 10.1109/VLSIT.2016.7573380
[6] D. Reilly, "Engineering the quantum-classical interface of solid-state
Technology,
Honolulu,
2016,
HI,
pp.
2002,
[19] M. H. Na, E. J. Nowak, W. Haensch and J. Cai, "The effective drive
current in CMOS inverters," Digest. International Electron Devices
Meeting, San Francisco, CA, USA,
doi: 10.1109/IEDM.2002.1175793
[20] D. Jacquet, F. Hasbani, P. Flatresse, R. Wilson, F. Arnaud, G. Cesana, T.
Di Gilio, C. Lecocq, T. Roy, A. Chhabra, C. Grover, O. Minez, J. Uginet,
G. Durieu, C. Adobati, D. Casalotto, F. Nyer, P. Menut, A. Cathelin, I.
Vongsavady, and P. Magarshack, "A 3 GHz Dual Core Processor ARM
Cortex TM -A9 in 28 nm UTBB FD-SOI CMOS With Ultra-Wide
Voltage Range and Energy Efficiency Optimization," IEEE Journal of
Solid-State Circuits, vol. 49, no. 4, pp. 812-826, April 2014. doi:
10.1109/JSSC.2013.2295977
pp.
1-2.
[21] M. Bhushan and M.B. Ketehen, "CMOS Power-Performance-density
Metrics" in CMOS test and Evaluation, 1st ed., New-York, NY, USA:
springer science, business media New-York, pp. 347-398, 2015.
qubits", npJ Quantum
doi: 10.1038/npjqi.2015.11
Information, vol. 1, no. 15011, 2015.
121-124.
8,
no.
no.
vol.
Rev.
1766,
Appl.,
044009,
[7] L. M. K. Vandersypen, H. Bluhm, J. S. Clarke, A. S. Dzurak, R. Ishihara,
A. Morello, D. J. Reilly, L. R. Schreiber, and M. Veldhorst, "Interfacing
spin qubits in quantum dots -- hot, dense and coherent", npJ Quantum
Information vol. 3, no. 34, 2017. doi:10.1038/s41534-017-0038-y
[8] M. Veldhorst, H. G. J. Eenink, C. H. Yang, and A. S. Dzurak, "Silicon
CMOS architecture for a spin-based quantum computer", Nature
Communications,
doi: 10.1038/s41467-017-01905-6
[9] P. Clapera, S. Ray, X. Jehl, A. Valentian, S. Barraud, and M. Sanquer,
"Design and Cryogenic operation of a hybrid quantum-CMOS circuit",
Phys.
vol. 4,
doi: 10.1103/PhysRevApplied.4.044009
[10] E. Charbon, F. Sebastiano, A. Vladimirescu, H. Homulle, S. Visser, L.
Song, and R.M. Incandela, "Cryo-CMOS for quantum computing," 2016
IEEE International Electron Devices Meeting (IEDM), San Francisco,
CA, 2016, pp. 13.5.1-13.5.4. doi: 10.1109/IEDM.2016.7838410.
[11] R.M. Incandela, L. Song, H.A.R. Homulle, F. Sebastiano, E. Charbon,
and A. Vladimirescu, "Nanometer CMOS characterization and compact
modeling at deep-cryogenic temperatures", 47th European Solid-State
Device Research Conference (ESSDERC), Leuven, pp. 58-61, 2017. doi:
10.1109/ESSDERC.2017.8066591
[12] H. Homulle, L. Song, E. Charbon and F. Sebastiano, "The Cryogenic
Temperature Behavior of Bipolar, MOS, and DTMOS Transistors in
Standard CMOS," in IEEE Journal of the Electron Devices Society, vol.
6, no. 1, pp. 263-270, 2018. doi: 10.1109/JEDS.2018.2798281
[13] B. Patra, R. Incandela, J. P. G. van Dijk, H. A. R. Homulle, L. Song, M.
Shahmohammadi, R. B. Staszewski, A. Vladimirescu, M. Babaie, F.
Sebastiano, and E. Charbon, "Cryo-CMOS Circuits and Systems for
Quantum Computing Applications," IEEE Journal of Solid-State Circuits,
vol. 53, no. 1, pp. 309-321, Jan. 2018. doi: 10.1109/JSSC.2017.2737549
[14] H. Homulle and E. Charbon, "Performance characterization of Altera and
Xilinx 28 nm FPGAs at cryogenic temperatures," 2017 International
Conference on Field Programmable Technology (ICFPT), Melbourne,
VIC, 2017, pp. 25-31. doi: 10.1109/FPT.2017.8280117
[15] Edmundo A. Gutiérrez-D., M. Jamal Deen, and C. Claeys, "Silicon
Devices and Circuits" in Low Temperature Electronics, 1st ed., San
Diego, CA, USA: Academic Press,
doi: 10.1063/1.1485590
[16] E. Beigne, J.-F. Christmann, A. Valentian, O. Billoint, E. Amat, D.
Morche, "UTBB FDSOI Technology Flexibility for ultra-low-power
Internet-of-things Applications", 45th European Solid State Device
Research
doi: 10.1109/ESSDERC.2015.7324739
(ESSDERC),
[17] N. Planes, O. Weber, V. Barral, S. Haendler, D. Noblet, D. Croain, M.
Bocat, P.-O. Sassoulas, X. Federspiel, A. Cros, A. Bajolet, E. Richard, B.
Dumont, P. Perreau, D. Petit, D. Golanski, C. Fenouillet-Béranger, N.
Guillot, M. Rafik, V. Huard, S. Puget, X. Montagner, M.-A. Jaud, O.
Rozeau, O. Saxod, F. Wacquant, F. Monsieur, D. Barge, L. Pinzelli, M.
Mellier, F. Boeuf, F. Arnaud, and M. Haond, "28nm FDSOI technology
low-voltage digital applications," 2012
platform for high-speed
Symposium on VLSI Technology (VLSIT), Honolulu, HI, 2012, pp. 133-
134. doi: 10.1109/VLSIT.2012.6242497
[18] H. Bohuslavskyi, S. Barraud, M. Cassé, V. Barral, B. Bertrand, L. Hutin,
F. Arnaud, P. Galy, M. Sanquer, S. De Franceschi, and M. Vinet, "28nm
Fully-depleted SOI technology: Cryogenic control electronics for
quantum computing", 2017 Silicon Nanoelectronics Workshop, Kyoto,
2017, pp. 143-144. doi: 10.23919/SNW.2017.8242338
conference
2001,
pp.
2017.
2015.
pp.
2014,
Grenoble,
[22] A. Beckers, F. Jazaeri, H. Bohuslavskyi, L. Hutin, S. De Franceschi, C.
Enz, "Design-oriented Modeling of 28nm FDSOI CMOS Technology
down to 4.2K for Quantum Computing", 2017 Joint International
EUROSOI Workshop and International Conference on Ultimate
Integration on Silicon (EUROSOI-ULIS), Granada, 2018.
temperature
[23] M. Shin, M. Shi, M. Mouis, A.Cros, E. Josse, G. T. Kim, and G. Ghibaudo
"Low
characterization of 14nm FDSOI CMOS
devices", 2014 11th International Workshop on Low Temperature
Electronics
29-32.
doi:10.1109/WOLTE.2014.6881018
[24] J. P.G. van Dijk, E. Kawakami, R. N. Schouten, M. Veldhorst, L. M. K.
Vandersypen, M. Babaie, E. Charbon, and F. Sebastiano, "The impact of
classical control electronics on qubit fidelity" arXiv:1803.06176 [quant-
ph], March 2018.
(WOLTE),
[25] B. Zhai, S. Pant, L. Nazhandali, S. Hanson, J. Olson, A. Reeves, M.
Minuth, R. Helfand, T. Austin, D. Sylvester, and D. Blaauw, "Energy-
Efficient Subthreshold Processor Design," IEEE Transactions on Very
Large Scale Integration (VLSI) Systems, vol. 17, no. 8, pp. 1127-1137,
Aug. 2009. doi: 10.1109/TVLSI.2008.2007564
[26] M. L. Schneider, C. A. Donnelly, S. E. Russek, B. Baek, M. R. Pufall, P.
F. Hopkins, P. D. Dresselhaus, S. P. Benz, W. H. Rippard, "Ultralow
power artificial synapses using nanotextured magnetic Josephson
junctions",
doi: 10.1126/sciadv.1701329
Advances,
Science
vol.
no.
4,
1,
2018.
105-240.
2015.
|
1810.08978 | 1 | 1810 | 2018-10-21T15:37:24 | Autonomous Deployment of a Solar Panel Using an Elastic Origami and Distributed Shape Memory Polymer Actuators | [
"physics.app-ph",
"cond-mat.soft"
] | Deployable mechanical systems such as space solar panels rely on the intricate stowage of passive modules, and sophisticated deployment using a network of motorized actuators. As a result, a significant portion of the stowed mass and volume are occupied by these support systems. An autonomous solar panel array deployed using the inherent material behavior remains elusive. In this work, we develop an autonomous self-deploying solar panel array that is programmed to activate in response to changes in the surrounding temperature. We study an elastic "flasher" origami sheet embedded in a circle of scissor mechanisms, both printed with shape memory polymers. The scissor mechanisms are optimized to provide the maximum expansion ratio while delivering the necessary force for deployment. The origami sheet is also optimized to carry the maximum number of solar panels given space constraints. We show how the folding of the "flasher" origami exhibits a bifurcation behavior resulting in either a cone or disk shape both numerically and in experiments. A folding strategy is devised to avoid the undesired cone shape. The resulting design is entirely 3D printed, achieves an expansion ratio of 1000% in under 40 seconds, and shows excellent agreement with simulation prediction both in the stowed and deployed configurations. | physics.app-ph | physics | Autonomous deployment of a solar panel using an Elastic Origami and Distributed
Shape Memory Polymer Actuators
Tian Chena, Osama R. Bilalb, Robert Langc, Chiara Daraiob∗, Kristina Sheaa∗
a Department of Mechanical and Process Engineering, ETH Zurich, Zurich 8092, Switzerland
b Division of Engineering and Applied Science, California Institute of Technology, Pasadena, CA 91125, USA and
c Lang Origami, Alamo, CA 94507, USA
(Dated: October 23, 2018)
Deployable mechanical systems such as space solar panels rely on the intricate stowage of passive
modules, and sophisticated deployment using a network of motorized actuators. As a result, a
significant portion of the stowed mass and volume are occupied by these support systems. An
autonomous solar panel array deployed using the inherent material behavior remains elusive. In this
work, we develop an autonomous self-deploying solar panel array that is programmed to activate
in response to changes in the surrounding temperature. We study an elastic "flasher" origami
sheet embedded in a circle of scissor mechanisms, both printed with shape memory polymers. The
scissor mechanisms are optimized to provide the maximum expansion ratio while delivering the
necessary force for deployment. The origami sheet is also optimized to carry the maximum number
of solar panels given space constraints. We show how the folding of the "flasher" origami exhibits a
bifurcation behavior resulting in either a cone or disk shape both numerically and in experiments.
A folding strategy is devised to avoid the undesired cone shape. The resulting design is entirely 3D
printed, achieves an expansion ratio of 1000% in under 40 seconds, and shows excellent agreement
with simulation prediction both in the stowed and deployed configurations.
I.
INTRODUCTION
The surging demand for deployable mechanical sys-
tems is predominantly driven by the need to explore ever
more inaccessible environments and to deliver increas-
ingly large and complex payloads. Existing need for de-
ployable systems includes space-based solar power [1] [2],
geoengineering [3], antennas [4], and propulsion [5]. The
design of these mechatronic systems largely considers the
stowage of the payload [6] and the actuation which trans-
form the payload from the stowed to the deployed con-
figuration [7]. Traditionally, the actuation is achieved
through a network of controlled actuators, which unpacks
the passive payload. While numerous studies are done to
ensure a robust stowage and deployment process, with
ever more complex systems, autonomous large-scale de-
ployment, weight and volume of power supply, compo-
nent jamming and failure, placement of the actuators,
and advanced control mechanisms continue to challenge
engineers.
The nascent field of programmable matter, where en-
gineers integrate material and functionality [8] to create
more reliable, adaptive and robust designs, is utilized
here to address these challenges. Programmable matter
refers to a material or a structure whose shape or stiff-
ness can be controlled [9, 10]. Numerous designs have
started to demonstrate shape reconfiguration [11]. For
example, fabricating swelling hydrogel to create doubly
curved surfaces [12] or flower petals [13], programming
and activation of shape memory polymer to transform a
sheet into a box [14] or to deploy a shape that was stowed
∗ To whom correspondence may be addressed.
Email:
[email protected] or [email protected]
in a cylinder [15] and transformable sheets using liquid
crystal polymer [16] or printing pre-strain [17]. When
the objects are fully 3D printed, the area is also called
4D printing, i.e. objects that reconfigure themselves in
time.
Functionalities beyond shape reconfiguration have also
been demonstrated. Pneumatics have been used to actu-
ate soft crawlers [18], a robotic octopus [19], and trans-
formable surfaces [20].Electrical components have been
embedded within the material itself as well to create
soft electronics [21] and cardiac micro-physiological de-
vices [22]. Instability has been used to create multi-stable
structures without mechanical hinges [23], the actuation
of soft directional swimmers [24], and stable propagation
of elastic energy [25]. While these works eliminate the
need for a physical tether or rigid machineries, the active
components are still being treated as separate entities
that need to be embedded into a passive vessel. This
complicates the design process and makes the system er-
ror prone, e.g. jamming.
By utilizing advanced material systems,
it becomes
possible to embed and distribute sensing, control, actu-
ation, and logic within the material itself rather than
as discrete components in an assembly [26]. Using a self-
deploying soft solar panel array as an example, we aim to
demonstrate stable self-deployment through distributed
actuation using the shape memory effect. The proposed
mechanical system consists of an outer ring and an inner
substrate (1a,b). The outer ring serves as the primary
structural system and actuation mechanism. The inner
substrate carries the solar panels and serves as the sec-
ondary actuation mechanism.
In such a configuration,
we show that the inner substrate and the solar panels fit
entirely within the inevitable void of the collapsed outer
ring.
8
1
0
2
t
c
O
1
2
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
1
v
8
7
9
8
0
.
0
1
8
1
:
v
i
X
r
a
2
FIG. 1. The schematic of the deployment strategy from (a)
to (b) of the presented mechanical system. An outer ring acts
as the primary actuation mechanism and structural support.
An inner substrate forms the secondary actuation mechanism
and provides the surface to carry the solar panels. Both com-
ponents are fabricated using a shape memory polymer as the
actuation mechanism.
To achieve a tunable expansion ratio, the outer ring
adopts the design of the Hoberman Sphere [27] where a
series of scissor mechanisms in a ring configuration serves
as the structural system and primary actuator. An elas-
tic "flasher" origami is used as a deployable surface and
a secondary actuator. We pose the maximization of ex-
pansion ratio and the maximization of the number of
fitted solar panels as discrete problems that are solved
analytically considering only kinematics. The resulting
mechanical system is able to achieve an area expansion
ratio of ten, from approximately 0.05 to 0.5 m2. The
dynamics of folding the "flasher" origami is studied in
detail as a bifurcation behavior is observed in both sim-
ulation and experiments. It is shown that depending on
the rate of folding, the origami folds into either a cone
or a disk shape. As a result, a rotational mechanism is
devised for collapsing of the whole system into its stowed
configuration.
For actuation, by distributing a shape memory poly-
mer in both the Hoberman ring and in the origami sub-
strate, we enable tuning of material stiffness by sens-
ing the surrounding temperature. In both stowage and
operating configurations where the temperature is lower
than the glass transition temperature (Tg) of the poly-
mer, both components contribute to the system's overall
stiffness. During collapsing and deployment, sensing a
temperature higher than (Tg), the material reduces in
stiffness by orders of magnitude and becomes compliant.
In this example, collapsing to the stowed configuration
is manual and programs a pre-strain in the shape mem-
ory polymers. The autonomous self-deployment occurs
when the reduction in stiffness relaxes the pre-strain and
reconfigures the system.
II. RESULTS AND DISCUSSION
The design and functionality of the self-deploying solar
panel array is described in detail first through its compo-
nents, i.e., the Hoberman ring, and the elastic "flasher"
origami. Then, the behavior of the entire assembly is
FIG. 2. The design of the autonomous self-deploying Hober-
man ring. a) A ring of the Hoberman Sphere toy in both
collapsed and expanded configuration. b) Zoomed to show
the mechanism with hubs and pins indicated. c) Fabricated
design of the 3D printed Hoberman ring in both collapsed and
expanded configurations showing 10 times change in area. d)
Zoomed in render showing the shape memory polymer hubs
and articulated pins.
presented. The Hoberman ring is described in terms of
its 1) mechanism, 2) analysis of its ratio of expansion,
and 3) means of actuation. The elastic "flasher" origami
is then described with respect to 1) the parametric op-
timization of the crease pattern, 2) dynamics of folding
that result in a bifurcation behavior, 3) means of actua-
tion. The description of the entire assembly consists of
the self-deploying experiment which shows two distinct
phases of deployment.
A. Hoberman Ring
The Hoberman Sphere [27] was initially patented by
Hoberman before being popularized as a toy. It makes
clever use of the classic scissor mechanism [28] to create
a sphere-like object which can expand to several times
its collapsed volume (Fig. 2a). Hubs are placed at the
intersection of the scissors, and pins are used to enable
the scissoring mechanism (Fig. 2b).
FIG. 3. Parametric design of the Hoberman ring accounting
for fabrication limitations and width of the members. Ex-
pansion ratio as a function of number of edges of the initial
polygon plotted for different member thickness, w. For mem-
ber thickness w = 4, the optimal expansion ratio occurs at
n = 20. Derivation can be found in the Appendix.
We adopt a 2D version (i.e., ring) of this design to
make an expanding planar mechanism (Fig. 2c).
In-
spired by numerous works which convert pure mecha-
nisms into mechanical meta-materials [29, 30], we en-
able autonomous expansion within the ring by replac-
ing the hubs with shape memory polymers as actuators
(Fig. 2d). The shape memory hubs have two polymer
states, glassy or rubbery, depending on whether they are
above or below the glass transition temperature, Tg. At
a temperature above Tg, the hubs can be deformed in a
predefined shape. If the hubs are held in this shape while
cooled down, they retrain the shape conferred upon heat-
ing, by "locking in" the strains. When reheated past Tg,
the shape memory hubs relax this pre-strain and conse-
quently deploy the structure.
The Hoberman ring is constructed using a regular poly-
gon as the base geometry. The number of edges, n, of this
polygon determines the expansion ratio of the mechanism
given a set of physical and fabrication constraints. We
define the expansion ratio as ∆A = Aexpanded/Acollapsed.
3
The radius of each configuration is measured as the dis-
tance between the polygon center and the outer most
vertex of the geometry. To maximize this expansion ra-
tio, we derive the relationship between ∆A against n.
The schematics (Fig. 3 top) show the geometric configu-
ration of collapsed and expanded rings, for different num-
ber of edges, n, assuming zero thickness, w, of the mem-
bers.
In this theoretical scenario, the inner vertices of
the ring coincides at the center of the polygon [31] [32].
By introducing finite thickness, the inner vertices col-
lide with each other as the ring collapses, forming an
inner circular void. Quantitatively, it is observed that
for a given w > 0 with increasing n, the expansion ratio
reaches a maximum (Fig. 3 bottom). This maximum oc-
curs with smaller n for larger values of w as the thickness
of the members prevents the mechanism from collapsing
to the center. Accounting for fabrication limitations and
robustness to repeated cycles of collapse-expansion, we
choose a thickness of w = 4 mm for testing. For this
thickness, the maximum expansion ratio of ∆A = 10.79
occurs for n = 20 edges.
The optimized ring including the SMPs is fabricated
in the expanded configuration and programmed to the
collapsed configuration. As the Hoberman mechanism
has a Poisson's Ratio of −1 at all strains, the program-
ming is done under T ≥ Tg by uniformly compressing
the ring until it fits into a predefined mold. The mold
is dimensioned such that the area expansion ratio is 10,
which is slightly less than the predicted value (10.79).
The specimen is cooled while in the mold and then ex-
tracted. The fabricated specimens demonstrate the pre-
dicted area change as well as structural stability in both
collapsed and expanded configurations.
The radial pressure required during compression and
exerted during expansion can be predicted by under-
standing the force displacement behavior of the shape
memory hubs.
It is known that the recovery force is
dependent on the dimensions of the shape memory ma-
terial, strain rate, and the ambient temperature [15].
Compression experiments are done to a segment of the
Hoberman ring under temperature T > Tg with 10 mm
of displacement being prescribed to both sides of the seg-
ment at a rate of 0.5 mm s−1. The ends of the segment
are pin-connected to the compression plates. The reac-
tion forces are recorded. Due to the small forces exerted
by each shape memory hub, force-displacement measure-
ments of six specimens are obtained under the same con-
dition. The mean is plotted (Fig. 4a). A load cell with a
maximum load of 250 N is used, the measurement error
is ±0.005% (±0.001 25 N). The same geometry is simu-
lated using finite element analysis. The shape memory
hub is modeled using a linear viscoelastic constitutive
model [23] and deformed with a prescribed displacement
d = [0− 10 mm]. The shape memory hub exhibits a non-
linear force displacement behavior (Fig. 4a) as it bends.
Numerical and experimental results are comparable, de-
spite the noise introduced due to the low force output.
The radial pressure is calculated with trigonometry
4
B. Elastic "flasher" origami
Origami structures, like the Miura fold [33], have been
studied extensively in the context of deployable sys-
tems [34]. In the case of rigid origami, the path of folding
and unfolding can be predicted with precision [35]. The
ability to predict the folding and unfolding paths allows
the design of origami mechanisms that can be actuated
using motors or thrusters. Reconfigurable systems can
benefit from the large volume change achievable by some
origami patterns. In existing deployable designs, origami
surfaces typically serve as the passive components in a
motorized unfolding mechanism [5].
Indeed, while the
concept of using a Hoberman like scissor mechanism to
unfold solar panels has already been proposed [36], its ac-
tuation or the space to store rigid panels remained an un-
resolved problem. In additional to origami's folding kine-
matics, the mechanical behavior of folded origami struc-
tures is becoming a focus of study [6]. With the introduc-
tion of 4D printing, fold lines within an origami pattern
could be programmed to self-fold [37]. This reconfigura-
bility has recently been exploited in the design of acoustic
waveguides [38], prismatic reconfigurable materials [39]
and tunable thermal expansion meta-material [40].
To create a foldable panel with the smallest possible
collapsed footprint, we adopt an origami pattern called
the "flasher"(Fig. 5a) [41]. The "flasher" origami starts
as a flat sheet, and can then be folded into a cylindrical
element, by allowing each section (arm) to fold around a
center core. During the experiments, a mock up of the
solar panels are printed on acetate which has comparable
stiffness [42] and same dimension [43] of commercially
available flexible solar panels (Fig.5b).
Much work has been done on the deployability of rigid
foldable versions of the "flasher" origami [3, 44]. Our
flexible crease pattern can be categorized with two vari-
ables [44]. The first variable is the rotational order of
the "flasher" n which is also the number of sides of the
outer polygon. The second variable is the number of cir-
cumferential layers in the crease pattern, c . We do not
physically discretize the facades in the radial direction,
but we allow them to fold flexibly.
As we intend the origami to serve as a hosting platform
for solar panels, we optimize the crease pattern to fit the
maximum number of solar panels of a given size. While
flexible solar panels can bend to a certain curvature, if
they are placed across folding lines, they would increase
the folding resistance and potentially break during fold-
ing. Therefore, they are only placed on the facades in
between crease lines. In the generation of the fold pat-
tern, we neglect the thickness of the facades, and allow
them to be straight, to better accommodate rectangular
solar panels. The strain during folding is assumed to be
absorbed by the elasticity of the selected materials. The
optimization problem used for design is stated in Eq. 2,
where we maximize the number of solar panels that can
fit within a given pattern. Constraints to this problem
are as follows, 1) the outer dimension of the unfolded
FIG. 4. Mechanics of the shape memory hubs and the re-
lationship with the radial pressure of the Hoberman ring.
a)Force displacement plot of a single shape memory hub from
experiments and simulations. b) Overall pressure produced
by all the shape memory joints of a n = 20 Hoberman ring as
a function of the angle in the scissor mechanism.
(Fig. 4b). As FH is horizontal at all stages of loading,
the force in the radial direction depends only on the an-
gle α. We empirically measure α from the expanded to
the collapsed configuration and plot against the radial
pressure (Eq. 1), where n is the number of polygon edge
and Rexp is the outer radius.
p = FH
n
2πRexp
tan(α)
cos( π
n )
.
(1)
5
height of the origami in the folded configuration, rmax is
the inner void space of the Hoberman ring (as discussed
in the previous section, and detailed in Appendix I).
The feasible region of the objective landscape (Fig. 6)
show optimized solutions of 60 solar panels at (n, c) =
(7, 5), (10, 3), (14, 1), (14, 2). As expected, as the number
of rotational order n increase, more solar panels can be
fitted. The constraints however dictate that beyond a
certain n or c, the folded dimension rf,o exceeds that of
the inner dimension of the Hoberman ring rmax. The
adopted solution is a decagon (n = 10) with three layers
(c = 3).We adopt this design because n = 10 is a fraction
of the number of scissor mechanisms in the Hoberman
ring. In this way, periodic connections can be made be-
tween the two. The derivation of the folded dimension of
the origami is detailed in Appendix II.
The design is fabricated with two layers: The 3D
printed base layer and the solar panel layer with thick-
nesses 0.15 mm and 0.24 mm, respectively. The thickness
of the facades provide structural rigidity during the ini-
FIG. 6. Maximizing the number of flexible solar panels by
changing the "flasher" origami crease pattern. Two variables,
n - number of edges in the pattern's outer polygon, c - number
of radial layers in the pattern, are optimized to fit the largest
number of solar panels of a given dimension. The contour
plot shows the number of solar panels that can fit for every
pair of (n, c).
FIG. 5. Elastic "flasher" origami in both collapsed and ex-
panded configurations. a) The "flasher" origami with crease
pattern indicated. b) Fabricated specimen with placement of
flexible solar panels indicated. The crease pattern is set such
that a maximum number of solar panels of a given dimension
can fit.
and folded origami must fit within the expanded and col-
lapsed Hoberman ring respectively, and 2) Each facade
of the crease pattern must be at least as wide as a given
solar panel plus the width of the fold lines. Preference is
given to patterns that can construct periodic linkages to
the Hoberman ring.
minimize
c,n
where
subject to
− max [p(c, n), 0]
(cid:34) c(cid:88)
(cid:18)
ru,o − wPV + hf (i − 1)
tan(π/n)
(cid:19)(cid:35)
p(c, n) = floor
i=1
rf,o − rmax ≤ 0
wPV − hf ≤ 0
(2)
Where ru,o and rf,o are the outer polygon radii of the
unfolded and folded configuration respectively, wPV is the
smaller dimension of a rectangular solar panel, hf is the
6
FIG. 7. Bifurcation during folding of the "flasher" origami as explained by the elastic energy within the fold lines. Two distinct
folded shapes are observed during folding both in experiment and in simulation. a) Plot of total height of the origami as a
function of the fold angle (between 0 and π). b) Plot of the total elastic energy as a function of the fold angle. c) Simulated
folded shapes (i.e., α = π) resulting in either a cone, or a disc.
tial folding process, but is flexible enough to be wound
around the center core. The gaps between the facades
form the lines of fold. The gaps width is greater than 2t
where t is the thickness of the facades. Due to the inher-
ent elasticity of the material in the rubbery state, both
mountain and valley crease lines are assigned the same
width. Voids are placed at the intersection of multiple
crease lines to reduce stress concentration.
For the programming of the origami and consequently
of the whole assembly, we apply a rotation at the cen-
ter core while keeping the mountain and valley fold lines
of the origami in the correct half plane(i.e. z > 0 and
z < 0) [45]. To apply this rotation, flexible wires are
connected between the inner most vertices of the origami
and a center rotational core (See Fig. 12). It is noted that
during folding, the center may "pop" out of plane, forc-
ing the whole origami to irreversibly assume a cone-like
shape. To better understand this phenomenon, the fold-
ing process is simulated by assuming the crease pattern is
an edge-node network. The coordinates of the nodes are
solved such that there is minimal axial strain along the
bars, and that the dihedral fold angle between all crease
lines equal to a prescribed value θ [46]. By incrementing
θ, a pseudo-dynamic folding simulation is achieved [47].
As this method is used for rigid origami, the facades of
the "flasher" in the radial direction are finely segmented
to mimic a flexible folding behavior. A bifurcation behav-
ior is observed, i.e. two different heights can be achieved
from the same fold angle (Fig. 7a, c1−5 and d1−5). The
cone shape is triggered if the change in the folding an-
gle is small between two pseudo time steps, and the disc
shape is triggered if the change is large. We see that the
strain energy or the error is much larger for smaller fold
angles for the disc than it is for the cone. This correlation
inverses at a larger fold angle (Fig. 7b). This explains the
center of the origami "popping-up" during folding, and
is avoided in experiments by manually holding the center
of the "flasher" down.
C. Autonomous self-deploying solar panel array
While the proposed design can be fabricated using a
3D printer monolithically, it is made as an assembly for
demonstration purposes. Both the Hoberman ring and
the elastic origami are fabricated in the expanded config-
uration and assembled using rotational hooks(Fig. 8a).
While both components have been programmed individ-
ually for demonstration (Fig. 2c and Fig. 5b), the assem-
bly is programmed in one step in a heated environment
7
FIG. 8. Deployable solar panel array. a) Rendering of the design assembly with the different components labeled. b) Photos
showing the collapsed and the expanded configuration of the solar panel array. c) Video frames of the deployment process in
water above Tg.
(T ≥ Tg). The folding is done by a single person, in con-
trast to the common practice of large origami patterns
that usually requires as many people as the number of
arms. The collapse starts by rotating the core which the
facades are connected to. As the core rotates, the whole
geometry is pulled in and wrapped. Then a cylindrical
mold that would give the desired collapsing ratio is used
to demonstrate that the collapsed design can fit within a
given space. While confined in this mold, the collapsed
design is cooled to the stowing temperature (Fig. 8b).
The mold is then removed.
The design must be deployed in an environment heated
past Tg. In our experiments, it is deployed under water,
to simulate a reduced gravity environment. First a rapid
rotational motion is observed, where the deployment is
largely initiated by the origami itself. At a certain stage,
the rotation stops and the remaining folds are flattened
by the Hoberman ring (Fig. 8c). This deployment behav-
ior mimics a proposed two-stage mechanized deployment
of a solar sail,[5] yet requires no actuators or power sup-
ply.
III. CONCLUSION
We exploit the large area expansion capability of the
Hoberman ring to autonomously deploy an array of flex-
ible solar panels mounted on an elastic "flasher" origami
substrate. The design of the system is created by an-
alyzing the Hoberman ring and the "flasher" origami
separately. First we propose to replace the rotational
hubs of the Hoberman ring with shape memory poly-
mers to enable temperature-triggered deployment. We
then parametrize and optimize the expansion ratio of the
Hoberman ring assuming rigid behavior. Lastly, we ana-
lyze the mechanics of collapse and expansion of a single
shape memory hub.
The analysis of the "flasher" origami begins by param-
8
eterizing and optimizing the crease pattern to accom-
modate the largest number of solar panels. Second, a
simplified strain-energy model is proposed to explain the
seemingly bifurcating behavior exhibited by the structure
while folding.
As our proposed design is under the highest mechani-
cal stress during the collapse process, we guarantee that
during deployment no material failure can occur. We
demonstrate the controlled deployment of a mechanical
system without external actuators, sensors or power sup-
ply. Rather, it is achieved by embedding and distributing
these functionalities within the material itself.
ACKNOWLEDGMENTS
C.D. acknowledges support from the Army Research
Office Grant W911NF-17-1-0147 and the Caltech/JPL
President's and Director's Fund Program.
[1] J. O. McSpadden and J. C. Mankins, IEEE Microwave
413 (2016), arXiv:arXiv:1011.1669v3.
Magazine 3, 46 (2002).
[14] Y. Mao, K. Yu, M. S. Isakov, J. Wu, M. L. Dunn, and
[2] T. McGuire, M. Hirsch, M. Parsons, S. Leake,
and
H. Jerry Qi, Scientific Reports 5, 13616 (2015).
J. Straub,
, 98650C (2016).
[3] D. Sigel, B. P. Trease, M. W. Thomson, D. R. Webb,
P. Willis, and P. D. Lisman, in Proceedings of the ASME
2014 International Design Engineering Technical Confer-
ences & Computers and Information in Engineering Con-
ference (IDETC/CIE 2014) (Buffalo, New York, 2014)
pp. DETC2014 -- 34315.
[4] Y. Zhang, D. Yang,
and S. Li, Mechatronics 35, 71
(2014).
[5] O. Mori, H. Sawada, R. Funase, M. Morimoto, T. Endo,
T. Yamamoto, Y. Tsuda, Y. Kawakatsu, J. Kawaguchi,
Y. Miyazaki,
and Y. Shirasawa, Transactions of the
Japan Society for Aeronautical and Space Sciences,
Aerospace Technology Japan 8, To 4 25 (2010).
[6] M. Schenk, S. G. Kerr, a. M. Smyth, and S. D. Guest,
Proceedings of the First Conference Transformables , 1
(2013).
[7] J. Block, M. Straubel, and M. Wiedemann, Acta Astro-
nautica 68, 984 (2011).
[8] G.-Z. Yang, P. Fischer,
and B. J. Nelson, Science
Robotics 2, eaap9294 (2017).
[9] E. Hawkes, B. An, N. M. Benbernou, H. Tanaka, S. Kim,
E. D. Demaine, D. Rus, and R. J. Wood, Proceedings of
the National Academy of Sciences 107, 12441 (2010).
[15] M. Wagner, T. Chen,
and K. Shea, 3D Printing
and Additive Manufacturing September, 3dp.2017.0027
(2017).
[16] T. J. White and D. J. Broer, Nature Materials 14, 1087
(2015), arXiv:arXiv:1011.1669v3.
[17] A. A. Bauhofer, S. Krodel, J. Rys, O. R. Bilal,
A. Constantinescu, and C. Daraio, Advanced Materials
1703024, 1703024 (2017).
[18] R. F. Shepherd, F. Ilievski, W. Choi, S. a. Morin, A. A.
Stokes, A. D. Mazzeo, X. Chen, M. Wang, and G. M.
Whitesides, Proceedings of the National Academy of Sci-
ences 108, 20400 (2011), arXiv:84855503964.
[19] M. Wehner, R. L. Truby, D. J. Fitzgerald, B. Mosadegh,
G. M. Whitesides, J. A. Lewis, and R. J. Wood, Nature
536, 451 (2016).
[20] A. J. H. Pikul, S. Li, H. Bai, R. T. Hanlon, I. Cohen,
and R. F. Shepherd, Science , 1 (2017).
[21] A. D. Valentine, T. A. Busbee, J. W. Boley, J. R. Raney,
A. Chortos, A. Kotikian, J. D. Berrigan, M. F. Durstock,
and J. A. Lewis, Advanced Materials 1703817, 1 (2017).
[22] J. U. Lind, T. A. Busbee, A. D. Valentine, F. S.
Pasqualini, H. Yuan, M. Yadid, S.-J. Park, A. Kotikian,
A. P. Nesmith, P. H. Campbell, J. J. Vlassak, J. A. Lewis,
and K. K. Parker, Nature Materials 16, 303 (2016).
[10] O. R. Bilal, A. Foehr, and C. Daraio, Advanced materials
[23] T. Chen and K. Shea, 3D Printing and Additive Manu-
29, 1700628 (2017).
facturing June (2018), arXiv:1711.00452.
[11] Z. Chen, G. Huang, I. Trase, X. Han, and Y. Mei, Phys-
[24] T. Chen, O. R. Bilal, K. Shea, and C. Daraio, arXiv , 1
ical Review Applied 5, 1 (2016), arXiv:1512.04967.
(2017), arXiv:1710.04723.
[12] D. Raviv, W. Zhao, C. McKnelly, A. Papadopoulou,
A. Kadambi, B. Shi, S. Hirsch, D. Dikovsky, M. Zyracki,
C. Olguin, R. Raskar, and S. Tibbits, Scientific Reports
4, 7422 (2014).
[13] A. Sydney Gladman, E. A. Matsumoto, R. G. Nuzzo,
L. Mahadevan, and J. A. Lewis, Nature materials 15,
[25] J. R. Raney, N. Nadkarni, C. Daraio, D. M. Kochmann,
J. A. Lewis, and K. Bertoldi, Proceedings of the National
Academy of Sciences 113, 9722 (2016).
[26] R. L. Truby and J. A. Lewis, Nature 540, 371 (2016).
[27] C. Hoberman, "Reversibly Expandable Doubly-Curved
Truss Structure," (1990).
[28] C. L. Larson, "Scissor-Lift Mechanism," (1966).
[29] K. Bertoldi, V. Vitelli, J. Christensen, and M. van Hecke,
Nature Reviews Materials 2, 17066 (2017).
[30] S. Janbaz, M. McGuinness, and A. A. Zadpoor, Physical
Review Applied 9, 64013 (2018).
[31] Z. You and S. Pellegrino, International Journal of Solids
and Structures 34, 1825 (1997).
[32] J. Patel and G. K. Ananthasuresh, International Journal
of Solids and Structures 44, 6279 (2007).
[33] Z. Y. Wei, Z. V. Guo, L. Dudte, H. Y. Liang,
and
L. Mahadevan, Physical Review Letters 110, 1 (2013),
arXiv:1211.6396.
[34] K. Miura, The Institute of Space and Astronautical Sci-
ence report 618, 1 (1985).
[35] T. Tachi, Origami 4 Fourth International Meeting of
Origami Science Mathematics and Education , 175
(2009).
[36] C. Hoberman, "Folding Covering Panels for Expanding
Structures," (2004).
[37] Q. Ge, C. K. Dunn, H. J. Qi, and M. L. Dunn, Smart
Materials and Structures 23, 1 (2014).
[38] S. Babaee, J. T. B. Overvelde, E. R. Chen, V. Tournat,
and K. Bertoldi, Science Advances 2, e1601019 (2016).
[39] J. T. B. Overvelde, J. C. Weaver, C. Hoberman, and
K. Bertoldi, Nature 541, 347 (2017).
[40] E. Boatti, N. Vasios, and K. Bertoldi, Advanced Mate-
rials 29 (2017), 10.1002/adma.201700360.
[41] J. Shafer, Origami to Astonish and Amuse, 1st ed. (St.
Martin's Griffin, 2001) pp. 1 -- 256.
[42] A. Czanderna and F. Pern, Solar Energy Materials and
Solar Cells 43, 101 (1996).
[43] Flex Solar Cells, PowerFilm Wireless Electronics Series
SP3-12, Tech. Rep.
[44] S. A. Zirbel, R. J. Lang, M. W. Thomson, D. A. Sigel,
P. E. Walkemeyer, B. P. Trease, S. P. Magleby,
and
L. L. Howell, Journal of Mechanical Design 135, 111005
(2013).
[45] W. E. Lanford, "Folding Apparatus," (1961).
[46] M. Schenk and S. D. Guest, Origami 5 Fifth International
Meeting of Origami Science Mathematics and Education
, 1 (2011).
[47] A. Ghassaei, "OrigamimSimulator," (2017).
Appendix A: Derivation of area expansion ratio of
the Hoberman ring
The Hoberman sphere is created using a regular poly-
hedron as a base, similarly, the ring can be created using
a regular polygon. We derive the analytical relationship
between the change in area from the expanded to the
collapsed configuration of the Hoberman ring in relation
to the number of edges of the initial polygon. Further,
we introduce three variables to account for fabrication
constraints (Fig. 9). Given the circumscribing diameter
of the polygon, dOC, as dictated by the maximum fab-
rication dimension, we calculate the edge length s, the
inscribing diameter dIC, the external angle θEA and the
internal angle θIA.
θEA =
2π
n
; θIA = (n − 2)
π
n
9
(A2)
We introduce an isogonal polygon with 2n edges for
the construction of the double scissor mechanism. The
edges of this polygon are alternatively grouped into two
sets e1 and e2 with lengths l∗
2 respectively. The
perpendicular distance from the origin to edge set e1 is
defined by varying the inscribing diameter dIC with ξ to
create the isogonal polygon. Edge set e2 rests on the
original regular polygon.
1 and l∗
a =
1
2
dIC + ξ
(A3)
This is done such that the nominal length of scissors
with pin ends are shorter than the one with hub ends.
The pin ended scissors require an added length as a hole
must be made at the end of the link,(Fig. 10), where as
the hub is a flexible polymer which deforms. By adding
ξ, we can tailor the system such that the physical embod-
iment would be closer to equal length. Length of edge set
e1 is then,
l∗
1 = (
1
2
dOC − a) tan(
θIA
2
)
(A4)
To calculate the length of the edge set e2 we first cal-
culate the two angles, θ1 and θ2, formed by connecting
the ends of each set of edges to the center,
(cid:18) l∗
(cid:19)
θ1 = tan−1
1
a
,
(A5)
θ2 =
θEA − θ1.
1
2
Then we find the length of edge set e2,
l∗
2 = tan(θ2)
dIC
2
.
(A6)
(A7)
If the scissor mechanism has an initial angle α,(Fig. 2b)
the lengths of the half scissor become l1 and l2,
l1 = l∗
1
1
cos(α)
l2 = l∗
2
;
1
cos(α)
.
(A8)
Without loss of generality, we assume ξ ≥ 0 since a
negative ξ of the same magnitude would only mirror the
structure. With this assumption, we see that l1 < l2. We
define l = l2 and θ = θ2. Now we assign physical dimen-
sions to the collapsed ring. Assuming that the width of
each scissor, w, equals the edge length of a fictitious in-
ner polygon of edge 2n, and an inscribing diameter dIC,f ,
then we can calculate an angle βmax as the angle between
two scissors in the collapsed configuration:
s = dOC sin(
π
n
) ; dIC = s cot(
π
n
)
(A1)
dIC,f = w cot(
π
2n
).
(A9)
10
FIG. 9. Construction of a Hoberman ring. Variables used in the derivation are shown.
FIG. 10. Rationale for two different lengths in the scissor segments, the pin ends are longer than the hub ends.
The maximum angle of rotation between the scissors
is therefore,
βmax =
− θ − arcsin
π
2
dIC,f sin(θ)
2l
.
(A10)
We define the ratio of area expansion as:
∆A =
[dIC + 2 sin(α)l]2
[4l sin(βmax) + dIC,f ]2 .
(A11)
In figure 1 of the manuscript, we plot ∆A vs. n for
different values of w.
Appendix B: Optimization landscape and constraint
boundaries of the "flasher" origami
In this work, we adopt the design of a "flasher" origami
as the basis to attach the solar panels. We relate the
shape and folds of a flash origami pattern to the num-
ber of solar panels that can fit on such a pattern. With
this, we are able to generate a crease pattern that can
accommodate the given number of solar panels of a cer-
tain size. We also use this derivation to relate the radius
of the unfolded pattern to the folded one.
Given two concentric polygons of n sides, we rotate the
outer polygon such that the edges of the inner polygon, if
extended outward, bisects the edges of the outer polygon.
This defines the outer and inner limits of the unfolded
pattern. Two other polygons of n sides are used to define
the outer and the inner limits of the folded pattern. The
circumscribing radius, inscribing radius, and edge length
are denoted as R, r, and s respectively.
The superscript ∗ refers to two outer polygons offset
by the asymmetry of the origami pattern (Fig. 11). The
relationship between the offset quantities is,
Ru,o = (Ru,o
∗2 + R2
u,i)
1
2
(B1)
These four polygons are denoted with subscripts u,o,
f,o, u,i, f,i referring to unfolded outer, folded outer, un-
folded inner, and folded inner respectively. The known
quantities are Rf,i, Ru,o
∗, Ru,i.
We define the radius of the folded geometry, Rf,o, as
Ru,i plus the number of wrapped layers, p. The variables
c and t refers to the number of levels in the "flasher"
pattern, and the thickness of origami sheet respectively.
Rf,o = Rf,i + p(2ct)
(B2)
The number of edges each arm of the outer polygon can
wrap around the inner polygon is calculated as the ratio
between the inscribed radius of the unfolded geometry
and the edge length of the folded geometry, Sf,o. Note
that this edge length increases as more layers are wrapped
and the radius increases.
p =
(B3)
∗
ru,o
sf,o
The inscribed radius is ru,o = Ru,o cos π
n and the edge
length is su,o = 2Ru,o sin π
n . Similarly, we calculate the
edge lengths sf,i, su,i, and the inscribed radius of the in-
ner polygon ru,i. we calculate the offset inscribed radius,
∗ = (ru,o
2 − ru,i
2)
ru,o
1
2 − su,i
2
(B4)
11
FIG. 11. Construction of a crease pattern of a "flasher" origami. Variables used in the derivation are shown.
and the edge length of the folded polygon follows,
We can solve for Rf,o, which forms the g2 constraint
sf,o = 2Rf,o sin(
π
n
).
(B5)
in Figure 2 of the manuscript.
Rf,o =
n
(cid:118)(cid:117)(cid:117)(cid:116)sin
(cid:16) π
(cid:114)(cid:16)
Rf,i sin
+2ct
(cid:17)(cid:32)
(cid:16) π
(cid:16) π
(cid:17)
cos
n
n
n
cos
Ru,i + 4ct
(cid:114)(cid:16)
(cid:17)
(cid:16) π
2 +(cid:0)−2ctRu,i + Rf,i
(cid:17)(cid:32)
(cid:16) π
(cid:16) π
(cid:17)
−4ct sin
n
n
−4ct sin
R∗
(cid:17)(cid:17)2
(cid:118)(cid:117)(cid:117)(cid:116)sin
+
u,o
n
u,o
R∗
(cid:17)(cid:17)2
(cid:16) π
(cid:17)(cid:41)
(cid:16) π
2(cid:1) sin
(cid:114)(cid:16)
n
·
Ru,i + 4ct
cos
(cid:16) π
(cid:17)
n
2 + sin
(cid:33)
2
Rf,i
Rf,i
(cid:16) π
(cid:17)(cid:17)2
n
R∗
u,o
2 + sin
(cid:16) π
(cid:17)
n
2
Rf,i
(cid:33)−1
(B6)
The first constraint requires the height of the folded
origami to be greater than the smaller dimension of the
solar panels. The height is simply the edge length of the
unfolded polygon divided by twice number of layers in
the "flasher":
(cid:113)
hf =
Ru,i
2
2 + R∗
c
u,o
sin
.
(B7)
(cid:16) π
(cid:17)
n
Appendix C: Collapsing procedure
Thermomechanically, the procedure to collapse the
system from the deployed configuration to the stowed
configuration is referred to as "programming". It is done
at a temperature higher than the glass transition temper-
ature of the shape memory polymers within the system.
Components of the system (i.e. the "flasher" origami,
the scissor mechanisms, and the connections between the
two) are fabricated in the deployed configuration and as-
sembled along with the solar panels (Fig. 12a,c). A center
core is installed and connected to the "flasher" origami
through a number of flexible wires (Fig. 12b,d).
The assembled system is submerged in heated water
(T ≥ Tg), and the center core is rotated as the crease
lines are folded. This continues until the collapsed sys-
tem fits within a prescribed cylinder (Fig. 12e). Once
in the cylinder and cooled to below the glass transition
temperature, the cylinder is released (Fig. 12f).
12
FIG. 12. "Programming" of the solar panel array. a,c) The procedure starts with the deployed configuration. b,d) A center
rotational core is installed and connected with the origami via flexible wires. e) In a heated environment, the system is rotated
and folded until it fits within a cylindrical mold. f) After cooling within the mold, it is released and stable on its own.
|
1903.12341 | 1 | 1903 | 2019-03-29T03:33:42 | Enhanced Ferroelectric Functionality in Flexible Lead Zirconate Titanate Films with In-Situ Substrate-Clamping Compensation | [
"physics.app-ph"
] | Much attention has been given recently to flexible and wearable integrated-electronic devices, with a strong emphasis on real-time sensing, computing and communication technologies. Thin ferroelectric films exhibit switchable polarization and strong electro-mechanical coupling, and hence are in widespread use in such technologies, albeit not when flexed. Effects of extrinsic strain on thin ferroelectric films are still unclear, mainly due to the lack of suitable experimental systems that allow cross structural-functional characterization with in-situ straining. Moreover, although the effects of intrinsic strain on ferroelectric films, e.g. due to film-substrate lattice mismatch, have been investigated extensively, it is unclear how these effects are influenced by external strain. Here, we developed a method to strain thin films homogenously in-situ, allowing functional and structural characterization while retaining the sample under constant straining conditions in AFM and XRD. Using this method, we strained the seminal ferroelectric, PbZr0.2Ti0.8O3 that was grown on a flexible mica substrate, to reduce substrate clamping effects and increase the tetragonality. Consequently, we increased the domain stability, decreased the coercive field value and reduced imprint effects. This method allows also direct characterization of the relationship between the lattice parameters and nanoscale properties of other flexible materials. | physics.app-ph | physics | Enhanced Ferroelectric Functionality in Flexible Lead Zirconate Titanate Films
with In-Situ Substrate-Clamping Compensation
Rachel Onn Winestook,1,2,† Cecile Saguy,
2,† Chun-Hao Ma,3,4,5
Ying-Hao Chu3,4,5 and Yachin Ivry1,2*
1Department of Materials Science and Engineering, Technion -- Israel Institute of Technology,
Haifa 3200003, Israel.
2Solid State Institute, Technion -- Israel Institute of Technology, Haifa 3200003, Israel.
3Department of Materials Science and Engineering, National Chiao Tung University, Hsinchu
30010, Taiwan.
4Institute of Physics, Academia Sinica, Taipei 11529, Taiwan.
5Material and Chemical Research Laboratories, Industrial Technology Research Institute,
Hsinchu, 31040, Taiwan.
†These authors contributed equally to the current work.
*Correspondence to: [email protected].
Keywords:
Flexible electronics, ferroelectric stability, nano-domains, in-situ AFM, nanoscale strain
Abstract
Much attention has been given recently to flexible and wearable integrated-electronic
devices, with a strong emphasis on real-time sensing, computing and communication
technologies. Thin ferroelectric films exhibit switchable polarization and strong electro-
mechanical coupling, and hence are in widespread use in such technologies, albeit not when
flexed. Effects of extrinsic strain on thin ferroelectric films are still unclear, mainly due to the lack
of suitable experimental systems that allow cross structural-functional characterization with in-
situ straining. Moreover, although the effects of intrinsic strain on ferroelectric films, e.g. due to
film-substrate lattice mismatch, have been investigated extensively, it is unclear how these
effects are influenced by external strain. Here, we developed a method to strain thin films
homogenously in-situ, allowing functional and structural characterization while retaining the
sample under constant straining conditions in AFM and XRD. Using this method, we strained the
seminal ferroelectric, PbZr0.2Ti0.8O3 that was grown on a flexible mica substrate, to reduce
substrate clamping effects and increase the tetragonality. Consequently, we increased the
domain stability, decreased the coercive field value and reduced imprint effects. This method
allows also direct characterization of the relationship between the lattice parameters and
nanoscale properties of other flexible materials.
Introduction
The growing interest in flexible and foldable electronics raises the need for functional
materials with suitable characteristics, mainly in the form of thin films. Ferroelectrics are
functional materials that exhibit reversible spontaneous polarization, while they are a sub-group
of piezoelectrics and hence demonstrate strong electro-mechanical coupling. Ferroelectrics are
thus used in a broad range of applications that may benefit from mechanical flexibility, including
sensors and medical monitoring devices as well as mobile-phone antennae and non-volatile
memory devices. However, characterizing functional properties of thin ferroelectric films at the
device-relevant length scale while flexing the material is a great challenge.
Effects of mechanical strain on the piezoresponse and polarization hysteresis loops of thin
ferroelectric films were observed first macroscopically (e.g. with optical interference methods),
demonstrating clockwise or counterclockwise rotation of the hysteresis loop under tensile and
compressive strain respectively.[1 -- 9] Later, local measurements took place by means of
piezoresponse force microscopy (PFM).[3] These measurements that were done at the sub-
micrometer scale, i.e. relevant for miniaturized devices, revealed an imprint effect or longitudinal
shift of the hysteresis loop rather than rotation.[10] For instance, individual PZT capacitors were
examined e.g. for 1.5 x 1.5 µm devices of 200-nm thick PZT films grown on Si.[11] The domain
distribution and hysteretic behavior of the ferroelectric film were tested with a local probe, by
means of piezoresponse force microscopy (PFM) before and after bending the Si substrate by
using a holder with a fixed-curvature radius of 30 cm. It was shown that flexing the material
results in domains with a single-polarization and a heavily imprinted state, compressive strain
inducing positive imprint (hysteresis loop shifts to lower voltages), while tensile strain induces
negative imprint (hysteresis loop shifts to the higher voltage values). This observation was then
confirmed[12] when a three-point bending stage was used to strain a similar system of PZT on a
thinned substrate as well as with dog-bone straining stage.[13] Despite the clear significance of
these experiments, rigorous testing of the effects of strain on functionality requires a careful
choice of both the flexing method as well as the material examined, mainly because thinned
substrate are typically not stable mechanically, while not all substrates can be thinned
reproducibly. Moreover, because structural and functional properties are coupled
in
ferroelectrics, characterizing the hysteresis loop is not sufficient. Rather, the crystallographic
structure of the material also needs to be examined under the same strain conditions. Moreover,
hysteresis loop is not the only important functional property and also domain stability of the
flexed films also has to be characterized.
Among the flexing methods, four-point stressing produces the most homogeneous strain
distribution over a
long range, enabling characterization of the functional properties
independent of the position of the PFM probe as well as under the larger-scale XRD beam. The
applied stress (σ) and hence the induced strain (ε) along a beam according to the torque behavior
is described as follows[14,15]:
𝜎 =
𝑀𝛿
2𝐼
=
3𝐹𝐿1
𝑏𝛿
(1),
where 𝑀 =
𝐹𝐿1
4
is the torque that is applied to the load span 𝐿1 (i.e. the distance between the
two inner supports of the four stressing points), 𝐹 is the force, 𝛿 and 𝑏 are respectively the
thickness and width of the bent sample and 𝐼 =
𝑏𝛿3
12
is the moment of inertia. The strain is then
extracted from Hook's law: 𝜖 =
𝜎
Ε
where Ε is Young's modulus. In four-point bending, the induced
torque is constant and the corresponding strain is thus homogeneous along the load span (see
Figure 1a). However, applying four-point bending in-situ while performing PFM characterization
is a complex task and thus far this method has been used only to characterize ferroelectrics in
the bulk form.[16]
Figure 1 Tunable homogeneous strain with in-situ four-point bending. (a) Schematic
illustration of a four-point bending setup (top) and the resultant homogeneous torque and strain
distribution at the load span, i.e. between the inner supports (bottom). (b) Optical photo of 1×1
cm2 PZT-SRO-mica in the four-point bending stage that was fabricated by a 3D metallic printer to
allow in-situ tunable straining during structural (XRD) and functional (PFM) characterizations
(contrast difference in the PZT layer is due to inhomogeneous bottom silver paste spread that is
observable through the transparent sample).
In addition to extrinsic strain, thin ferroelectric films experience intrinsic strain that must be
considered as well. Thin films of perovskite ferroelectrics exhibit typically inferior functional
characteristics with respect to their bulk form because they are inherently strained by the
substrate. Significant effort has been put in understanding and controlling the effect of substrate-
film lattice mismatch during the film growth on key properties, such as piezoresponse,
polarization retention, polarization imprint and coercivity, with an emphasize on the commonly
used perovskite ferroelectric, PZT.[13,17 -- 19] It is possible to reduce the effects of substrate clamping
by growing thicker films or by lithographically patterning isolated island structures.[5,7,20 -- 24]
However, these methods require special treatments and processing steps, which complicates the
integration of ferroelectric-based devices in existing technologies.[23] Another strategy that has
proved useful for reducing substrate clamping and maintaining the high functional performance
of ferroelectrics even as thin films is choosing a substrate with minimal lattice mismatch for the
ferroelectric film.[1,25] Nevertheless, because both the film and the substrate are brittle the
technological applicability of this strategy is not ideal for flexible devices. Moreover, although the
substrate lattice constant can be engineered rather freely to produce a broad range of substrate-
film lattice mismatch, this method does not allow strain tunability in a given film, suggesting that
direct examination of the effect is limited. Moreover, most commercially relevant ferroelectric
films are deposited at high temperature as paraelectrics and they undergo a structural phase
transition when being cooled down to the ferroelectric state, so that the lattice matching can be
engineered either to the paraelectric or to the ferroelectric state, but not to both. Hence, having
a suitable material (films and substrates) and a stressing method that allow in-situ tunability of
homogeneous strain is attractive both for understanding the effects of substrate clamping on
ferroelectricity as well as for integrating ferroelectrics in flexible and foldable electronic devices.
Nevertheless, to-date, experimental realization of the simultaneous extrinsic (flexing) and
intrinsic (substrate clamping) effects of strain for the sake of understanding the effects of strain
and structure on ferroelectric functionality in a given material is still lacking.
Here, we developed a method for straining a flexible ferroelectric film homogeneously with
a four-point bending stage and characterizing both its structural and functional properties in-situ.
We used this method to tune the substrate-film lattice mismatch while straining the material in
the XRD. We then transferred the sample while maintaining the same strain conditions to
perform PFM
imaging and spectroscopic characterization. We demonstrated that by
compensating the substrate clamping with external strain the ferroelectric functional properties
are enhanced.
Experimental
The samples investigated in this work are 30-nm thick PbZr0.2Ti0.8O3 (PZT) films grown on a
flexible freshly cleaved mica muscovite substrate (30-m thick) by pulsed laser deposition
(PLD).[26] A CoFe2O4 seeding layer has been used to form a Van de Waals quasi-epitaxy structure
(atomic layer of interdiffusion was found, please see details elsewhere[26,27]), while an
intermediate layer of 100-nm thick SrRuO3 (SRO) bottom electrode was deposited below the PZT
film. The tetragonal PZT composition was chosen to be far away from the morphotropic transition
for identifying confidently the effects of structural changes and substrate-film lattice matching
on the ferroelectric properties. To apply homogeneous tensile strain on the film, we developed
and fabricated a four-point-bending stage that allows in-situ straining while performing
nanoscale PFM imaging and spectroscopy and XRD structural characterization as well as
transferring the sample between the two, while retaining the same strain conditions (see Figure
1b). This method allowed us to strain the sample during the XRD characterization, and once the
obtained strain was as required, the sample was transferred to the PFM while remaining in the
holder and subject to the same strain. Hence, in-situ strain tunability was obtained as well as
reliable cross-characterization of the structural-functional properties. The bending stage was
manufactured using a novel 3D metal-printer (Arcam A2Xm) using Ti6Al4V titanium alloy, which
has high mechanical strength, even for patterns as small as 0.6 mm as well as high melting and
creep temperatures and high electrical conductivity. The method is based on melting of pre-
alloyed metal powder particles by an electron beam.
A comparison between the functional properties of the unstrained and strained PZT thin
films was done by means of PFM imaging and PFM spectroscopy (Asylum Research, MFP
Infinity).[28] The measurements were performed with Si cantilevers coated with titanium silicide,
with nominal force constant of 2.4 N∙m-1 and 70 kHz resonance frequency. The local amplitude
and phase hysteresis loops were performed by positioning the PFM tip at various points of the
surface and cycling a DC voltage between -10 V and +10 V with 0.1 V increments, superimposed
with the AC PFM imaging.[29] Each loop was averaged over at least 85 consecutive cycles. The
crystal structure properties were determined with a Rigaku SmartLab 9 kW high-resolution
diffractometer. A Cu kα rotating-anode source at 45 kV tube voltage was used, with a 200-mA
tube current as well as a 0-dimension silicon drift detector. Reciprocal space mappings (RSM)
were done with a Ge(220)X2 monochromator in parallel-beam geometry. The peaks' positions
were found using Rigaku 3D Explore software, where the peaks fit to a 2D Gaussian curve.
Results and discussion
The procedure we used to corroborate the structural and functional properties of the same
individual samples in unstrained and strained states while applying in-situ external stress was as
follows. First, we determined domain stability, coercive field and remanent piezoresponse by
PFM imaging and local hysteresis measurements of unstrained samples. Secondly, we used the
RSM to measure the lattice parameters of both the PZT and SRO films, allowing us to extract the
PZT-SRO lattice mismatch (
𝑎𝑙−𝑎𝑠
𝑎𝑠
) as well as the PZT tetragonality (
𝑐𝑙
𝑎𝑙
), where 𝑎𝑠 is the SRO lattice
parameter and 𝑎𝑙, and 𝑐𝑙 are the short- and long- axis PZT lattice parameters. Thirdly, using the
bending stage, we varied the PZT-SRO lattice mismatch by tuning the strain in the sample while
performing XRD profiling. Finally, once we minimized the lattice mismatch, we transferred the
sample to the PFM and characterized the strained samples by both imaging and hysteresis
measurements. This iterative process was performed when tuning a given sample controllably
in-situ.
Structure (x-ray and RSM)
The out-of-plane 𝜃 − 2𝜃 scans were measured (see Figure SI1), indicating that the major
planes parallel to the sample surface are PZT(111) and SRO(111) on mica(001). To determine the
crystallographic orientation and lattice parameters we performed an RSM scan. Figure 2a shows
asymmetric reciprocal space mapping of PZT(310), PZT(103) and SRO(301) of the unstrained
sample. No mica peaks were observed around these reciprocal-space coordinates. The peaks of
SRO and PZT were both aligned to 𝑄𝑥, which is the direction parallel to the film in the reciprocal
space. That is, the RSM data confirm that the PZT layer was laid in epitaxial registry with the SRO
layer. The lattice parameters of both the PZT and the SRO were extracted from the relationship
between the inter-planar spacing 𝑑ℎ𝑘𝑙 and the Miller indices (ℎ, 𝑘 and 𝑙):
1
𝑑(ℎ𝑘𝑙)
2 =
ℎ2+𝑘2
2 +
𝑎𝑙
𝑙2
2 (2),
𝑐𝑙
allowing us to extract both the film-substrate lattice mismatch and the PZT tetragonality (for
extracting the SRO lattice parameters, we replaced in this expression both 𝑎𝑙 and 𝑐𝑙 with 𝑎𝑠). We
then strained the sample in-situ with the four-point bending stage and extracted the
corresponding change in lattice parameters by mapping the RSM signal around two different
orientations. Figure 2b shows the RSM signal of the strained sample around the PZT(111)
orientation (i.e. a symmetric RSM, which allowed us to confirm the hetero-epitaxial mosaic
structure of the PZT-SRO layers that are Van der Waals-bonded attached to the mica flakes
substrate.[26] Likewise, Figure 2c shows the RSM signal of the same sample, under the same
tensile strain, but around the PZT(310) orientation (i.e. asymmetric RSM).
Figure 2 Reciprocal space mapping of a flexible PZT film without and with strain. (a)
Asymmetric RSM of an unstrained PZT-SRO film on mica. (b) Symmetric and (c) asymmetric RSM
of the same material that was now strained in-situ. The position of the {310} and (111)
orientations within the reciprocal space axes parallel (𝑄𝑥) and perpendicular (𝑄𝑧) to the film
allows quantitative calculation of the PZT and SRO lattice parameters at both the unstrained and
strained states (see Table 1).
Table 1 Effects of strain on lattice parameters, film-substrate lattice mismatch and
tetragonality. Lattice parameters of the PZT layer and SRO substrate with and without in-situ
strain that were extracted from the corresponding RSM measurements (Figure 2). The reduction
of PZT-SRO lattice mismatch and increase in PZT tetragonality in the strained state are also
presented.
PZT-SRO lattice mismatch
Tetragonality
PZT film state 𝒂𝒍(Å)
𝒄𝒍(Å) 𝒂𝒔(Å)
Unstrained
3.957
4.106 3.93
Strained
3.954
4.170 3.95
𝑎𝑙−𝑎𝑠
𝑎𝑠
[%]
0.67
0.15
𝑐𝑙
𝑎𝑙
1.038
1.054
The lattice parameters extracted for the PZT and SRO layers of the same sample both
unstrained and under strain are given Table 1. Using these parameters, we calculated both the
PZT-SRO lattice mismatch and the PZT tetragonality (note that here, the tetragonality is used as
a figure of merit to describe the strain in the film). Table 1 also shows that straining the samples
reduced the lattice mismatch to less than a quarter of its original values with respect to the
unstrained film. Likewise, straining the sample increased the PZT tetragonality by more than 1.5
at %.
Functionality (PFM imaging and spectroscopy)
To determine the effects of strain on ferroelectricity we tested both the domain stability as
well as the spectroscopic switching parameters. First, we imaged engineered domains (areas
scanned with positive and negative voltages that exceed the coercive value) of the unstrained
and of the strained sample. The imaging was done by means of amplitude and phase PFM signals
of the vertical (out-of-plane) and lateral (in-plane) piezoresponse (typically, proprtional to the
polarization) mapping as seen in Figure 3.
Figure 3 Polarization distribution of domains in flexible PZT film without and with extrinsic
strain immediately after patterning. (a) Out-of-plane PFM amplitude and (b) phase signals and
the simultaneously imaged (c) in-plane PFM amplitude and (d) signals showing the polarization
distribution in an area that was pre-patterned by applying voltage between the tip and SRO layer
as illustrated in the map in (e). (f) The simultaneously imaged topography of the same area is
given as a reference. (g) The corresponding PFM out-of-plane amplitude and (h) phase as well as
(i) in plane amplitude and (j) phase signals demonstrating the polarization distribution in a pre-
patterned area, while the film experienced in-situ strain. (k) The domain patterning map and (l)
topography of the area of the strained film experiment.
To determine the domain stability, we imaged the relaxation dynamics of the engineered
domains (Figures 3b and 3g) over time. The vertical PFM phase image of the relaxing domains in
both the strained (Figure 4a-d) and unstrained (Figure 4f-i) show a significant change in the
domain relaxation behavior. The unstrained domains relaxed already after about one and a half
hours, while the strained sample, with the engineered domains remained stable even after
twenty one hours in the strained sample with the reduced PZT-STO lattice mismatch. The native
domain distribution of the unstrained and strained samples are given as a reference (Figures 4e
and 4j respectively), showing no significant difference between them. To illustrate the clear
increase in domain stability for the strained samples, we plotted in Figure 4k the evolution of the
percentage of down-polarization domains (dark areas in Figure 4) over time. The native domain
distribution of the unstrained and strain films are also given in Figure 4a and 4f, respectively.
Figure 4k shows that the native domain were randomly distributed for the unstrained sample
and have a preferable 'down' orientation for the strained samples (corresponding to the data
points that are marked with 'x').
Figure 4 Domain relaxation in a PZT film with and without external strain. (a)-(j): Out-of-plane
PFM phase signal of the (a) native domain distribution in an unstrained film. (b) The same area
immediately after domain patterning, following the scan map in Figure 3e. (c) The domain
evolution in this area 12 min, (d) 23 min and (e) 87 min after the completion of the imaging in (b)
showing significant relaxation of the patterned domains. After straining the film (see Table 1),
the out-of-plane PFM signal was recorded in a different area to show the (f) native domain
distribution as well as (g) the polarization distribution immediately after domain patterning,
following the scan map in Figure 3k. The areas were then scanned repeatedly, demonstrating
that the domain distribution was almost unchanged after (h) 230 min, (i) 537 min and (j) 1260
min. The percentage of area with down polarization (dark areas in a-g) a function of time is
plotted (k) to compare the domain stability in the unstrained and strained scenarios. Filled red
circle and the blue square correspond to relaxation in the unstrained and flexed sample,
respectively. Empty data points at time '0' designated with 'x' represent the native domain
distribution.
Next, to complete the examination of the effects of strain on the functional properties of
the ferroelectric material, we measured the switching parameters of the hysteresis loop in the
PZT film. Local hysteresis loop measurements were done in a set of 8×7 points spread in a 25×25
m2 area. At each point, four consecutive switching cycles were performed. For examining the
effect of strain on the hysteresis loops, the entire experiment was done both for the unstrained
and strained sample with the same experimental conditions (same cantilever, voltage-sweeping
conditions etc.) with only a few hours gap between the measurements of the unstrained and
strained sample.
Figure 5a shows the average butterfly (PFM amplitude) hysteresis loops of the unstrained
(red circles) and strained (blue squares) states of the same film. The negative and positive
coercive voltages measured from the amplitude butterfly hysteresis-loops are -5.75 V and 6.70 V
for the unstrained state, and -3.60 V and 5.4 V for the strained state. That is, straining the
ferroelectric films gives rise to reduction in the coercive voltage value with respect to the native
unstrained state with the larger film-substrate lattice mismatch. We also compared the
piezoresponse hysteresis loops (amplitude times cosine the phase of the PFM signal) of the
unstrained and strain states of the ferroelectric film as seen in Figure 5b. Here, we clearly see
that the straining the sample not only has increased the saturation values of the piezoresponse
signal (which is proportional to the polarization) and increased the area in the hysteresis loop,
but also made the hysteresis loop more symmetric than the unstrained state with respect both
to the remanent piezoresponse and coercive voltage. The parameters extracted from the
switching spectroscopy are given in Table 2. We should note that Figures 5a and 5b are the
average of the last three (out of four) consecutive measurements that were done in each point.
For the sake of completeness, the average of the first switching cycle for all the 8×7 points of the
unstrained and strained cases are given for both the butterfly (PFM amplitude signal, Figure 5c)
and piezoresponse (PFM amplitude times the cosine of the PFM phase signal, Figure 5d) are given
as a reference.
Figure 5 Hysteresis loop measurements of the PZT film both unstrained and under in-situ
strain. (a) Butterfly (PFM amplitude signal) and (b) piezoresponse (amplitude times cosine the
phase signal) hysteresis loops measured in the unstrained (full red circles) PZT and under in-situ
strain (empty blue squares). Each hysteresis curve is an average of the last three (out of four)
switching cycles from 56 different locations. The averaged (c) butterfly and (d) piezoresponse
hysteresis loops of the first switching measurements in these 56 points is given as a reference.
Table 2 Functional properties of unstrained and strained PZT. The switching parameters that
are extracted from the averaged hysteresis loops (see Figure 5) of the PZT film both without and
with external strain. The positive and negative coercive values, positive and negative saturation
piezoresponse (which is proportional to the saturation polarization) and remanent voltage were
extracted (see highlighted data points in Figure 5).
Positive coercive voltage [V]
Negative coercive voltage [V]
Positive piezoresponse saturation [a.u.]
Negative piezoresponse saturation [a.u]
Remanent voltage [V]
Unstrained
6.70
-5.75
1.68
-2.21
0.77
Strained
5.4
-3.60
2.120
-3.320
1.00
Figure 6 Functional properties of PZT films as a functional of their tetragonality: comparison
with literature.[30 -- 33] (a) Remanent polarization and remanent piezoresponse vs. tetragonality.
(b) Coercive field and coercive voltage vs. tetragonality. (See Table SI1 for details).
Our results show a clear relationship between the structure and functionality of
ferroelectrics by means of direct observations. Reducing the film-substrate lattice mismatch
between the PZT film and the SRO sample is accompanied by significant increase in stabilization
of switched ferroelectric domains (retention). Reduction of this lattice matching is also
accompanied by reduction of the asymmetry of the ferroelectric hysteresis loop (imprint and
surface charging). Straining the sample led also to another change in the structural properties --
increase in the PZT tetragonality, which in turn was accompanied by an increase of the remanent
piezoresponse as well as the area integrated within the hysteresis loop. We suggest that these
observations are in agreement with the existing literature related to the effects of either lattice
matching engineering (by means of varying the substrates)[1,33 -- 35] or in-situ bending[11,12,36] on the
hysteresis loop and stability of switched domains. However, the in-situ strain tunability during
structural and functional characterization allows bridging between these two effects of substrate
clamping and flexing in ferroelectrics, i.e. between the effects of intrinsic and extrinsic strain. The
observed change in the ferroelectric properties under flexing conditions of the PZT films may
open some technological applications for ferroelectrics as flexible electronics. We should note
though that the effects of flexing on the ferroelectric properties may vary for different PZT
compositions. For instance, existing literature suggests that PZT near the morphotropic transition
may exhibit a larger degree of stability in the ferroelectric properties when flexed.[26] To assess
the effects of extrinsic strain reported here, we compared it to the effect of intrinsic strain for
the switching properties. Figure 6 shows literature data of the dependence of remanent
polarization as well as coercive field on tetragonlity for strain obtained by varying the substrate,
the PZT composition or the PZT temperature (raw data are given in Table SI1). These data is
compared to the dependence of remanent piezoresponse and coercive voltage as a function of
tunable tetragonality, which is reported here. This comparison shows that trend of the
dependence of these values on the tetragonality varies between the extrinsic strain and the
intrinsic strain. The exact origin of this opposite trend is still not completely known to us, mainly
with respect to the coercive value. A possible explanation is that as mentioned above, in addition
to change in tetragonlity, the strain changed also the lattice-substrate lattice mismatch, which
may play a role in these two values. However, a comparison that accounts for both changes in
tetragonality and substrate clamping is not straightforward, while not all data do not yet exist in
the literature. We therefore encourage further experimental and theoretical investigation of
different ferroelectric materials and compositions for flexible-electronic purposes. Finally, we
hope our work will encourage further examinations of the relationship between structure and
nanoscale functionality of materials that are subject to in-situ tunable strain.
Acknowledgements
This research was supported by a grant from the Ministry of Science & Technology, Israel
& the Ministry of Science and Technology of Taiwan. The Technion group acknowledges also
financial support from the Zuckerman STEM Leadership Program. The authors would like to thank
Mr. Garry Muller from the Israeli Institute of Metals for assistance with the 3D printing of the 4-
point bending stage as well as Mr. Shimon Cohen and Mr. Alon Hendler Avidor for earlier
discussion regarding the design of this stage. Moreover, we thank Dr. Maria Koifman and Prof.
Semën Gorfman for their assistance with the XRD and RSM characterizations and data
interpretation.
References
[1] O. Nesterov, G. Rispens, J. A. Heuver, H. M. Christen, B. Noheda, S. Matzen, M. Biegalski,
Nat. Commun. 2014, 5, 4415.
[2] N. Ledermann, P. Muralt, J. Baborowski, S. Gentil, K. Mukati, M. Cantoni, A. Seifert, N.
Setter, Sensors Actuators, A Phys. 2003, 105, 162.
[3] G. L. Brennecka, W. Huebner, B. A. Tuttle, P. G. Clem, J. Am. Ceram. Soc. 2004, 87, 1459.
[4] M. B. Kelman, P. C. McIntyre, B. C. Hendrix, S. M. Bilodeau, J. F. Roeder, J. Appl. Phys.
2003, 93, 9231.
[5]
J. F. Shepard, P. J. Moses, S. Trolier-McKinstry, Sensors Actuators A Phys. 1998, 71, 133.
[6]
J. F. Shepard, F. Chu, I. Kanno, S. Trolier-McKinstry, J. Appl. Phys. 1999, 85, 6711.
[7] G. H. Haertling, J. Am. Ceram. Soc. 1999, 82, 797.
[8] H. Nakaki, Y. K. Kim, S. Yokoyama, R. Ikariyama, H. Funakubo, K. Nishida, K. Saito, H.
Morioka, O. Sakata, H. Han, S. Baik, J. Appl. Phys. 2009, 105, 14107.
[9] M. D. Nguyen, M. Dekkers, H. N. Vu, G. Rijnders, Sensors Actuators, A Phys. 2013, 199,
98.
[10] P. Muralt, A. Kholkin, M. Kohli, T. Maeder, Sensors Actuators A Phys. 1996, 53, 398.
[11] A. Gruverman, B. J. Rodriguez, C. Dehoff, J. D. Waldrep, A. I. Kingon, R. J. Nemanich, J. S.
Cross, Appl. Phys. Lett. 2005, 87, 82902.
[12] A. Alsubaie, P. Sharma, G. Liu, V. Nagarajan, J. Seidel, Nanotechnology 2017, 28, 075709.
[13] S. Yagnamurthy, I. Chasiotis, J. Lambros, R. G. Polcawich, J. S. Pulskamp, M. Dubey, J.
Microelectromechanical Syst. 2011, 20, 1250.
[14] X. Yan, W. Huang, S. Ryung Kwon, S. Yang, X. Jiang, F. G. Yuan, Smart Mater. Struct. 2013,
22, 085016.
[15] K. Prume, P. Muralt, F. Calame, T. Schmitz-Kempen, S. Tiedke, J. Electroceramics 2007,
19, 407.
[16] P. Zubko, G. Catalan, A. Buckley, P. R. L. Welche, J. F. Scott, Phys. Rev. Lett. 2007, 99,
167601.
[17] A. Gruverman, B. J. Rodriguez, A. I. Kingon, R. J. Nemanich, A. K. Tagantsev, J. S. Cross, M.
Tsukada, Appl. Phys. Lett. 2003, 83, 728.
[18] M. Alexe, C. Harnagea, D. Hesse, U. Gösele, Appl. Phys. Lett. 2001, 79, 242.
[19] J. Lee, R. Ramesh, Appl. Phys. Lett. 1996, 68, 484.
[20] F. Griggio, S. Jesse, A. Kumar, O. Ovchinnikov, H. Kim, T. N. Jackson, D. Damjanovic, S. V
Kalinin, S. Trolier-Mckinstry, Phys. Rev. Lett. 2012, 108, 157604.
[21] V. Nagarajan, A. Roytburd, A. Stanishevsky, S. Prasertchoung, T. Zhao, L. Chen, J.
Melngailis, O. Auciello, R. Ramesh, Nat. Mater. 2002, 2, 43.
[22] J. Cardoletti, A. Radetinac, D. Thiem, J. Walker, P. Komissinskiy, B.-X. Xu, H. Schlaak, S.
Trolier-McKinstry, L. Alff, AIP Adv. 2019, 9, 25017.
[23] M. Wallace, R. L. Johnson-Wilke, G. Esteves, C. M. Fancher, R. H. T. Wilke, J. L. Jones, S.
Trolier-McKinstry, J. Appl. Phys. 2015, 117, 54103.
[24] A. Palacios-Laloy, F. Nguyen, F. Mallet, P. Bertet, D. Vion, D. Esteve, J. Low Temp. Phys.
2008, 151, 1034.
[25] E. Snoeck, B. Noheda, D. H. A. Blank, A. Lubk, A. Janssens, G. Rijnders, C. Magen, G.
Rispens, G. Catalan, A. H. G. Vlooswijk, Nat. Mater. 2011, 10, 963.
[26] J. Jiang, Y. Bitla, C.-W. Huang, T. H. Do, H.-J. Liu, Y.-H. Hsieh, C.-H. Ma, C.-Y. Jang, Y.-H. Lai,
P.-W. Chiu, W.-W. Wu, Y.-C. Chen, Y.-C. Zhou, Y.-H. Chu, Sci. Adv. 2017, 3, e1700121.
[27] Y.-H. Chu, npj Quantum Mater. 2017, 2, 67.
[28] S. B. Lang, H. L. W. Chan, Frontiers of ferroelectricity : a special issue of the Journal of
materials science; Springer: New York, 2007, 1-288.
[29] A. Gannepalli, D. G. Yablon, A. H. Tsou, R. Proksch, Nanotechnology 2011, 22, 355705.
[30] H. Watanabe, T. Mihara, C. A. Paz De Araujo, Integr. Ferroelectr. 1992, 1, 293.
[31] Y. Sakashita, T. Ono, H. Segawa, K. Tominaga, M. Okada, J. Appl. Phys. 1991, 69, 8352.
[32] H. Maiwa, S.-H. Kim, N. Ichinose, Appl. Phys. Lett. 2003, 83, 4396.
[33] N. Floquet, J. Hector, P. Gaucher, J. Appl. Phys. 1998, 84, 3815.
[34] Y. Ivry, D. Chu, C. Durkan, Appl. Phys. Lett. 2009, 94, 162903.
[35] Y. Ivry, C. Durkan, D. Chu, J. F. Scott, Adv. Funct. Mater. 2014, 24, 5567.
[36] Y. Ivry, V. Lyahovitskaya, I. Zon, I. Lubomirsky, E. Wachtel, A. L. Roytburd, Appl. Phys. Lett.
2007, 90, 172905.
Supporting Information
Figure SI1 (a) Broad and (b) zoom-in range 𝜃 − 2𝜃 x-ray scans of the unstrained (red) and
strained (blue) PZT-SRO-mica heterostructure.
Table SI1 Structure-functionality relationship as extracted from this work, in comparison to
the existing literature, following Figure 6.
Tetragonality
Ec
Vc
Pr
[kV/cm]
[V]
[uC/cm^2]
𝒂𝒍
[Å]
𝒄𝒍
[Å]
3.965
4.109
3.98
4.082
4.005
4.045
4
4.14
4.01339
4.13
0.1
0.2
0.3
25
100
140
1.036
1.025
1.010
1.035
1.029
80
72
58
68.55
64
Composition
[Zr %wt][30]
Temperature [⁰C]
[31,32]
Substrate[33]
External strain
11
8
9
28
27.5
25
22
17
5.38
6.71
2.23 *
2.17*
4.0161 4.1308
1.028
49.21
STO (100)
MGO(100)
_
_
_
_
Unstrained
3.957
4.106
Strained
3.954
4.17
1.038
1.032
1.037
1.054
110
110
|
1811.12609 | 1 | 1811 | 2018-11-30T04:41:03 | Invited Article: 4D Printing as a New Paradigm for Manufacturing with Minimum Energy Consumption | [
"physics.app-ph"
] | 4D printing is a new manufacturing paradigm that combines stimuli-responsive materials, mathematics, and multi-material additive manufacturing to yield encoded multi-material structures with intelligent behavior over time. This emerging field has received growing interests from various disciplines such as space exploration, renewable energy, bioengineering, textile industry, infrastructures, soft robotics, and so on. Here, as a first attempt, we consider the energy aspect of 4D printing. By a thermodynamic analysis, we obtain the theoretical limit of energy consumption in 4D printing and prove that 4D printing can be the most energy-efficient manufacturing process. Before that, we clearly underpin 4D printing as a new manufacturing process and identify its unique attributes. | physics.app-ph | physics | Invited Article: 4D Printing as a New Paradigm for
Manufacturing with Minimum Energy Consumption
Farhang Momeni 1,* and Jun Ni 1
1 Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI 48109, US A .
* Correspondence: [email protected]
Featured Application: Carnot cycle yields the theoretical limit of the energy efficiency in "heat
engines". Here, we obtained the theoretical limit of minimum energy consumption in
"manufacturing processes" that can be approached by 4D printing .
Abstract: 4D printing is a new manufacturing paradigm that combines stimuli-responsive materials,
mathematics, and multi-material additive manufacturing to yield encoded multi-material structures
with intelligent behavior over time. This emerging field has received growing interests from various
disciplines such as space exploration, renewable energy, bioengineering, textile industry,
infrastructures, soft robotics, and so on. Here, as a first attempt, we consider the energy aspect of
4D printing. By a thermodynamic analysis, we obtain the theoretical limit of energy consumption
in 4D printing and prove that 4D printing can be the most energy-efficient manufacturing process.
Before that, we clearly underpin 4D printing as a new manufacturing process and identify its unique
attributes.
Keywords: energy efficiency; thermodynamics; advanced manufacturing; energy consumption;
theoretical limit; 4D printing
1. Introduction
Manufacturing industries consume about one-third (31%) of the global energy and are also
responsible for approximately one-third (36%) of CO2 emissions [1 -- 4]. Energy availability and costs
are the next issues in addition to environmental impacts [ 5,6]. Therefore, energy efficiency has been
the focus of many studies recently, and its eminence has been highlighted more than ever [ 7,8].
On the other hand, 4D printing [9 -- 14] provides a situation for using random (free) energy to
make non-random structures [9]. This goal is achieved by utilizing and encoding smart (stimuli-
responsive) materials in multi-material structures. 4D printing can revolutionize manufacturing and
construction industries [9]. This emerging field has received growing interests from various
disciplines such as space exploration [15 -- 17], renewable energy [18,19], bioengineering [20], textile
industry [21,22], soft robotics [23,24], electrical circuits [25], metamaterials [26 -- 28], kirigami [29],
origami [30], supercapacitors and batteries [31], infrastructures [9], and so on.
In the following, first, we underpin 4D printing as a new manufacturing process and identify its
unique attributes. Second, we organize various concepts of assembly unveiled over time. Third, we
introduce principles of self-assembly at manufacturing scale as it is in the early stage of development
and convergence of ideas onto the right direction is required. Fourth, we obtain the theoretical limit
of minimum energy consumption in manufacturing that can be approached by 4D printing.
2. 4D printing as a new manufacturing process with unique attributes
3D printing (additive manufacturing) is a well-known manufacturing process with its unique
attributes. Now, 4D printing needs to be clearly defined and described as a new manufacturing
process and its unique attributes should also be proved.
By analyzing natural shape morphing materials and structures [ 32], in addition to stimuli and
stimuli-responsive materials, one other thing is observed that is encoded anisotropy [32]. To enable
1
complicated shape-shifting behaviors required for accomplishing various tasks in natural structures
such as the pinecone, nature programs a specific arrangement of active and passive elements [32].
This encoded anisotropy is required to direct the response into the desired direction [32]. Now, 4D
printing is a good paradigm to meet this type of encoding in synthetic shape morphing structures.
By considering the aforementioned point in natural shape morphing structures and based on the 4D
printing concepts discussed in our previous review article [ 14], as well as some of the related works
[9 -- 12], we underpin 4D printing as a new manufacturing process as shown in Figure 1. We also
identify attributes of 4D printing as a new manufacturing process as Figure 2. Almost all applications
enabled by 4D printing can be categorized into self-assembly, self-adaptability, and self-repair [14]
that we call them here as "3S 4D printing applications" (Figure 2). 4D printing conserves the
advantages of 3D printing and further adds new features (Figure 2). The "complexity-free geometry"
attribute was introduced as the unique feature of structures made by 3D printing [33 -- 36]. Here, we
introduce "complexity-free geometry change" as the unique attribute of 4D printed structures. One
of the key goals of 3D printing is movement from form to functionality [37]. 4D printing goes further
and provides multi-functionality. Furthermore, 4D printing possesses the material-saving
characteristic arising from the general advantages of 3D printing and further adds the energy-saving
trait (that is the focus of this study).
Proposal for future: it should be highlighted that future "4D printers" should possess an
integrated inverse mathematics (as a software/hardware added to the current multi-material 3D
printers) to predict the shape-shifting behaviors for various (or categories of) materials and stimuli.
The term "4D printer" has already been mentioned in various studies in the literature of 4D printing.
However, "4D printer" is not simply achieved by extending a single-material 3D printer to a multi-
material 3D printer, or by combining several printing techniques (e.g., FDM and inkjet) in one 3D
printer. We should say that "4D printer" should be able to analyze and predict the "4 th D". To achieve a
4D printer, an "intelligent head" should be developed and added to the current printers (Figure 3).
This head (as an integrated software/hardware added to the existing multi-material 3D printers)
should be able to analyze and predict the 4 th D. It should be able to predict the appropriate
arrangement of active and passive materials (an encoded anisotropy) for the desired evolution after
printing. As we elaborated in our previous work [14], 4D printing mathematics is a link between four
main factors: printing path (arrangement of active and passive voxels), desired shape after printing,
stimulus properties, and materials properties. 4D printing mathematics is required to predict the
shape-shifting behavior after printing over time, prevent internal collisions, and decrease or even
eliminate trial-and-error tests for getting the desired shape-shifting [14]. Currently, 4D printing
process utilizes the inverse mathematical modeling in an offline manner (passively), as seen in Figure
1. However, the inverse mathematical modeling can systematically be incorporated into current 3D
printers to yield 4D printers that can analyze and predict the 4 th D.
2
Figure 1. 4D printing process.
3
Multi-material 3D printerInverse mathematics (to encode multi-materials during printing) 4DStimulus: can be a free energy in the environmentShape, property (other than shape), or functionality evolution over Time 4D printing processInteraction mechanismSelf-assembly to reach the final desired shape Self-adaptability between at least two different shapesSelf-repair (by self-disassembly and self-healing)oror
Figure 2. "3S 4D printing applications" and 4D printing attributes.
Figure 3. Future 4D printers. The term "4D printer" has already been mentioned in some studies.
However, "4D printer" is not simply achieved by extending a single -material printer to a multi-
material printer, or by combining several printing techniques (e.g., FDM and inkjet) in one 3D printer.
"4D printer" should be able to analyze and predict the "4th D". Thus, to achieve a 4D printer, an
"intelligent head" (i.e., an integrated software/hardware
problems of Figure 1) should be developed and added to the current multi-material 3D printers.
that incorporates inverse mathematic al
4
Self-adaptabilitySelf-assemblySelf-repair3S 4D printing Applications 1. Complexity-free geometry Originates from 3D printing. 2. Complexity-free geometry change Unique to 4D printing.1. Self-sensing (embedded sensor)By using stimuli-responsive materials during the printing.2. Self-actuating (embedded actuator)By combining active and passive materials in a special spatial arrangement (mainly by an inverse problem) to guide the mismatch-driven forces into the desired direction.1. Functionality Originates from 3D printing.2. Multi-functionality Unique to 4D printing.Geometry attributesFunctionality attributesSensing and actuating attributes1. Material saving Originates from 3D printing.2. Energy saving Unique to 4D printing.Material and energy attributesIntelligent head:an integrated software/hardware that incorporates inverse mathematical problems and is added to the current multi-material printers to predict and analyze the 4th D. Future 4D printers3D printer4D printer
3. (R)evolution of assembly concept in manufacturing
In this section, we analyze the literature [38 -- 50,9 -- 14] and organize various concepts of assembly
in manufacturing as seen in Figure 4.
A
S
S
E
M
B
L
Y
Design For
Assembly
(DFA)
Design For
No- (or Non-)
Assembly
(DFNA)
Design For
Self-
Assembly
(DFSA)
•
•
•
•
•
•
•
•
•
•
•
Approach 1.
Manual assembly: Human
assembles structures (Groover [38]).
Approach 2.
Automated assembly: robots
assemble structures (Groover [38]).
A hybrid of approaches 1 and 2 is
also possible.
An integrated structure with no
need for assembly (Kota et al.
[41 -- 43]).
Elaborated by 3D printing.
Type 1.
Separate components self-
assemble to the desired structure.
Elaborated by 3D or 4D printing
(if smart materials are used).
Type 2.
An encoded, multi-material, one-
time printed structure self-
assembles to the desired
structure, e.g. from a flat sheet to
a cube (Tibbits [9]).
Elaborated by 4D printing.
Can be combined with DFNA.
T
I
M
E
Figure 4. The (r)evolution of assembly concept in manufacturing .
4. Principles of self-assembly at manufacturing scale
Self-assembly is a process with low or no waste [51] and covers from the molecular- to the
galactic-scale applications [52]. Even though the self-assembly has been initiated at the molecular
scale [52], it can be more important for making products and structures at nano, meso, and macro
scales [53] that we call them here as manufacturing scale. By analyzing the principles of self-assembly
at the molecular scale [53], the concepts of assembly at manufacturing scale [40], and 4D printing
concepts [14], we introduce principles of self-assembly at manufacturing scale (covering both types I
and II discussed in Figure 4).
• Smart components
5
Physical components are needed for making structures through self-assembly [53]. In addition,
the components (whether separate components in self-assembly type I, or the multi-component
structure in self-assembly type II) should have smartness to self-assemble on their own. This
smartness can be embedded into structures by using smart materials and enabled by stimuli.
• Feature modeling
When two components (in self-assembly type I) or two parts of one multi-component structure
(in self-assembly type II) are planned to mate, at least, two features, each of which located on one of
the two parts are needed (Figure 5(b)) [40]. To have a systematic analysis for various purposes such
as achieving a specific distance or angle between two points of the assembled structure, the
assembling parts, and their features should be modeled by coordinate frames. After assigning
coordinate frames to each part (e.g., 𝐴) and its feature (e.g., 𝐹𝐴), the assembly process can be modeled
by matrix transformations (e.g., 𝑇𝐴−𝐹𝐴) to analyze translation and orientation of assembling parts
(Figure 5(b)) [40]. These features can be elaborated by the additive manufacturing.
• Variation analysis
In practice, various errors such as mislocations of features on parts, misshapen features, or
mating errors between features can occur [40]. Therefore, variation analysis, (in which 𝑇 ′ = 𝑇 𝑑𝑇 is
the central equation [40]), should be implemented in conjunction with the feature modeling (Figure
5(c)) [40].
• Constraint modeling
Each part in space can have three linear ( 𝜐𝑥, 𝜐𝑦 ,𝜐𝑧) and three angular (𝜔𝑥, 𝜔𝑦 ,𝜔𝑧 ) velocities
(motions). If a part is constrained in a linear/angular direction, it can resist a force (𝑓)/moment (𝑚) in
that direction. Otherwise, it has a motion in that direction. Screw Theory [ 54 -- 56] is a tool to
implement constraint/motion analysis by performing reciprocal, union, and intersection operations
on Twist and Wrench matrices of features (Figure 5(d)) [40]. This analysis is required to identify
under-constraint, over-constraint, and properly-constraint parts in an assembly and devise
appropriate decisions [40].
• Sequence analysis
Theoretically, several assembly sequences can be realized in an assembly, but not all of them are
feasible. First, the feasible assembly sequences should be obta ined by using one of the existing
methodologies such as AND/OR graph developed by De Mello and Sanderson (Figure 5(e)) [57,40].
Then, the best assembly sequence (s) can be chosen among the feasible ones based on technical and
business criteria [40]. The final thing that should be considered in assembly modeling is the Datum
Flow Chain (DFC) for extra filtering of assembly sequences and a systematic analysis of Key
Characteristics (KCs) (Figure 5(e)) [40].
• Stimulus
A stimulus such as heat, light, water, humidity, etc., which can be a free resource in the
environment is needed to trigger the self-assembly process.
•
Interactions
In a self-assembling system, interactions between separate components [53] in self-assembly
type I or between active and passive components (materials) in the multi-component structure in self-
assembly type II, as well as interactions between stimulus and component (s) should be analyzed.
• Mobility
6
Mobility under the right stimulus is required in self-assembling systems. At large scales, gravity
and friction forces should also be taken into consideration [ 53].
• Reversibility
"Self-assembly" is different from "formation" [52]. Reversibility is one of the primary
requirements for a self-assembly process that also allows for adjustability and error-correct [53].
The initial 4D printed structures [58,11,10] were generally reversible. However, their
reversibility could be purposely prevented by devising some self-locking mechanisms. These self-
locking mechanisms are specific arrangements of active and passive materials that are planned in
multi-material structures to intentionally prevent the reversibility when the s timulus is off.
Furthermore, in future, 4D printing can freely provide both the reversible and irreversible structures
[59], depending on needs.
•
Inverse problem
In the self-assembly at manufacturing scale, there is no human or robot to put the components
in the right positions (in self-assembly type I) or to bring various parts of a multi-component, one-
time printed structure into correct locations (in self-assembly type II). Therefore, an inverse problem
should be modeled by considering all the previous principles to achieve the final desired structure,
accurately.
7
𝑇𝐴 −𝐵 = 𝑇𝐴 −𝐹𝐴 𝑇𝐹𝐴 −𝐹𝐵 𝑇𝐹𝐵 −𝐵
𝑇′ = 𝑇 𝐷𝑇
= 𝑇𝐴−𝐹𝐴 𝐷𝑇𝐹𝐴 −𝐹′
𝐴
𝐴
𝑇𝐴 −𝐹′
Liaison
𝑇𝑖 = [𝜔𝑥 𝑖 𝜔𝑦𝑖
𝑊𝑖 = [𝑓𝑥 𝑖 𝑓𝑦𝑖
𝜔𝑧𝑖
𝑓𝑧𝑖
𝜐𝑦 𝑖
𝜐𝑥 𝑖
𝑚𝑥 𝑖 𝑚𝑦𝑖
𝜐𝑧𝑖 ]
𝑚𝑧𝑖 ]
Figure 5. The principles of self-assembly at manufacturing scale. By analyzing the principles of self-
assembly at the molecular scale [53], the concepts of assembly at manufacturing scale [40], and 4D
printing concepts [14], we introduce principles of self-assembly at manufacturing scale.
8
(b) Feature modeling(a) Smart components(f) Stimulus(h) MobilityType IType II(g) Interactionsµ Type IType II(c) Variation analysisgv, a(i) Reversibility(d) Constraint modelingScrew Theory x yzOperations: Reciprocal (recip), Union (U), and Intersection ( ) Twist (T) & Wrench (W) for feature i: (e) Sequence analysis{A, B, C}{B, C}{A}{B}{C}ABC{A, B}{A}{B}{C}: infeasibleFinal desired structure (that needs to be achieved on its own) Type I: Encoded componentsType II: An encoded multi-component (-material) structureInverse mathematical modeling (j) Inverse problemAND/OR graphDFC analysis -1 =...KC F BACF: Fixture: Mate: ContactA, B, C : PartsFeature BPart AFeature APart BAAFBFBTA-FATF -BBAFBFTF -FABBAInterface errors between two features Mislocation of feature A on part A Misshapen feature AAFKey equation in variation analysis:AFE.g.:variednominalA FAF
5. Energy aspect of 4D printing as a new process for self-assembly at manufacturing scale
A manufacturing process, in the most general form, can be modeled as Figure 6 [60 -- 64]. There
are three types of energy transfer mechanisms between a system and its surroundings: heat, work,
and mass flow (in fact, mass is energy, and mass flow is present in open thermodynamic systems)
[65]. The energy transfer between a system and its surroundings causes entropy transfer between
them so that 𝑆
𝑚𝑎𝑠𝑠 = 𝑚 𝑠 (there is no entropy transfer by work) [65].
Energy is conserved (i.e., it cannot be destroyed or generated), while entropy can be generated [65].
The first and second laws of thermodynamics deal with energy and entropy, respectively, and in the
most general forms are [65]:
𝑤𝑜𝑟𝑘 = 0, and 𝑆
ℎ𝑒𝑎𝑡 =
, 𝑆
𝑄
𝑇
Energy balance: 𝐸𝑖𝑛 − 𝐸𝑜𝑢𝑡 = ∆𝐸𝑠𝑦𝑠
{
Entropy balance: 𝑆𝑖𝑛 − 𝑆𝑜𝑢𝑡 + 𝑆𝑔𝑒𝑛 = ∆𝑆𝑠𝑦𝑠
rate form
⇒ 𝑆
𝑖𝑛 − 𝑆
rate form
⇒ 𝐸
𝑖𝑛 − 𝐸
𝑜𝑢𝑡 =
𝑑𝐸𝑠𝑦𝑠
𝑑𝑡
𝑜𝑢𝑡 + 𝑆
𝑔𝑒𝑛 =
.
𝑑𝑆𝑠𝑦𝑠
𝑑𝑡
(1)
By ignoring kinetic and potential energies (and other macroscopic forms of energy), the two
thermodynamics laws are [66]
Energy balance: 𝑄
𝑖𝑛 − 𝑄
𝑜𝑢𝑡 + 𝑊
𝑖𝑛 − 𝑊
𝑜𝑢𝑡 + ∑ 𝑚 ℎ
− ∑ 𝑚 ℎ
=
𝑄
𝑖𝑛
𝑇𝐻
𝑄
𝑜𝑢𝑡
𝑇𝐿
− ∑
𝐿
𝐻
𝑖𝑛
𝑜𝑢𝑡
+ ∑ 𝑚 𝑠
− ∑ 𝑚 𝑠
+ 𝑆
𝑔𝑒𝑛 =
𝑖𝑛
𝑜𝑢𝑡
𝑑𝐸𝑠𝑦𝑠
𝑑𝑡
.
𝑑𝑆𝑠𝑦𝑠
𝑑𝑡
(2)
Entropy balance: ∑
{
Here, we adopted subscripts 𝐻 and 𝐿 to indicate high and low temperatures, respectively
(Figure 6).
By applying equation (2) to a general manufacturing process (Figure 6) operating under steady
conditions (
𝑑
𝑑𝑡
= 0), we have
{
Energy balance: 𝑄
𝑖𝑛 − 𝑄
𝑄
𝑖𝑛
𝑇𝐻
Entropy balance:
−
𝑜𝑢𝑡 + 𝑊
𝑄
𝑜𝑢𝑡
𝑇𝐿
𝑖𝑛 − 𝑊
𝑜𝑢𝑡 + (𝑚 ℎ)𝑖𝑛 − (𝑚 ℎ)𝑜𝑢𝑡 = 0
+ (𝑚 𝑠)𝑖𝑛 − (𝑚 𝑠)𝑜𝑢𝑡 + 𝑆
𝑔𝑒𝑛 = 0
.
(3)
By multiplying both sides of the entropy balance equation by 𝑇𝐿 and equating the left-hand
sides of the energy balance and the resulting entropy balance equation, we arrive at
𝑊
𝑖𝑛 = 𝑊
𝑜𝑢𝑡 + (
𝑇𝐿
𝑇𝐻
− 1) 𝑄
𝑖𝑛 + (𝑚 ℎ)𝑜𝑢𝑡 − (𝑚 ℎ)𝑖𝑛 − 𝑇𝐿[(𝑚 𝑠)𝑜𝑢𝑡 − (𝑚 𝑠)𝑖𝑛] + 𝑇𝐿𝑆
𝑔𝑒𝑛.
(4)
Equation (4) is a more general form of the required input work than that obtained in the
literature [63,64], which further considers the output work. Up to here, we followed the same
approach devised by Gutowski et al. [61 -- 64] and further extended their required-input-work
equation.
Now, the required input power (energy rate) would be
𝐸
𝑖𝑛 = 𝑊
𝑖𝑛 + 𝑄
𝑖𝑛 ⇒
𝐸
𝑖𝑛 = 𝑊
𝑜𝑢𝑡 +
𝑇𝐿
𝑇𝐻
𝑄
𝑖𝑛 + (𝑚 ℎ)𝑜𝑢𝑡 − (𝑚 ℎ)𝑖𝑛 − 𝑇𝐿[(𝑚 𝑠)𝑜𝑢𝑡 − (𝑚 𝑠)𝑖𝑛] + 𝑇𝐿𝑆
𝑔𝑒𝑛.
This equation can also be written in non-rate form as
𝐸𝑖𝑛 = 𝑊𝑜𝑢𝑡 +
𝑇𝐿
𝑇𝐻
𝑄𝑖𝑛 + (𝑚ℎ)𝑜𝑢𝑡 − (𝑚ℎ)𝑖𝑛 − 𝑇𝐿[(𝑚𝑠)𝑜𝑢𝑡 − (𝑚𝑠)𝑖𝑛] + 𝑇𝐿𝑆𝑔𝑒𝑛.
(5)
(6)
Equation (6) is a general equation that gives the required input energy for any manufacturing
process.
9
Now, for 4D printing as a manufacturing process that en ables self-assembly at the
manufacturing scale, this equation can be further analyzed and simplified. Let us consider the
following key-points:
• Key-point 1. Self-assembly is a spontaneous and reversible process [67 -- 69,51 -- 53]. For such a
process, the central thermodynamic equation is ∆𝐺𝑝𝑟𝑜𝑐𝑒𝑠𝑠 = 0 [67,65].
• Key-point 2. In a reversible process, the system is in thermodynamic equilibrium with its
surroundings. One of the necessities of thermodynamic equilibrium is thermal equilibrium.
When two bodies are in thermal equilibrium, their temperatures are the same. Therefore, during
a reversible process, 𝑇𝑠𝑦𝑠 ≈ 𝑇𝑠𝑢𝑟𝑟 [70,71,65].
• Key-point 3. For a reversible process, 𝑆𝑔𝑒𝑛 = 0 [65].
By considering Key-point 2, for any manufacturing process in the reversible condition, equation
(6) can be written as (we emphasize that from equation (1), the convention is ∆⦾ = ⦾𝑖𝑛 − ⦾𝑜𝑢𝑡):
𝐸𝑖𝑛 = 𝑊𝑜𝑢𝑡 +
𝑇𝐿
𝑇𝐻
𝑄𝑖𝑛 − ∆𝐺 + 𝑇𝐿𝑆𝑔𝑒𝑛,
(7)
where ∆𝐺 = ∆𝐻 − 𝑇∆𝑆 and 𝐺 is the Gibbs free energy. Then, by using key-points 1 and 3, we
have the following equation for 4D printing that enables self-assembly at manufacturing scale:
𝐸𝑖𝑛 = 𝑊𝑜𝑢𝑡 +
𝑇𝐿
𝑇𝐻
𝑄𝑖𝑛.
(8)
Equation (8) is the minimum theoretical limit of required input energy for 4D printing as a new
process that enables self-assembly at manufacturing scale. In addition, generally, ∆𝐺 ≤ 0 and 𝑆𝑔𝑒𝑛 ≥
0 and thus both terms −∆𝐺 and 𝑇𝐿𝑆𝑔𝑒𝑛 are positive or zero ( ≥ 0 ). Therefore, by comparing
equations (7) and (8), it can be concluded that 4D printing can have the minimum energy
consumption among various manufacturing processes.
10
Figure 6. A manufacturing process in the most general thermodynamic model (this figure has been
drawn based on the concepts in [60 -- 64]). The energy and entropy flows have also been illustrated.
6. Conclusion
We derived the theoretical limit of minimum energy consumption in 4D printing as a
manufacturing process and proved that 4D printing could be the most energy-efficient process
among various manufacturing processes. This minimum energy consumption limit in manufacturing
obtained here can be approached by 4D printing process and future 4D printers. It does not necessarily
mean that this limit is practically achieved in 4D printing processes. One of the main reasons is that,
currently, in 4D printing, although the self-assembly process can be triggered by environmental free
energy, fabrication of the initial self-assembling components requires electrical energy for running
the printers. Nevertheless, future 4D printers may somehow incorporate environmental free energy
for the whole manufacturing process (that is, fabrication of the self-assembling components and then
self-assembly of them). Sadi Carnot worked on the energy efficiency of "heat engines" and the Carnot
cycle gives the theoretical limit in heat engines. Here, we worked on the energy efficiency of
"manufacturing processes" and obtained the theoretical limit of minimum energy consumption in
manufacturing that can be approached by 4D printing. In this study, w e have also clearly
underpinned 4D printing as a new manufacturing process with unique attributes.
Acknowledgments: F.M. wants to thank the graduate-level course "Assembly Modeling for Design and
Manufacturing" taught by Prof. Kazuhiro S aitou at the University of Michigan-Ann Arbor, for which F.M.
served as GS I (Graduate S tudent Instructor). This course was helpful to analyze the concepts of "assembly" at
manufacturing scale.
11
Author Contributions: Conceptualization, Farhang Momeni; Investigation, Farhang Momeni; Methodology,
Farhang Momeni; Visualization, Farhang Momeni; Writing -- original draft, Farhang Momeni; Writing -- review
& editing, Farhang Momeni and Jun Ni.
References
Bunse, K., Vodicka, M., S chönsleben, P., Brülhart, M., & Ernst, F. O. (2011). Integrating energy efficiency
performance in production management -- gap analysis between industrial needs and scientific literature .
2007a. Tracking
Journal of Cleaner Production, 19(6-7), 667-679.
IEA,
https://www.iea.org/publications/freepublications/publication/tracking_emissions.pdf
2018).
Industrial,
Energy Efficiency and CO2 Emissions. Available
at:
(Accessed May
1.
2.
3.
ElMaraghy, H. A., Youssef, A. M., Marzouk, A. M., & ElMaraghy, W. H. (2017). Energy use analysis and
local benchmarking of manufacturing lines. Journal of Cleaner Production, 163, 36 -48.
4. D. S aygin, M.K. Patel, D.J. Gielen, (2010). Global industrial energy efficiency benchmarking: an energy
policy tool R. Wright (Ed.), UNIDO Working Paper, Vienna, Austria.
5.
6.
Rahimifard, S ., S eow, Y., & Childs, T. (2010). Minimising Embodied Product Energy to support energy
efficient manufacturing. CIRP annals, 59(1), 25-28.
S eow, Y., & Rahimifard, S . (2011). A framework for modelling energy consumption within manufacturi ng
systems. CIRP Journal of Manufacturing S cience and Technology, 4 (3), 258-264.
7. Nilakantan, J. M., Huang, G. Q., & Ponnambalam, S . G. (2015). An investigation on minimizing cycle time
and total energy consumption in robotic assembly line systems. Journal of Cleaner Production, 90, 311 -325.
Liu, Y., Dong, H., Lohse, N., Petrovic, S., & Gindy, N. (2014). An investigation into minimising total energy
consumption and total weighted tardiness in job shops. Journal of Cleaner Production, 65, 87-96.
8.
Tibbits, S . The emergence of "4D printing". TED Conference (2013).
9.
10. Tibbits, S . (2014). 4D printing: multi‐material shape change. Architectural Design, 84(1), 116 -121.
11. Tibbits, S ., McKnelly, C., Olguin, C., Dikovsky, D., & Hirsch, S . (2014). 4D Printing and universal
transformation.
12. Gladman, A. S ., Matsumoto, E. A., Nuzzo, R. G., Mahadevan, L., & Lewis, J. A. (2016). Biomimetic 4D
printing. Nature materials, 15(4), 413-418.
13. Ge, Q., Qi, H. J., & Dunn, M. L. (2013). Active materials by four-dimension printing. Applied Physics
Letters, 103(13), 131901.
14. Momeni, F., Liu, X., & Ni, J. (2017). A review of 4D printing. Materials & Design, 122, 42-79.
15. Chen, T., & S hea, K. (2018). An Autonomous Programmable Actuator and S hape Reconfigurable S tructures
using Bistability and S hape Memory Polymers. 3D Printing and Additive Manufacturing.
16. Athanasopoulos, N., & S iakavellas, N. (2018, July). Variable emissivity through multilayer patterne d
surfaces for passive thermal control: preliminary thermal design of a nano -satellite. 48th Internatio nal
Conference on Environmental S ystems. Albuquerque, New Mexico
17. Athanasopoulos, N., & S iakavellas, N. J. (2017). S mart patterned surfaces with programmable thermal
emissivity and their design through combinatorial strategies. Scientific reports, 7(1), 12908.
18. Momeni, F., & Ni, J. (2018). Nature -inspire d smart solar concentrators by 4D printing. Renewable
Energy, 122, 35-44.
19. Momeni, F., S abzpoushan, S ., Valizadeh, R., Morad, M. R., Liu, X., & Ni, J. (2019). Plant leaf-mimetic smart
wind turbine blades by 4D printing. Renewable Energy, 130, 329-351.
20. Ashammakhi, N., Ahadian, S ., Zengjie, F., S uthiwanich, K., Lorestani, F., Orive, G., ... & Khademhosseini,
A. (2018). Advances and future perspectives in 4D bioprinting. Biotechnology journal, 1800148.
21. Zhang, W., Zhang, F., Lan, X., Leng, J., Wu, A. S ., Bryson, T. M., ... & Chou, T. W. (20 18). Shape memory
textile functional composites. Composites Science and
behavior and recovery force of 4D printed
Technology, 160, 224-230.
22. Leist, S . K., Gao, D., Chiou, R., & Zhou, J. (2017). Investigating the shape memory properties of 4D printed
polylactic acid (PLA) and the concept of 4D printing onto nylon fabrics for the creation of smart
textiles. Virtual and Physical Prototyping, 12(4), 290-300.
23. Zolfagharian, A., Kouzani, A. Z., Khoo, S . Y., Moghadam, A. A. A., Gibson, I., & Kaynak, A. (2016).
Evolution of 3D printed soft actuators. Sensors and Actuators A: Physical, 250, 258-272.
12
24. de Marco, C., Pané, S ., & Nelson, B. J. (2018). 4D printing and robotics. Science Robotics, 3(18), eaau0449.
25. Deng, D., Yang, Y., Chen, Y., Lan, X., & Tice, J. (2017) . Accurately controlled sequential self-folding
structures by polystyrene film. Smart Materials and Structures, 26(8), 085040.
26. Bodaghi, M., Damanpack, A. R., & Liao, W. H. (2017). Adaptive metamaterials by functionally graded 4D
printing. Materials & Design, 135, 26-36.
27. Zhang, Q., Zhang, K., & Hu, G. (2016). S mart three -dimensional lightweight structure triggered from a thin
composite sheet via 3D printing technique. Scientific reports, 6, 22431.
28. Xu, H., & Pasini, D. (2016). Structurally efficient three -dimensional metamaterials with controllable thermal
expansion. Scientific reports, 6, 34924.
29. Liu, Z., Du, H., Li, J., Lu, L., Li, Z. Y., & Fang, N. X. (2018). Nano -kirigami with giant optical chirality. Science
advances, 4(7), eaat4436.
30. Ge, Q., Dunn, C. K., Qi, H. J., & Dunn, M. L. (2014). Active origami by 4D printing. Smart Materials and
Structures, 23(9), 094007.
31. Armelin, E., Pérez-Madrigal, M. M., Alemán, C., & Díaz, D. D. (2016). Current status and challenges of
biohydrogels for applications as supercapac itors and secondary batteries. Journal of Materials Chemistry
A, 4(23), 8952-8968.
32. Oliver, K., S eddon, A., & Trask, R. S . (2016). Morphing in nature and beyond: a review of natural and
synthetic shape-changing materials and mechanisms. Journal of Materials S cience, 51(24), 10663-10689.
33. Kruth, J. P., Leu, M. C., & Nakagawa, T. (1998). Progress in additive manufacturing and rapid prototyping.
CIRP Annals-Manufacturing Technology, 47(2), 525-540.
34. Mahajan, R. L., Mueller, R., Williams, C. B., Reed, J., Campbell, T. A., & Ramakrishnan, N. (2012,
November). Cultivating emerging and black swan technologies. In AS ME 2012 International Mechanical
35.
Engineering Congress and Exposition (pp. 549-557). American S ociety of Mechanical Engineers.
Jin, Y., Qin, S . J., & Huang, Q. (2016, August). Prescriptive analytics for understanding of out-of-plane
deformation in additive manufacturing. In Automation S cience and Engineering (CAS E), 2016 IEEE
International Conference on(pp. 786-791). IEEE.
36. Costabile, G., Fera, M., Fruggiero, F., Lambiase, A., & Pham, D. (2017). Cost models of additive
manufacturing: A literature review. International Journal of Industrial Engineering Computations, 8(2),
263-283.
37. Lewis, J. A. (2015). 3D Printing: Making the Future, S cience Research Public Lectures.
38. Groover, M. P., (2008). Automation, production systems, and computer-Integrated manufacturing. Pearson
Education Inc .
39. Boothroyd G, Dewhurst P, Knight WA (2011). Product design for manufacture and assembly. CRC Press, Boca
Raton.
40. Whitney, D. E. (2004). Mechanical assemblies: their design, manufacture, and role in product development.
Oxford University Press Inc, New York, NY.
41. Kota, S ., & Ananthasuresh, G. K. (1995). Designing compliant mechanisms. Mechanical Engineering-
CIME, 117(11), 93-97.
42. Kota, S . (1998). Role of Mechanical Compliance in Product Design: Design for No -Assembly. In International
Forum on DFMA" Foundations & Frontiers.
43. Kota, S . (2001). Compliant systems using monolithic mechanisms. Smart Materials Bulletin, 2001(3), 7-10.
44. Mavroidis, C., DeLaurentis, K. J., Won, J., & Alam, M. (2001). Fabrication of non-assembly mechanisms and
robotic systems using rapid prototyping. Journal of mechanical design, 123(4), 516-524.
45. Mastrangeli, M., Abbasi, S ., Varel, C., Van Hoof, C., Celis, J. P., & Böhringer, K. F. (2009). S elf-assembly
from milli-to nanoscales: methods and applications. Journal of micromechanics and microengineering, 19(8),
083001.
46. Olson, A. J. (2015). S elf-assembly gets physical. Nature nanotechnology, 10(8), 728.
47. Tibbits, S . (2012). Design to S elf‐Assembly. Architectural Design, 82(2), 68-73.
48. Tibbits, S . (2017). From Automated to Autonomous Assembly. Architectural Design, 87(4), 6-15.
49. Tibbits, S . (2012). The S elf-Assembly Line.
50. S ass, L. (2004, S eptember). Design for self assembly of building components using rapid prototyping. In
Architecture in the Network S ociety [22nd eCAADe Conference Proceedings] Copenhagen (Denmark) (pp.
15-18).
13
51. Bensaude-Vincent, B. S elf-Assembly, S elf-Organization: A Philosophical Perspective on Converging
Technologies Paper prepared for France/S tanford Meeting Avignon, December 2006.
52. Whitesides, G. M., & Grzybowski, B. (2002). S elf-assembly at all scales. Science, 295(5564), 2418-2421.
53. Whitesides, G. M., & Boncheva, M. (2002). Beyond mo lecules: S elf-assembly of mesoscopic and
macroscopic components. Proceedings of the National Academy of Sciences, 99(8), 4769-4774.
54. Ball, R. S . (1900). A Treatise on the Theory of Screws. Cambridge university press.
55. Waldron, K. J. (1966). The constraint analysis of mechanisms. Journal of Mechanisms, 1(2), 101-114.
56. Konkar, R., "Incremental Kinematic Analysis and S ymbolic S ynthesis of Mechanisms," Ph.D.
Dissertation, S tanford University, S tanford, CA 1993.
57. De Mello, L. H., & S anderson, A. C. (1991). A correct and complete algorithm for the generation of
mechanical assembly sequences. IEEE transactions on Robotics and Automation, 7(2), 228 -240.
58. Raviv, D., Zhao, W., McKnelly, C., Papadopoulou, A., Kadambi, A., S hi, B., ... & Raskar, R. (2014). Active
printed materials for complex self-evolving deformations. Scientific reports, 4, 7422.
59. Lee, A. Y., An, J., & Chua, C. K. (2017). Two -way 4D printing: A review on the reversibility of 3D-printe d
shape memory materials. Engineering, 3(5), 663-674.
60. Gyftopoulos, E. P., & Beretta, G. P. (2005). Thermodynamics: foundations and applications. Courier
Corporation.
61. Gutowski, T., Dahmus, J., & Thiriez, A. (2006, May). Electrical energy requirements for manufacturi ng
processes. In 13th CIRP international conference on life cycle engineering (Vol. 31, pp. 623-638). CIRP
International Leuven, Belgium.
62. Gutowski, T., Dahmus, J., Thiriez, A., Branham, M., & Jones, A. (2007, May). A thermodynamic
characterization of manufacturing processes. In Electronics & the Environment, Proceedings of the 2007 IEEE
International Symposium on(pp. 137-142). IEEE.
63. Branham, M., Gutowski, T. G., Jones, A., & S ekulic, D. P. (2008, May). A thermodynamic framework for
analyzing and improving manufacturing processes. In Electronics and the Environment, 2008. IS EE 2008.
IEEE International S ymposium on (pp. 1-6). IEEE.
64. Gutowski, T. G., Branham, M. S ., Dahmus, J. B., Jones, A. J., Thiriez, A., & S ekulic, D. P. (2009).
Thermodynamic analysis of resources used in manufacturing processes. Environmental science &
technology, 43(5), 1584-1590.
65. Cengel, Y. A., & Boles, M. A. (2015). Thermodyna mics: An engineering approach. New York: McGraw-Hill
Education.
66. Bejan, A. (2016). Advanced engineering thermodynamics. John Wiley & S ons.
67. Bergström, L. M. (2011). Thermodynamics of self-assembly. In Application of Thermodynamics to Biological
and Materials Science. InTech.
68. S teed, J. W., Turner, D. R., & Wallace, K. (2007). Core concepts in supramolecular chemistry and nanochemistry.
John Wiley & S ons.
69. Frewer, L. J., Norde, W., Fischer, A., & Kampers, F. (Eds.). (2011). Nanotechnology in the agri-food sector:
implications for the future. John Wiley & S ons.
70. S ears, F. W., S alinger, G. (1975). Thermodynamics, kinetic theory, and statistical thermodynamics., 3rd edn.
Addison-Wesley.
71. Zumdahl S S , DeCoste JD (2016) Chemical principles, 8th edn. Cengage Learning.
14
|
1712.03119 | 1 | 1712 | 2017-12-08T15:21:57 | Selective etching of PDMS: etching as a negative tone resist | [
"physics.app-ph"
] | In this work authors present for the first time how to apply the additive-free, cured PDMS as a negative tone resist material, demonstrate the creation of PDMS microstructures and test the solvent resistivity of the created microstructures. The PDMS layers were 45 um and 100 um thick, the irradiations were done with a focused proton microbeam with various fluences. After irradiation, the samples were etched with sulfuric acid that removed the unirradiated PDMS completely but left those structures intact that received high enough fluences. The etching rate of the unirradiated PDMS was also determined. Those structures that received at least 7.5*10^15 ion*cm-2 fluence did not show any signs of degradation even after 19 hours of etching. As a demonstration, 45 um and 100 um tall, high aspect ratio, good quality, undistorted microstructures were created with smooth and vertical sidewalls. The created microstructures were immersed into numerous solvents and some acids to test their compatibility. It was found that the unirradiated PDMS cannot, while the irradiated PDMS microstructures can resist to chloroform, n-hexane, toluene and sulfuric acid. Hydrogen fluoride etches both the unirradiated and the irradiated PDMS. | physics.app-ph | physics | Selective etching of PDMS: etching as a negative tone resist
S.Z. Szilasi1,*, L. Juhasz2
1 Institute for Nuclear Research, Hungarian Academy of Sciences,
H-4001 Debrecen, P.O. Box 51, Hungary
2 University of Debrecen, Dept. of Solid State Physics,
H-4010 Debrecen, P.O. Box 2, Hungary
* Corresponding author. S.Z. Szilasi
Address: Institute of Nuclear Research of the Hungarian Academy of Sciences
H-4026 Debrecen Bem tér 18/c, Mail: H-4001 Debrecen, POB. 51 Hungary
E-mail address: [email protected]
Tel: +36 52 509 200; Fax: +36 52 416 181
Abstract
In this work authors present for the first time how to apply the additive-free, cured
PDMS as a negative tone resist material, demonstrate the creation of PDMS microstructures
and test the solvent resistivity of the created microstructures.
The PDMS layers were 45 m and 100 m thick, the irradiations were done with a
focused proton microbeam with various fluences. After irradiation, the samples were etched
with sulfuric acid that removed the unirradiated PDMS completely but left those structures
intact that received high enough fluences. The etching rate of the unirradiated PDMS was also
1.
determined. Those structures that received at least 7.5×1015 ion × cm-2 fluence did not show
any signs of degradation even after 19 hours of etching.
As a demonstration, 45 m and 100 m tall, high aspect ratio, good quality, undistorted
microstructures were created with smooth and vertical sidewalls.
The created microstructures were immersed into numerous solvents and some acids to
test their compatibility. It was found that the unirradiated PDMS cannot, while the irradiated
PDMS microstructures can resist to chloroform, n-hexane, toluene and sulfuric acid.
Hydrogen fluoride etches both the unirradiated and the irradiated PDMS.
Keywords: PDMS; Resist; Development; Etching; Irradiation; Proton Beam Writing (PBW)
1. Introduction
The rapid development in the field of micro/nanofluidics, micro/nanooptics or micro-
and nanoelectromechanical systems (MEMS/NEMS) demands the continuous development of
lithographic techniques. This includes not only the improvement of the various exposure or
irradiation techniques but also the research and development of new resist materials. By the
introduction of new resist materials, the quality and/or the dimension of the microstructures
may improve and the lithographic processes may become simpler or more reliable.
Poly(dimethylsiloxane) (PDMS) is not unknown in microtechnology. It is widely used
mostly as a mold, a casting or replicating material in soft lithography [1] but recent researches
showed that it is possible to pattern the polymer with some direct writing techniques also. In
2002, Constantoudis et.al. created structures in liquid, uncured PDMS prepolymer with
electron beam lithography and then used the structures as a hard mask. In 2009, Szilasi et.al.
irradiated cured PDMS with a high energy focused proton beam and observed significant
2.
compaction at the irradiated areas [2]. The compaction effect was applied for the creation of
parallel lines with curved surfaces [3] and microlenses [4] in one step, without the need of any
further development. In 2011, Tsuchiya et.al. reported that the uncured, liquid phase PDMS
polymer crosslinks, thus acts as a negative tone resist if it is exposed to proton irradiation and
made microstructures in it [5]. Bowen et.al. created structures by electron beam lithography
and studied the change of Young's modulus as a function of the delivered dose in 2012 [6]. In
2016, Gorissen et.al. patterned PDMS through SU8 mask by reactive ion etching (RIE) [7].
Others made the PDMS pre-polymer photosensitive by various additives [8,9].
The application of PDMS as a resist material in direct writing lithography is based on
the chemical modification of the polymer due to irradiation. The absorbed radiation creates
excited states, ions and free radicals [10] in the polymer that initiate a variety of chemical
reactions. The result may be cross-linking, chain scissioning, or the two simultaneously. In
cured PDMS, chain scissioning prevails that results in the degradation of the polymer
structure. Due to irradiation, the main Si-O-Si chain brakes, functional groups split and the
volatile products (e.g. H2, CH4 and C2H6 gases) leave the irradiated volume [10]. These
processes lead to the transformation of the polymer to SiOx [11] an inorganic, silica-like
product. The material properties of the irradiated and degraded PDMS are significantly
different from the unirradiated polymer. During irradiation, the initially elastic material
becomes hard, rigid and glass-like. Its Young's modulus depends on the irradiation dose and
can be varied over approximately seven orders of magnitude [6]. The refractive index can be
also tuned by changing the irradiation parameters [12].
Thanks
to a
range of advantageous properties,
it
is not surprising
that
poly(dimethylsiloxane) is probably the most widely used silicon-based, cross-linkable
polymer. Besides it is cost effective and easy to use, the cross-linked PDMS is elastic,
optically clear, hydrophobic, chemically resistive, stable and inert. These properties make it a
3.
good choice in various applications such as microfluidic chips [13, 14], microreactors [15],
hydrophobic valves [16], microlenses [17], contact lenses [18], microstamps [19] or even
medical implants [20].
Although the presence of PDMS in numerous research fields and applications is
significant, up to now it has not been known how to use the cured polymer as a negative tone
lithographic resist material. The creation of micro- or nanostructures in cured PDMS has a lot
of advantages compared to the lithography in the liquid pre-polymer. The layer thickness of
the cured layer can be arbitrary while the thickness of the liquid phase is limited by the flow
parameters of the polymer (viscosity, temperature, orientation of the sample). The cured
samples need less attention during sample handling, irradiation and storage also because the
cured layer protects the created structures from outside impacts before development. Since
PDMS is an insulator material, it charges up at the area of irradiation during exposure to
charged beams (electron or ion beams). Due to charging, the liquid PDMS layer flows apart
making the creation of structures in infinitely thick layers impossible. This problem, of
course, does not arise in case of the cured polymer. The above make the creation of arbitrarily
tall structures possible, since the height of the structure is only limited by the penetration
depth of the used radiation. If a thin conductive layer is necessary during irradiation due to
excessive charging, a thin metal layer may simply be evaporated on the top surface of the
cured polymer sample. This does not hinder the adhesion between the substrate and the
polymer layer and can be either removed or kept in the development process.
In this paper authors present for the first time how to apply the additive-free, cured
poly(dimethylsiloxane) as a negative resist material in proton beam writing (PBW), a direct
writing lithography technique.
2. Experimental
4.
The samples were created by using Sylgard 184 kit from Dow-Corning, the mixing ratio
of the base polymer and the curing agent was 10:1. Glass substrates were cut and cleaned in
piranha solution (H2SO4:H2O2 - 3:1) for 5 minutes. The PDMS polymer was spin-coated on
the glass substrates in 45 m and 100 m thicknesses and then baked at 125 ◦C for 30
minutes.
The samples were irradiated at the nuclear microprobe facility of HAS-ATOMKI,
Debrecen, Hungary [21]. The 45 m and 100 m thick samples were patterned by 2 MeV and
2.5 MeV protons, respectively. The size of the beam spot was ~2.5 m × 2.5 m, the beam
current was 1.3 nA. The penetration depth of the different energy protons was calculated by
the SRIM [22] code. These calculations showed that the range of the 2 MeV protons is ~85
m, while that of the 2.5 MeV protons is ~120 m in PDMS. Since the polymer layers were
much thinner than the penetration depths of protons in the corresponding samples, the
particles easily penetrate through the resist layer without suffering considerable lateral
scattering creating structures with vertical sidewalls.
To test the etching method, two kinds of patterns were created in the samples. The so
called fluence test samples consisted of fifteen parallel lines. Each line was numbered and
received different fluences in increasing order. In case of one kind of fluence test samples, the
fluences ranged from 1.25×1015 ion × cm-2 (2 000 nC×mm-2) to 1.88×1016 ion × cm-2 (30 000
nC×mm-2) with approximately 1.25×1015 ion × cm-2 (2 000 nC×mm-2) increments, while other
fluence test samples had better fluence resolution and received fluences between 1.25×1015
ion × cm-2 (2000 nC×mm-2) and 5.63×1015 ion × cm-2 (9000 nC×mm-2) in 3.13×1015 ion ×
cm-2 (500 nC×mm-2) increments.
The demonstration test samples consisted of various shape microstructures, such as
squares, circles, lines with various widths, dot and column matrices. These samples received
1.25×1016 ion × cm-2 (20000 nC×mm-2) fluence.
5.
To develop the samples, concentrated sulfuric acid was used. It was found that dilute
sulfuric acid does not remove the unirradiated PDMS, so 98% concentration was used. For
the best result, the samples were etched for 15 minutes at 35 oC then placed in distilled water
for 2 minutes. In case of high aspect ratio structures, intensive stirring of the etchant or the
water is not advised because the structures may brake off.
The etched structures were investigated by a Zeiss Axio Imager Optical Microscope and
a Hitachi-S4300-CFE scanning electron microscope (SEM).
Since the above mentioned process makes possible the creation of microfluidic
elements, it is important to test which solvents the developed microstructures are compatible
with. In the framework of the solvent compatibility test two samples were placed in every
solvents, a 45 m thick unirradiated PDMS sample and some developed microstructures. The
experiment happened at room temperature, the time duration was 30 minutes. After removing
the samples from the solvents they were dried and examined with an optical microscope.
Besides organic solvents some acids were also tested. A solvent or an acid was considered
compatible with the microstructures or the polymer layer if after 30 minutes no visible
changes (whitening, swelling, any degradation, delamination, etc.) could be observed on
them.
3. Results and discussion
In the first test, a 45 m thick fluence test sample with parallel lines was etched for 5
minutes. It was found that the sulfuric acid removed only the unirradiated PDMS and did not
etch the structures that received high fluences. The lowest fluence line (1.25×1015 ion × cm-2
or 2 000 nC×mm-2) in the test structure disappeared completely but the others remained. The
first good quality line that was not damaged by the etchant in 5 minutes was the one that
6.
received 5×1015 ion × cm-2 fluence (8 000 nC×mm-2). Below this fluence the quality of the
structures decreased with decreasing fluences but above it seemed to be uniform.
The other test sample with smaller fluence steps (3.13×1015 ion × cm-2 or 500 nC×mm-2)
was used to find the fluence threshold of the development with better accuracy. After 5
minutes of etching, it was observed that the quality and integrity of the lines increased
steadily with increasing fluences until it reached 4.38×1015 ion × cm-2 (7000 nC×mm-2) above
which they were uniform (Figure 1).
Figure 1. A fluence test sample with 3.13×1015 ion × cm-2 (500 nC×mm-2) resolution after 5 minutes
etching. The first line with good quality and integrity is #11 which received 4.38×1015 ion × cm-2 (7000
nC×mm-2) fluence. (a) and (b) are 10x and 20x optical microscope images, (c) and (d) are SEM images
taken under 55 degree tilt angle.
To determine how the various fluence microstructures degrade over time in the etchant,
the test sample with wider fluence range was placed back into fresh, 98% sulfuric acid. After
20 minutes of etching, the 5×1015 ion × cm-2 fluence (8 000 nC/mm2) fluence line slightly
started to degrade, which becomes obvious at about 50 minutes. After about 3 hours of
etching, the 6.25×1015 ion × cm-2 fluence (10 000 nC/mm2) structure became thinner by 1 m
and showed some signs of degradation. After 19 hours spent in sulfuric acid, some small
7.
cracks could be seen on the edge of this line, and it became thinner by ~1 m again. Despite
the long etching time, all the other structures that received larger fluences remained intact and
were in a good condition. This shows that the etching has very high selectivity above this
fluence threshold. This concludes that if the creation of microstructures needed with
considerable resistance to strong acids, a fluence above 7.5×1015 ion × cm-2 (12 000 nC/mm2)
is needed to be delivered to the structures.
The demonstration test samples could be developed successfully and in a good quality.
The etchant cleaned the microstructures well, the cured but unirradiated PDMS was removed
completely. The walls were vertical and smooth, the shape of the microstructures were not
deformed.
2. Figure The first 45 m tall microstructures created by particle irradiation in cured PDMS. Figure (a) -
overview image, (b) - various diameter discs, (c) - various width lines, the 2 m wide one broke off (d) - the
edges of a square shape structure – the vertical edge is rounded due to the dose distribution
Due to the irradiation, the elastic PDMS becomes hard and glass like. Some of the high
aspect ratio microstructures broke off due to their rigidity (Figure 2/c), but their adhesion to
the glass substrates were very good. The larger structures could not be easily removed from
the glass substrate mechanically even by touching them with hand or scratching them with a
needle. The explanation for this is probably that the composition and the structure of the
8.
highly degraded PDMS is very similar to the glass', so the radicals that formed due to
irradiation at the interface attached the two medium together strongly.
When energetic particles penetrate inside a material they suffer scattering. The
scattering is more pronounced towards the end of their path where the energy of the ion has
already decreased significantly. These irradiations were designed the way that the protons
penetrate through the polymer layer and stop inside the glass substrate. This way it can be
achieved that the PDMS is modified all the way to the substrate and chemically bonds to it.
Since the 100 m layer thickness is relatively large compared to the 120 m penetration depth
of 2.5 MeV protons in PDMS, the scattering causes visible widening at the bottom of the
microstructures (Figure 3.b). The exact height of these microstructures was measured by SEM
and it turned out to be 103 m (Figure 3). The diameter of the narrowest columns was 7 m at
the tip and 15 m at the bottom. Smaller diameter columns were also irradiated but they broke
off from the substrate during etching. This can probably be avoided by further improvement
of the development method.
9.
Figure 3. 103 um tall structures: (a) overview image, (b) the narrowest columns, their diameter was 7 m
at the tip and 15 m at the bottom, (c) and (d) closeups of the sidewall of a column
At the development of tall and narrow microstructures, the evaporation of the
developing or rinsing liquids may cause problems. If the liquid wets the surface of the
microstructures and/or the substrate, and the structures are in contact with the surface of the
evaporating liquid then the surface tension and the capillary forces may deform the
microstructures or make them collapse. To minimize this effect, the sample has to be kept wet
during the development process and the final rinsing liquid has to be non-wetting to the
degraded PDMS and the substrate. Since the microstructures become rigid due to irradiation
and do not deform easily, they are less sensitive to deformation or collapsing that might occur
during drying of the liquids used in the development process.
The average etching rate of the cured, unirradiated PDMS that is immersed into 35 oC
sulfuric acid has also been determined during the experiments and it turned out to be about
0.35 m/sec. This means that a 45 m high structure develops in about 2 minutes. However,
few microns thin layers etch much faster (~1.5 m/sec) because the reaction products of
10.
PDMS with the sulfuric acid cannot accumulate close to the sample surface and hamper the
fresh etchant to reach the microstructures.
The reaction products of PDMS with concentrated sulfuric acid were studied by Lee et
al. by IR and mass spectrometry [23]. It turned out that the white precipitate that forms in the
reaction
consists of
low molecular weight oligomers having
the
structure
(CH3)3Si[OSi(CH3)2]xOSi(CH3)3.
The results of the solvent resistivity test (Table 1) showed that the unirradiated PDMS
do not, while the irradiated PDMS microstructures do resist to chloroform, n-hexane, toluene
and of course sulfuric acid (98%). This is the consequence of changing the material structure
of the polymer due to high fluence irradiation. Hydrogen fluoride (38%) etches both the
unirradiated and the irradiated PDMS. Previous studies [23] reported that some tested solvents
swell PDMS in a significantly longer time and/or at elevated temperatures. At room
temperature after the duration of the test, we did not observe the above mentioned effects.
Besides the substances listed in Table 1, the effects of 30% potassium hydroxide (KOH) and
30% sodium hydroxide (NaOH) solutions were also tested on irradiated samples. It was found
that both solutions etched effectively those areas of the sample that were irradiated with
sufficiently high fluences. This means that KOH and NaOH can be used to etch PDMS as a
positive resist material. Further results and the details of this study will be published in a
separate paper. During the development experiments, it was found that the 30 wt% KOH + 20
wt% IPA + 50 wt% DI water solution at 70 oC temperature etched away both the irradiated
and non-irradiated PDMS in 20 minutes. This solution can be used to clean any PDMS
residues off of glass wafers.
According to other studies [23], trifluoroacetic acid, dipropylamine and Tetra-n-
butylammonium fluoride (TBAF) + tetrahydrofuran (THF) solution also dissolve the cured
11.
PDMS polymer. These substances may also be good candidates to selectively etch the cured
and micropatterned PDMS.
Solvent
Acetic anhydride
Acetone
Acetonitrile
Benzene
Chloroform
Cyclohexane
Dibutyl ether
Diethyl ether
Diethylene glycol monobutyl ether
Ethanolamine
Ethyl alcohol
Hydrochloric acid (37%)
Hydrogen fluoride (38%)
Hydrogen peroxide (30%)
Isopropanol
Methanol
Morpholine
n-Butyl alcohol
n-Hexane
Nitric acid (68%)
Petroleum ether
Sulfuric acid (98%)
tert-Butyl alcohol
Tetrahydrofuran
Toluene
Water
Xylene
Unirradiated PDMS
Irradiated PDMS
microstructures
Compatible
Not
compatible
Compatible
Not
compatible
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Swells
Etches
Swells
Etches
Swells
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Etches
1. Table The results of the 30 minute solvent and acid compatibility test
PDMS is capable of the creation of cured polymer layers with arbitrary thicknesses
starting from the nanometre regime. Due to the hardness, chemical resistivity, cost-
effectiveness and good adhesion of the micro-/nanostructures, this polymer may be a great
12.
choice for hard masks that need to resist the erosion of wet or dry etching in various
lithographic processes.
4. Conclusions
Authors have found how the additive-free, cured and proton irradiated PDMS can be
selectively etched as a negative tone resist material. In this experiment, 45 m and 100 m
thick cured PDMS layers were irradiated with a focused proton microbeam with various
fluences and then etched with 98% sulfuric acid. The etchant removed all the unirradiated
PDMS very well while left the irradiated structures intact if the irradiation fluence was high
enough. The etching rate of the unirradiated PDMS was about 0.35 m/sec. In case the
irradiation fluence exceeded 7.5×1015 ion × cm-2 (12 000 nC/mm2), no signs of degradation
was observable on the structures even after 19 hours of etching in concentrated sulfuric acid.
This indicates how high the selectivity of this etching method is. If the creation of
microstructures with considerable resistance to strong acids or organic solvents is necessary,
at least the above fluence needs to be delivered to the structures.
With this technique, good quality, smooth and vertical sidewall, undistorted, high aspect
ratio microstructures were created in 45 m and 100 m thick PDMS layers. The
microstructures become glass-like, rigid and adhered to the glass substrate very well.
The solvent resistivity of the created microstructures was also tested. It was found that
the unirradiated PDMS cannot while the irradiated PDMS microstructures can resist to
chloroform, n-hexane, toluene and of course sulfuric acid (98%). Hydrogen fluoride (38%)
etches both the unirradiated and the irradiated PDMS.
It was also an important finding that KOH and NaOH solutions could be used to
selectively etch PDMS as a positive resist material. These results will be presented in a
separate paper.
13.
5. Acknowledgements
This work was supported by the National Research, Development and Innovation Fund
No. PD 121076, by the Hungarian Scientific Research Fund OTKA No. K 108366 and by the
TAMOP 4.2.2.A-11/1/ KONV-2012-0036 project, which is co-financed by the European
Union and European Social Fund.
The work was also supported by the GINOP 2.3.2-15-2016-00041 (co-financed by the
European Union and the European Regional Development Fund).
References:
[1] W.
Ieong, H. Chih-Ming, Surface molecular property modifications
for
poly(dimethylsiloxane) (PDMS) based microfluidic devices, Microfluidics and
Nanofluidics 7 (2009) 291-306
[2] S.Z. Szilasi, R. Huszank, A. Csik, C. Cserháti, I. Rajta, PDMS patterning by proton beam,
Nucl. Instrum. and Meth. in Phys. Res. B 267 (2009) 2296–2298
[3] S.Z. Szilasi, J. Kokavecz, R. Huszank, I. Rajta, Compaction of poly(dimethylsiloxane)
(PDMS) due to proton beam irradiation, Appl. Surf. Sci. 257 (2011) 4612–4615
[4] S.Z. Szilasi, N. Hegman, A. Csik, I. Rajta, Creation of convex microlenses in PDMS with
focused MeV ion beam, Microelectronic Eng. 88 (2011) 2885-2888
[5] R. Tsuchiya, H. Nishikawa, Fabrication of silica-based three-dimensional structures by
changing fluence using proton beam writing, Transactions of the Mater. Res. Society of
Japan 36.3 (2011) 325-328
[6] J. Bowen, D. Cheneler, A.P.G. Robinson, Direct e-beam lithography of PDMS,
Microelectronic Eng. 97 (2012) 34–37
14.
[7] B. Gorissen, C. Van Hoof, D. Reynaerts, M. De Volder, SU8 etch mask for patterning
PDMS and its application to flexible fluidic microactuators, Microsyst. & Nanoeng. 2
(2016) 16045
[8] K. Tsougeni, A. Tserepi, E. Gogolides, Photosensitive poly(dimethylsiloxane) materials
for microfluidic applications, Microelectronic Engineering 84 (2007) 1104–1108
[9] A. Romeo, S.P. Lacour, Concurrent photopatterning of elastic modulus and
structures in photosensitive silicone elastomers, Extreme Mechanics Letters 3 (2015) 1–7
[10] A. Kondyurin, M. Bilek, Interactions of energetic ions with polymers: chemical picture,
in: Ion Beam Treatment of Polymers (Second Edition), 2015, pp 29-67
[11] M. Ouyang, C. Yuan, R.J. Muisener, A. Boulares, J.T. Koberstein, Conversion of some
siloxane polymers to silicon oxide by UV/ozone photochemical processes, Chem. Mat.
12 (6) (2000) 1591-1596
[12] S.Z. Szilasi, J. Budai, Z. Pápa, R. Huszank, Z. Tóth, I. Rajta, Refractive index depth
profile and its relaxation in polydimethylsiloxane (PDMS) due to proton irradiation,
Mater. Chem. and Phys. 131 (2011) 370–374
[13] X. Zhang, X. Wu, R. Peng, D. Li, Electromagnetically controlled microfluidic chip for
DNA extraction, Measurement 75 (2015) 23–28
[14] J. Chen, D. Chen, T. Yuan, X. Chen, Y. Xie, H. Fu, D. Cui, X. Fan, M K.K. Oo, Blood
plasma separation microfluidic chip with gradual filtration, Microelectronic Eng.128
(2014) 36–41
[15] R. Chen, L. Li, X. Zhu, H. Wang, Q. Liao, M.X. Zhang, Highly-durable optofluidic
microreactor for photocatalytic water splitting, Energy 83 (2015) 797-804
[16] T. Fujii, PDMS-based microfluidic devices for biomedical applications, Microelectronic
Eng. 61–62 (2002) 907–914
15.
[17] K. Zhong, Y. Gao, F. Li, Z. Zhang, N. Luo, Fabrication of PDMS microlens array by
digital maskless grayscale lithography and replica molding technique, Optik -
International J. for Light and Electron Optics, 125 (2014) 2413-2416
[18] C.H. Lin, Y.H. Yeh, W.C. Lin, M.C. Yang, Novel silicone hydrogel based on PDMS
and PEGMA for contact lens application, Colloids and Surfaces B: Biointerfaces, 123
(2014) 986-994
[19] P.G. Shao, J.A. van Kan, K. Ansari, A.A. Bettiol, F. Watt, Poly (dimethyl siloxane)
micro/nanostructure replication using proton beam written masters, Nucl. Instrum. and
Meth. in Phys. Res. B 260 (2007) 479–482
[20] M. Chua, C.K. Chui, Probabilistic predictive modelling of carbon nanocomposites for
medical implants design, J. of the Mechanical Behavior of Biomed. Mater. 44 (2015)
164–172
[21] I. Rajta, I. Borbély-Kiss, Gy. Mórik, L. Bartha, E. Koltay, Á.Z. Kiss, The new ATOMKI
scanning proton microprobe, Nucl. Instr. and Meth. B109 (1996) 148
[22] J.F. Ziegler, SRIM-2008.03. <http://www.srim.org>
[23] J.N. Lee, C. Park, G.M. Whitesides, Solvent Compatibility of Poly(dimethylsiloxane)-
Based Microfluidic Devices, Analytical Chemistry 75 (2003) 6544-6554
16.
|
1907.08593 | 1 | 1907 | 2019-07-11T18:30:24 | Microwave-to-optical conversion using lithium niobate thin-film acoustic resonators | [
"physics.app-ph",
"physics.optics",
"quant-ph"
] | We demonstrate conversion of up to 4.5 GHz-frequency microwaves to 1500 nm-wavelength light using optomechanical interactions on suspended thin-film lithium niobate. Our method utilizes an interdigital transducer that drives a free-standing 100 $\mu$m-long thin-film acoustic resonator to modulate light travelling in a Mach-Zehnder interferometer or racetrack cavity. Owing to the strong microwave-to-acoustic coupling offered by the transducer in conjunction with the strong photoelastic, piezoelectric, and electro-optic effects of lithium niobate, we achieve a half-wave voltage of $V_\pi$ = 4.6 V and $V_\pi$ = 0.77 V for the Mach-Zehnder interferometer and racetrack resonator, respectively. The acousto-optic racetrack cavity exhibits an optomechancial single-photon coupling strength of 1.1 kHz. Our integrated nanophotonic platform coherently leverages the compelling properties of lithium niobate to achieve microwave-to-optical transduction. To highlight the versatility of our system, we also demonstrate a lossless microwave photonic link, which refers to a 0 dB microwave power transmission over an optical channel. | physics.app-ph | physics |
Microwave-to-optical conversion using lithium
niobate thin-film acoustic resonators
LINBO SHAO,1,6 MENGJIE YU,1 SMARAK MAITY,1 NEIL
SINCLAIR,1, 2 LU ZHENG,3 CLEAVEN CHIA,1 AMIRHASSAN
SHAMS-ANSARI,1 CHENG WANG,1,4 MIAN ZHANG,1,5 KEJI LAI,3
AND MARKO LON CAR1,7
1John A. Paulson School of Engineering and Applied Sciences, Harvard University, 29 Oxford Street,
Cambridge, MA 02138, USA
2Division of Physics, Mathematics and Astronomy, and Alliance for Quantum Technologies (AQT),
California Institute of Technology, 1200 E. California Blvd., Pasadena, California 91125, USA
3Department of Physics, University of Texas at Austin, Austin, Texas 78712, USA
4Department of Electrical Engineering & State Key Lab of THz and Millimeter Waves, City University of
Hong Kong, Kowloon, Hong Kong, China
5HyperLight Corporation, 501 Massachusetts Avenue, Cambridge, MA, 02139, USA
[email protected]
[email protected]
Abstract: We demonstrate conversion of up to 4.5 GHz-frequency microwaves to 1500
nm-wavelength light using optomechanical interactions on suspended thin-film lithium niobate.
Our method utilizes an interdigital transducer that drives a free-standing 100 µm-long thin-film
acoustic resonator to modulate light travelling in a Mach-Zehnder interferometer or racetrack
cavity. Owing to the strong microwave-to-acoustic coupling offered by the transducer in
conjunction with the strong photoelastic, piezoelectric, and electro-optic effects of lithium
niobate, we achieve a half-wave voltage of Vπ = 4.6 V and Vπ = 0.77 V for the Mach-Zehnder
interferometer and racetrack resonator, respectively. The acousto-optic racetrack cavity exhibits
an optomechancial single-photon coupling strength of 1.1 kHz. Our integrated nanophotonic
platform coherently leverages the compelling properties of lithium niobate to achieve microwave-
to-optical transduction. To highlight the versatility of our system, we also demonstrate a lossless
microwave photonic link, which refers to a 0 dB microwave power transmission over an optical
channel.
© 2019 Optical Society of America
Introduction
1.
Conversion of information between the microwave and optical domains is a key ingredient for
classical and quantum hybrid signal processing, computing, and networking [1 -- 5]. Among the
many approaches to achieve coherent quantum transduction, electrically-coupled optomechanical
systems have emerged as a promising candidate [6]. Experimental progress includes suspended
structures such as optical waveguides in microwave cavities [7], membranes in free-space Fabry-
Pérot cavities [8 -- 10], and nanoscale piezoelectric optomechanical crystals (OMCs) [11 -- 16].
While optical waveguides in bulk microwave cavities benefit from the high quality (Q) factors
of microwave resonance, the suspended membranes achieve a high photon number conversion
efficiency leveraging a triple resonance of microwave, mechanical, and optical fields. Large-
scale integration of these devices is, however, challenging and has not been demonstrated
yet. On the other hand, OMCs [17 -- 22] provide a fully-integrated platform featuring gigahertz
mechanical frequencies, and megahertz optomechanical coupling strengths, while limitations
due to surface effects are becoming more well-understood [17]. Microwave-to-mechanical (i.e.
electromechanical) coupling to OMCs has been achieved using piezoelectric materials, such as
aluminum nitride [14 -- 16], gallium arsenide [11,13], and lithium niobate (LN) [12]. However,
the demonstrated electromechanical couplings are inefficient due to the mismatch between the
mechanical resonant modes to microwaves [12,15,16] or travelling mechanical waves [11,13,14].
To address the weak microwave-to-mechanical conversion of current integrated devices,
we use free-standing LN thin-film acoustic (i.e. mechanical) resonators with low-loss optical
resonators. Using an interdigital transducer (IDT), our acousto-optic devices demonstrate up to
50% coupling efficiency from microwave inputs to acoustic resonator modes, thereby enabling
efficient optical modulation using Mach-Zehnder interferometers (MZI) and racetrack cavities.
Our approach benefits from the strong piezeoelectricity and electro-optic effects of LN [23 -- 25] in
conjunction with the photoelastic effect to achieve microwave-to-optical conversion. Specifically,
using a 100 µm long optical waveguide embeded within a 3.33 GHz acoustic resonator with
a Q factor of 3,600, our MZI exhibits a low half-wave voltage Vπ of 4.6 V. Moreover, the
half-wave-voltage-length product Vπ L, the figure of merit for optical modulators, is as low as
0.046 V·cm, which is a 50-fold reduction over the start-of-the-art electro-optic modulators [26].
This comes at expense of reduced microwave-to-optical conversion bandwidth of around 1 MHz,
which is significantly smaller than that of electro-optic MZI approaches [26], nonetheless it is
much greater than what has previously been demonstrated using using a microwave, mechanical,
and optical triple resonance [8]. Our racetrack cavity features an optical Q factor of over 2 × 106,
thereby enabling single optical sideband conversion with an effective Vπ of 0.77 V, a photon
number conversion efficiency of 0.0017 % for an optical power of 1 mW, and an acousto-optic (i.e
optomechanical) coupling strength of 1.1 kHz. Though this acousto-optic coupling strength is
lower than that of state-of-the-art OMCs, the overall microwave-to-optical conversion efficiency
is improved due to the enhanced microwave-to-mechanical coupling. Finally, to illustrate its
efficient microwave-to-optical conversion, we demonstrate a loss-less microwave-photonic link
with a ∼50 mW optical power routing on chip.
2. Device design and fabrication
We utilize IDTs to drive our acoustic resonators due to their efficient electromechanical coupling
and ease of fabrication. Notably, IDTs are widely used in electro-acoustic signal processing at up
to hypersonic (greater than 1 GHz) frequencies [27]. Furthermore, they have been used in optical
applications to diffract guided beams [28], modulate cavities [29], drive photonic molecules [30],
and even break time-reversal symmetry [31]. Therefore, integration of IDT-coupled acoustic
resonators [32, 33] with high performance optical devices fabricated in LN [34, 35] offers the
possibility for efficient acoustically-medicated microwave-to-optical conversion.
Figures 1(a) and 1(b) shows our acousto-optic devices in the MZI and the racetrack cavity
configurations, respectively.
IDT-coupled acoustic resonators host optical waveguides and
modulate the phase acquired by the optical signal propagating in one arm of the MZI, which
results in intensity modulation of transmitted optical signal (Fig. 1(a)). Similar effect is
responsible for modulating the optical resonance of the racetrack cavity (Fig. 1(b)). For our
optical components, we employ rib waveguides which are defined by etching a 0.8-µm-thick
X-cut LN thin film to a depth of 0.4 µm. The width of the optical waveguide is 0.95 µm and
1.3 µm in the case of MZI and the racetrack cavity, respectively. In the case of latter, the larger
waveguide cross-section is used to reduce the propagating loss due to sidewall roughness. To
reduce bending loss and to minimize mode conversion between transverse electric (TE) and
transverse magnetic (TM) modes, we employ quadratic Bézier curves for our optical waveguides,
which is a generalized form of more conventional Euler curves [36]. To confine the acoustic
waves, the LN waveguide is released by removing a sacrificial silicon dioxide layer ( Fig. 1(c)).
The suspended thin film acoustic resonator is formed by the long slots on both sides of a LN
waveguide section. The length of each acoustic resonator is 10 µm, and the width is 100 µm.
The electrode width of 90 µm is used to match the impedance of IDT with the 50 Ω impedance
of the driving electronics. Electrode pitch and width of IDTs are controlled to allow excitation
of acoustic modes at different frequencies. Specifically, an IDT with four electrodes, pitch of
0.6 µm, and width of 0.3 µm is used for the acoustic resonator in the MZI configuration. This
allows coupling to acoustic modes in the range of 1 to 4.5 GHz with peak coupling efficiency
occuring around 3 GHz. In the racetrack cavity configuration, we use an IDT featuring pitch
of 0.86 µm, width of 0.43 µm, and same number of electrodes (four). This allows the highest
coupling efficiency to acoustic modes around 2 GHz.
To fabricate each device, three layers of electron beam lithography are used to define the LN
optical waveguides, the opening slots for the acoustic resonator and the release of the LN layer,
and the metal electrodes needed for IDT. The LN is etched using reactive ion etching, and the
fabricated devices feature a sidewall angle of 70◦. The metal electrodes are defined using a
lift-off process: PMMA resist is patterned as a sacrificial layer, then a 75-nm-thick gold layer
with a 8-nm-thick chromium adhesion layer is deposited using electron-beam evaporation, and
the device is immersed in a solvent to lift off the resist. Finally, the release of the LN device from
the substrate is achieved using buffered oxide etchant, which removes the underneath sacrificial
oxide layer through the completely etched slots of the LN layer.
Fig. 1. Integrated acousto-optic devices on suspended thin-film LN. (a) Microscope image
of a suspended acousto-optic MZI. The interferometer is unbalanced to allow phase control
by laser detuning. (b) Microscope image of a suspended optical racetrack cavity with a
thin-film acoustic resonator. The suspended regions adjacent to the optical waveguide are
identified by a different color, which is darker than the plain substrate in (a) and lighter in
(b). (c) False color scanning electron microscope image of the acoustic resonator with an
IDT and an optical waveguide.
3. Description of acoustically-mediated microwave-to-optical conversion
Our acoustic resonator mediates the microwave-to-optical conversion by coupling to the microwave
input via the IDT and modulating light due to acoustic modes (Fig. 2(a)). The optical modulation
is enabled by a generalized acousto-optic interaction that comprises conventional optomechanical
couplings of photoelastic and weak moving boundary effects, as well as cascaded piezoelectric
and electro-optic effects, which feature a coupling strength comparable to photoelastic alone. We
perform simulations to understand and engineer the interplay between these three effects in order
to maximize the overall acousto-optic interaction.
100 μmThin-film acoustic resonatorSuspendedoptical waveguideContact pads for IDTMach -- Zehnder interferometer(a)100 μmSuspended optical racetrack cavityContact pads for IDTThin-film acoustic resonator(b)IDTThin-film acoustic resonatorOptical waveguide(c)20 μmOpening slots Fig. 2. Numerical simulation of the acoustically-mediated microwave-to-optical conversion.
(a) Couplings between the microwave, acoustic, and optical fields that facilitate microwave-
to-optical conversion. (b) Device schematic used for the 2D numerical simulation. The
crystal orientation and coordinate system are shown. The top width of the optical waveguide
is 0.95 µm (c) Electric field Ex of the fundamental TE optical mode. (d) syy component
of the acoustic strain field for the 3.24 GHz acoustic mode, and resulting (e) electric fields
Ex induced by piezoelectric effect. We note that syy has the largest contribution to the
photoelastic interaction shown in (f). (g) Electro-optic interactions between the optical TE
mode and acoustic fields, mediated by piezo-electric effect. In (f) and (g) the interaction
is described by an induced optical refractive index change, calculated by multiplying the
optical electric field, the acoustic field, and the interaction matrices. Color scale bars in (d)
and (e) are normalized individually, while those in (f) and (g) are the same.
The coupling strength of each interaction is evaluated using a 2D numerical model based on
the device cross section and crystal orientation shown in Fig. 2(b). To avoid double-counting
the coupling strength, we use photoelastic (electro-optic) coefficients under a constant electric
field (strain, i.e. clamped) condition following that presented in Ref. [25, 37]. Owing to the
LN crystal orientation chosen for our device, the generalized acousto-optic interaction is much
stronger for the guided transverse electric (TE) mode (Fig. 2(b)) than for the transverse magnetic
(TM) mode at 1550 nm, due to the strong electro-optic and photoelastic coefficients, r33 and p31,
respectively, for TE polarization. The elecro-optic coefficients form a third-order tensor, and r33
relates nZ Z, the optical index change of the crystal Z Z component (indicated by the first 3 in
the subscript), to EZ, the electrical field in Z direction (indicated by the first 3 in the subscript).
The photoelastic coefficients form a forth-order tensor, and p31 relates nZ Z to sX X, the strain
of crystal X X component (indicated by the 1 in the subscript). Thus the strain sX X (or syy in
simulation coordinate) contributes most to the photoelastic interaction. Detailed discussions
are provided in the Supplementary Material. Figures 2(d) and 2(e) plot the simulated acoustic
strain syy and electric field Ex of a 3.24 GHz mode, both of which have the same sign across
the optical waveguide region, thus contributing constructively to the overall modulation to the
optical refractive index.
The contributions of moving boundary, electro-optic and photoelastic effects are calculated
by integrating the products of the acoustic and optical field components with the corresponding
coupling matrices. Acousto-optic interaction strengths between various acoustic modes and
optical TE or TM mode are summarized in Table S1. For the 3.24 GHz acoustic mode, Fig. 2(f)
(b)Lithium niobate thin filmMetal electrodesOpticalwaveguideZXxyCrystalSimulationPiezoelectrically-induced electric field Electro-optic interactionAcoustic strain field syy Photoelastic interaction0-11-0.50.5Amplitude (arb. unit)010.5Optical electric field Ex(c)(d)(f)(e)(g)ℰxOptical fieldStrain fieldElectric fieldPiezoelectric effectPhotoelastic effectElectro-optic effectMoving boundary Electric field(microwave input)IDTPiezoelectric effectThin-film LN device(a) and 2(g) show the induced refractive index changes of the TE mode by the photoelastic and
electro-optic effects. Based on this result, we extract an acousto-optic single-photon coupling
strength g0 = 1.6 kHz for the racetrack cavity geometry. For the acousto-optic MZI, the half-
wave-voltage-length product Vπ L depends on the the acoustic Q factor and microwave-to-acoustic
coupling efficiency, where we use experimental values to avoid over-estimation by the 2D
simulation. A half-wave-voltage-length product Vπ L = 0.045 V·cm for the acousto-optic MZI
is extracted based on the simulated interaction strength, as well as the acoustic Q of 2,000 and
electrical-to-acoustic coupling efficiency of 0.5 (corresponding to a -3 dB dip in the S11 spectrum)
from typical experimental measurements. Details are provided in the Sec. 1 of Supplementary
Material.
4. Acousto-optic Mach-Zehnder interferometer
We experimentally characterize our acousto-optic MZI using a tunable C-band laser, a vector
network analyzer (VNA), and a photoreceiver that features a sensitivity of ∼800 V/W (Fig. 3(a)).
We use lensed fiber to couple light into and out of our structures with a fiber-to-fiber insertion
loss of 10 dB ( < 5 dB/facet) for our suspended LN chip. The periodic variation of optical
transmission with wavelength at 10 nm intervals is consistent with the optical path difference
in the MZI (Fig. 3(b)). To optimize the microwave-to-optical conversion efficiency, the laser
wavelength is chosen to be 1534 nm, corresponding to 50% transmission and indicating a π/2
phase difference between two optical paths.
We evaluate the acoustic resonances and the microwave-to-acoustic coupling of our devices by
measuring the microwave reflection (S11) of the IDT. Our acoustic resonator exhibits multiple
resonances in the range between 1.0 and 4.5 GHz (Fig. 3(d)). To correlate measured acoustic
modes with that of our the simulations, the acoustic electric field profiles are experimentally-
measured using transmission-mode microwave impedance microscopy [38, 39], see Sec. 4 in
Supplementary Material. We measure acoustic Q factors of up to 3,600 (Fig. 3(c)), similar to the
LN OMC devices [12,18], and our resonance at the microwave frequency of 3.273 GHz yields a
state-of-the-art frequency-quality-factor product of f Q > 1013 at room temperature [33].
We characterize the optical modulation induced by the acoustic resonance by the opto-acoustic
S21 spectrum, where the driving port 1 of the VNA is connected to the IDT of the acoustic
resonator and the detecting port 2 is connected to the photoreceiver [Fig. 3(a)]. The S21 spectrum
shown in Fig. 3(d) features several peaks indicating enhanced microwave-to-optical conversion at
acoustic resonances, with the strongest responses measured at the 2.24 and 3.33 GHz acoustic
modes, in agreement with our simulations (Table S1). The microwave-to-optical conversion
efficiency indicated by the S21 depends on both the acoustic Q factor and the overlap between the
acoustic mode and the optical mode. We extract the half-wave voltage Vπ of our acousto-optic MZI
from the experimental measurements of S21 spectrum. Under the conditions of our measurement,
the MZI half-wave voltage Vπ is related to the S21 by
(cid:18) π RPD Irec
(cid:19)2
Vπ
S21 =
,
(1)
in which RPD (Irec = 0.25 mW) is the sensitivity of (optical power at) the photoreceiver, with
derivation given in Sec. 2 of Supplementary Materials. We find Vπ = 4.6 V (5.8 V) using
S21 = −17.4 dB (-19.3 dB) at the resonance frequency of 3.33 GHz (2.24 GHz), and due to the
100 µm length of our acoustic resonator, we obtain Vπ L = 0.046 V·cm (0.058 V·cm), agreeing
with that predicted by our simulation.
5. Acousto-optic racetrack cavity
Compared with MZI, our racetrack cavity features loaded optical Q factor of 2.2 × 106 for
TE-polarized light of wavelength 1574.9 nm, corresponding to a linewidth of 95 MHz (Fig. 4(b)),
Fig. 3. Characterization of the acousto-optic MZI. (a) Simplified experimental schematic
for acousto-optic device characterization. (b) Optical transmission of the acousto-optic
MZI. (c) S11 reflection spectrum of the acoustic resonator. (d) S21 spectrum showing an
enhanced microwave-to-optical conversion at the resonances indicated by the S11 spectrum.
The optical power detected by the photoreceiver is 0.25 mW.
allowing us to operate in the microwave sideband-resolved regime. With the laser blue-detuned
by the acoustic resonant frequency from the optical resonance, we generate an optical sideband by
the acoustic resonant mode and enhance it using the racetrack resonator (Fig. 4(a)). Consequently,
we observe a high S21 = −7.5 dB at the acoustic resonant frequency of 2 GHz with a total
optical power of Irec = 0.13 mW measured at the photoreceiver (Fig. 4(c)). We note that the
S21 quadratically depend on the optical power, and we should consider S21 at the same optical
power level to compare the conversion efficiency. The S21 of the racetrack cavity thus results in a
much lower effective Vπ of 0.77 V than the MZI, as determined by the small signal response of a
intensity modulator (see Sec. 2 of Supplementary Material).
In the
sideband-resolved regime (Ωm (cid:29) κ) and for weak microwave inputs, the relation between the
S21 and g0 is given by
Next, we determine the overall acousto-optic single-photon coupling strength g0.
8g2
0 γe κ2
γ2Ω3
PD I2
e R2
rec
mκ2Rload
S21 =
,
(2)
where κ (γ) and κe (γe) are the total loss and external coupling rate of the optical (acoustic) mode,
respectively. Ωm is the frequency of the acoustic mode, and Rload = 50 Ω is the impedance of
the input microwave source. Eq. 2 is derived from the equation of motion for the dynamics of
the acousto-optic cavity (see Sec. 3 in Supplementary Material), We estimate the acousto-optic
single-photon coupling strength to be g0 ∼ 1.1 kHz between the 2.17 GHz acoustic mode and the
fundamental TE optical mode, which is in good agreement with our theoretical predictions (see
Table S1).
Another important figure of merit is the photon number conversion efficiency η from the
Network AnalyzerActive photoreceiverPort 2Port 1Tunable laserAcousto-optic MZIDriving acoustic resonator(a)14801490150015101520153015400.00.20.40.60.81.0Optical transmissionOptical wavelength (nm)(b)3.263.273.283.290.50.60.70.80.91.0Acoustic S 11Acoustic frequency (GHz)Q = 3,600m(c)1.01.52.02.53.03.54.04.55.0-80-60-40-200Opto-acoustic S (dB)21Acoustic frequency (GHz)-10-8-6-4-20Acoustic S (dB)11(d) Fig. 4. Characterization of the acousto-optic racetrack cavity. (a) Illustration of single-
sideband microwave-to-optical conversion using an acousto-optic cavity. (b) Transmission
spectrum of an high-Q optical resonance. (c) Acoustic S11 and opto-acoustic S21 spectra.
A high resolution measurement around 2 GHz is shown in dark blue. The optical pump
wavelength is set to maximize the power received at the photoreceiver.
microwave frequency to the optical sideband frequency, and it describes the device performance
at the single photon level. From the derivation described in Sec. 3 of the Supplementary Material,
the photon number conversion efficiency η is given by
η = C0 · ncav · 2γe
(3)
γ
· 2κe
κ
,
m
0/(γκ) is the single-photon cooperativity, ncav = κeIopt/(Ω2
ω0) is the intracavity
where C0 = 4g2
optical photon number for the blue-detuned pump light, 2κe/κ (2γe/γ) describes the external
coupling efficiency of optical (acoustic) mode. Based on the experimentally extracted rates
(Table S2), our acousto-optic cavity features an single-photon cooperativity C0 = 4 × 10−8, and
an photon number conversion efficiency η = 0.0017 % for an optical power of Iopt = 1 mW. This
efficiency could be further improved by acoustic and photonic engineering as discussed later in
Sec. 7.
As shown in Eq. 3, the photon number conversion efficiency depends on both optomechanical
cooperativity (C0) and the microwave-to-mechanical coupling strength (described by 2γe/γ).
Recent progress of LN OMCs [12] has demonstrated unitary optomechanical cooperativity, but
their microwave-to-mechanical coupling strength is as low as 10−8, which would limit the overall
photon number conversion efficiency. Benefiting from our up to 50% microwave-to-acoustic
coupling to the thin-film acousto-optic resonator, this photon number conversion efficiency could
be greater than the OMCs, even though the acousto-optic coupling strength is weaker than that of
the OMCs.
6. Demonstration of a microwave-photonic link
A microwave photonic link enables low-loss long haul transport and flexible manipulation of
microwave signals using optical devices by up-converting microwave frequencies to optical
frequencies. To benchmark our acousto-optic racetrack device, we demonstrate a narrow-band
microwave-photonic link using our acousto-optic racetrack cavity, and a link gain of 0 dB is
achieved without the need of an optical amplifier within the link (after the modulation of our
acousto-optic device), which would significantly increases the noise of the link. The optical
pump light is amplified to ∼500 mW by an erbium-doped fiber amplifier and is blue-detuned
by the acoustic resonant frequency Ωm from the optical resonance (Fig. 5(a)). We estimate that
∼ 150 mW of optical power is coupled into the suspended LN waveguide, resulting in ∼ 50 mW
(a)-20-10010200.00.20.40.60.81.0Optical transmissionOptical detuning (pm)l =1574.9 nmOptical loaded 6Q = 2.2×101.01.52.02.53.0-60-50-40-30-20-100Acoustic frequency (GHz)-8-6-4-20Opto-acoustic S (dB)21Acoustic S (dB)11Optical pump ωpΔ = ΩmOptical mode ω0Opticalsidebandω+ΩpmOpticalsidebandω−Ωpm(b)(c)Frequency Fig. 5. Demonstration of a microwave-photonic link. (a) Experimental schematic. EDFA:
Erbium-doped fiber amplifier. (b) Optical transmission of the racetrack cavity for different
microwave powers. (c) S21 spectrum features a peak microwave power transmission of ∼ 0
dB. The optical power received at the photodiode Irec is 50 mW. (d) Microwave spectrum of
the photodiode output signal with a microwave power of 5 mW applied to the IDT of the
thin-film acoustic resonator. The laser is blue-detuned from the optical mode by the acoustic
resonant frequency Ωm
reaching the photodiode. Importantly, no damage to the waveguide is observed, indicating the
ability of our suspended thin-film LN devices to handle large optical powers. A high power
photodiode, with responsivity RPD of 0.55 A/W (corresponding to a quantum efficiency of 40%),
is used to detect and down-convert the optical signal that we generate from our our racetrack
acousto-optic transducer back to the microwave domain. With these parameters, we measure an
overall microwave link gain of 0 dB at 1.572 GHz (Fig. 5(c)) for a small microwave input signal
at -20 dBm. By reducing the fiber-to-chip coupling loss to the previously-demonstrated value of
1.7 dB/facet [40], a microwave link with gain of 6.6 dB may be achieved in principle, with the
possibility of a gain of at least 9 dB if tapered fibers are employed [41,42].
Next, we characterize the response at higher powers of microwave input. With increasing
microwave powers at the acoustic resonant frequency Ωm ∼2 GHz, optical sideband dips are
observed in the transmission spectra (Fig. 5(b)), which agree with theoretical predictions (Fig. S3).
The red-shift of the optical resonance with increasing input microwave powers results from the
heating of the acoustic resonator. As a result of efficient microwave-to-optical conversion, a pair
of second order sideband dips can be observed in the optical transmission spectrum with only 5
mW of microwave input power.
Parking the laser at Ωm detuning from the optical mode, up to 3rd order harmonic signals are
123456-90-70-50-30Power spectral density (dB)Optical beating frequency (GHz)1.551.561.571.581.59-50-40-30-20-100Acoustic frequency (GHz)Opto-acoustic S (dB)21(d)(c)1578.81578.91579.01579.11579.2012345Optical transmissionWavelength (nm)Off1 mW2 mW3 mW5 mW microwave power(b)Tunable laserMicrowave inputAcousto-optic cavityEDFAPhotodiodeMicrowave outputDC bias(a) observed at the photodiode output with a microwave power of 5 mW (Fig. 5(d)). The additional
broad peak observed at 3.85 GHz is a result of the suspended optical racetrack cavity since it
is only observed when the pump laser is close to the optical resonance. We speculate that the
acoustic mode along the suspended optical waveguide causes this additional peak by spontaneous
Brillouin scattering, as our suspended racetrack cavity has the similar geometry with that in
Ref. [43].
7. Conclusions and outlook
We demonstrate an integrated acousto-optic platform on thin film LN, which converts acoustic
waves in microwave domain to optical light by a generalized acousto-optic interaction. Efficient
microwave-to-acoustic coupling has been achieved using our IDT-coupled LN thin-film acoustic
resonator. This addresses the coupling issue of current OMC-based platforms using mechanically-
mediated microwave-to-optical converters. To further improve the photon number conversion
efficiency, a variety of efforts in acoustic and photonic engineering can be made. For example,
the acoustic resonator can be operated under vacuum and cryogenic environments to achieve
higher Q factors, and the clamping loss of the suspended structures can be further reduced using
phononic crystals [20]. Optical cavities defined by photonic crystal mirrors could improve the
acousto-optic coupling strength g0 with larger overlap with the acoustic resonator, while the
presented racetrack cavity only partially sits in the acoustic resonator. Another order-of-magnitude
improvement could be obtained by bringing both the pump light and generated optical sideband
into resonance [3]. With a double optical resonance, which can be found in coupled cavities or
due to scattering in a single ring cavity, the term Ω2 (∼ GHz) in the denominator of intracavity
optical photon number ncav (Eq. 3) is replaced by the optical cavity loss κ2 (∼ 10s of MHz) and
could result in an improvement of 4 orders of magnitude.
Beyond microwave-to-optical conversion, our acousto-optic platform could also find appli-
cations in gigahertz frequency optical comb generation, on-chip optical routing and optical
mode conversion. In these applications, the acoustic resonator could strongly enhance the signal
in microwave domain and allow low microwave power operations. Compared to microwave
electromagnetic resonators, the high acoustic f Q product and smaller acoustic mode volume of
our on-chip acousto-optic resonator could enable quantum optomechanics at room temperatures
with smaller footprints.
Acknowledgment
This work is supported by the STC Center for Integrated Quantum Materials, NSF Grand
No. DMR-1231319, NSF E2CDA Grand No. ECCS-1740296, NSF CQIS Grand No. ECCS-
1810233, ONR MURI Grant No. N00014-15-1-2761, and NSF Grant No. DMR-1707372.
N.S. acknowledges the support of the Natural Sciences and Engineering Research Council of
Canada (NSERC) and the AQT Intelligent Quantum Networks and Technologies (INQNET)
research program.
References
1. D. Marpaung, J. Yao, and J. Capmany, "Integrated microwave photonics," Nat. Photonics 13, 80 -- 90 (2019).
2. G. Wendin, "Quantum information processing with superconducting circuits: a review," Rep. Prog. Phys 80, 106001
(2017).
3. A. Rueda, F. Sedlmeir, M. C. Collodo, U. Vogl, B. Stiller, G. Schunk, D. V. Strekalov, C. Marquardt, J. M. Fink,
O. Painter, G. Leuchs, and H. G. L. Schwefel, "Efficient microwave to optical photon conversion: an electro-optical
realization," Optica 3, 597 (2016).
4. K. Stannigel, P. Rabl, A. S. Sørensen, P. Zoller, and M. D. Lukin, "Optomechanical transducers for long-distance
quantum communication," Phys. Rev. Lett. 105, 220501 (2010).
5. M. Tsang, "Cavity quantum electro-optics," Phys. Rev. A 81, 063837 (2010).
6. M. J. A. Schuetz, E. M. Kessler, G. Giedke, L. M. K. Vandersypen, M. D. Lukin, and J. I. Cirac, "Universal quantum
transducers based on surface acoustic waves," Phys. Rev. X 5, 031031 (2015).
7. L. Fan, C.-L. Zou, N. Zhu, and H. X. Tang, "Spectrotemporal shaping of itinerant photons via distributed
nanomechanics," Nat. Photonics 13, 323 -- 327 (2019).
8. A. P. Higginbotham, P. S. Burns, M. D. Urmey, R. W. Peterson, N. S. Kampel, B. M. Brubaker, G. Smith, K. W.
Lehnert, and C. A. Regal, "Harnessing electro-optic correlations in an efficient mechanical converter," Nat. Phys. 14,
1038 -- 1042 (2018).
9. R. W. Andrews, R. W. Peterson, T. P. Purdy, K. Cicak, R. W. Simmonds, C. A. Regal, and K. W. Lehnert, "Bidirectional
and efficient conversion between microwave and optical light," Nat. Phys. 10, 321 -- 326 (2014).
10. T. Bagci, A. Simonsen, S. Schmid, L. G. Villanueva, E. Zeuthen, J. Appel, J. M. Taylor, A. Sorensen, K. Usami,
A. Schliesser, and E. S. Polzik, "Optical detection of radio waves through a nanomechanical transducer," Nature 507,
81 -- 85 (2014).
11. M. Forsch, R. Stockill, A. Wallucks, I. Marinkovic, C. GÃďrtner, R. A. Norte, F. van Otten, A. Fiore, K. Srinivasan,
and S. Gröblacher, "Microwave-to-optics conversion using a mechanical oscillator in its quantum ground state," arXiv
e-prints p. arXiv:1812.07588 (2018).
12. W. Jiang, R. N. Patel, F. M. Mayor, T. P. McKenna, P. Arrangoiz-Arriola, C. J. Sarabalis, J. D. Witmer, R. V. Laer,
and A. H. Safavi-Naeini, "Lithium niobate piezo-optomechanical crystals," Optica 6, 845 -- 853 (2019).
13. K. C. Balram, M. I. Davanco, J. D. Song, and K. Srinivasan, "Coherent coupling between radio frequency, optical,
and acoustic waves in piezo-optomechanical circuits," Nat. Photonics 10, 346 -- 352 (2016).
14. A. Vainsencher, K. J. Satzinger, G. A. Peairs, and A. N. Cleland, "Bi-directional conversion between microwave and
optical frequencies in a piezoelectric optomechanical device," Appl. Phys. Lett. 109 (2016).
15. J. Bochmann, A. Vainsencher, D. D. Awschalom, and A. N. Cleland, "Nanomechanical coupling between microwave
16. L. Fan, X. Sun, C. Xiong, C. Schuck, and H. X. Tang, "Aluminum nitride piezo-acousto-photonic crystal nanocavity
and optical photons," Nat. Phys. 9, 712 -- 716 (2013).
with high quality factors," Appl. Phys. Lett. 102 (2013).
17. G. S. MacCabe, H. Ren, J. Luo, J. D. Cohen, H. Zhou, A. Sipahigil, M. Mirhosseini, and O. Painter, "Phononic
bandgap nano-acoustic cavity with ultralong phonon lifetime," arXiv e-prints p. arXiv:1901.04129 (2019).
18. H. Liang, R. Luo, Y. He, H. Jiang, and Q. Lin, "High-quality lithium niobate photonic crystal nanocavities," Optica 4,
1251 (2017).
19. K. C. Balram, M. Davanco, J. Y. Lim, J. D. Song, and K. Srinivasan, "Moving boundary and photoelastic coupling in
GaAs optomechanical resonators," Optica 1, 414 -- 420 (2014).
20. K. Fang, M. H. Matheny, X. Luan, and O. Painter, "Optical transduction and routing of microwave phonons in
cavity-optomechanical circuits," Nat. Photonics 10, 489 -- 496 (2016).
21. M. J. Burek, J. D. Cohen, S. M. Meenehan, N. El-Sawah, C. Chia, T. Ruelle, S. Meesala, J. Rochman, H. A. Atikian,
M. Markham, D. J. Twitchen, M. D. Lukin, O. Painter, and M. Lončar, "Diamond optomechanical crystals," Optica 3,
1404 (2016).
22. M. Eichenfield, J. Chan, R. M. Camacho, K. J. Vahala, and O. Painter, "Optomechanical crystals," Nature 462, 78
(2009).
CA, 1989).
23. M. Lejman, G. Vaudel, I. C. Infante, I. Chaban, T. Pezeril, M. Edely, G. F. Nataf, M. Guennou, J. Kreisel, V. E.
Gusev, B. Dkhil, and P. Ruello, "Ultrafast acousto-optic mode conversion in optically birefringent ferroelectrics," Nat.
Commun. 7, 12345 (2016).
24. A. S. Andrushchak, B. G. Mytsyk, H. P. Laba, O. V. Yurkevych, I. M. Solskii, A. V. Kityk, and B. Sahraoui,
"Complete sets of elastic constants and photoelastic coefficients of pure and MgO-doped lithium niobate crystals at
room temperature," J. Appl. Phys. 106 (2009).
25. R. S. Weis and T. K. Gaylord, "Lithium niobate: Summary of physical properties and crystal structure," Appl. Phys.
A 37, 191 -- 203 (1985).
26. C. Wang, M. Zhang, X. Chen, M. Bertrand, A. Shams-Ansari, S. Chandrasekhar, P. Winzer, and M. Lončar,
"Integrated lithium niobate electro-optic modulators operating at CMOS-compatible voltages," Nature 562, 101 -- 104
(2018).
27. C. Campbell, Surface Acoustic Wave Devices and their Signal Processing Applications (Academic Press, San Diego,
28. C. S. Tsai, "Wideband acousto-optic bragg diffraction in linbo3 waveguide and applications," in Guided-Wave
Acousto-Optics: Interactions, Devices, and Applications, C. S. Tsai, ed. (Springer Berlin Heidelberg, Berlin,
Heidelberg, 1990), pp. 117 -- 203.
29. S. A. Tadesse and M. Li, "Sub-optical wavelength acoustic wave modulation of integrated photonic resonators at
microwave frequencies," Nat. Commun. 5, 5402 (2014).
30. S. Kapfinger, T. Reichert, S. Lichtmannecker, K. Muller, J. J. Finley, A. Wixforth, M. Kaniber, and H. J. Krenner,
"Dynamic acousto-optic control of a strongly coupled photonic molecule," Nat. Commun. 6, 8540 (2015).
31. D. B. Sohn, S. Kim, and G. Bahl, "Time-reversal symmetry breaking with acoustic pumping of nanophotonic circuits,"
Nat. Photonics 12, 91 -- 97 (2018).
32. L. Shao, S. Maity, L. Wu, A. Shams-Ansari, Y.-I. Sohn, E. Puma, M. N. Gadalla, M. Zhang, C. Wang, and M. Lončar,
33. Y. Yang, R. Lu, T. Manzaneque, and S. Gong, "1.7 GHz Y-cut lithium niobate MEMS resonators with FoM of 336
"High-Q gigahertz surface acoustic wave cavity on lithium niobate," arXiv e-prints: 1901.09080 (2019).
and fQ of 9.15 × 1012," in 2018 IEEE/MTT-S International Microwave Symposium - IMS, (2018), pp. 563 -- 566.
34. M. Zhang, C. Wang, R. Cheng, A. Shams-Ansari, and M. Lonvar, "Monolithic ultra-high-Q lithium niobate microring
resonator," Optica 4, 1536 -- 1537 (2017).
35. B. Desiatov, A. Shams-Ansari, M. Zhang, C. Wang, and M. Lončar, "Ultra-low-loss integrated visible photonics
using thin-film lithium niobate," Optica 6, 380 -- 384 (2019).
36. M. Cherchi, S. Ylinen, M. Harjanne, M. Kapulainen, and T. Aalto, "Dramatic size reduction of waveguide bends on a
micron-scale silicon photonic platform," Opt. Express 21, 17814 -- 23 (2013).
37. L. Marculescu and G. Hauret, "Étude de l'effet brillouin à température ordinaire dans le niobate de lithium," Compt.
1113 -- 1116 (2018).
2005).
Rend. Acad. Sci. (B) 276, 555 -- 558 (1973).
38. L. Zheng, H. Dong, X. Wu, Y.-L. Huang, W. Wang, W. Wu, Z. Wang, and K. Lai, "Interferometric imaging of
nonlocal electromechanical power transduction in ferroelectric domains," Proc. Natl. Acad. Sci. 115, 5338 -- 5342
(2018).
39. L. Zheng, D. Wu, X. Wu, and K. Lai, "Visualization of surface-acoustic-wave potential by transmission-mode
microwave impedance microscopy," Phys. Rev. Appl. 9, 061002 (2018).
40. L. He, M. Zhang, A. Shams-Ansari, R. Zhu, C. Wang, and L. Marko, "Low-loss fiber-to-chip interface for lithium
niobate photonic integrated circuits," Opt. Lett. 44, 2314 -- 2317 (2019).
41. M. J. Burek, C. Meuwly, R. E. Evans, M. K. Bhaskar, A. Sipahigil, S. Meesala, B. Machielse, D. D. Sukachev, C. T.
Nguyen, J. L. Pacheco, E. Bielejec, M. D. Lukin, and M. Lončar, "Fiber-coupled diamond quantum nanophotonic
interface," Phys. Rev. Appl. 8, 024026 (2017).
42. A. Sipahigil, R. Evans, D. Sukachev, M. Burek, J. Borregaard, M. Bhaskar, C. Nguyen, J. Pacheco, H. Atikian, and
C. Meuwly, "An integrated diamond nanophotonics platform for quantum-optical networks," Science 354, 847 -- 850
(2016).
43. N. T. Otterstrom, R. O. Behunin, E. A. Kittlaus, Z. Wang, and P. T. Rakich, "A silicon brillouin laser," Science 360,
44. R. E. Newnham, Properties of materials: anisotropy, symmetry, structure (Oxford University Press, New York,
Supplementary Material
1. Numerical simulation of the acousto-optic interaction
We perform a 2D numerical simulation of our device cross-section (Fig. S1(a)) using COMSOL
Multiphysics. Optical and acoustic modes are simulated independently and the acousto-optic
interactions are then calculated by the integral of acoustic and optical fields using corresponding
nonlinear coefficient matrices.
1.1. Simulation of optical and acoustic modes
The single-mode optical waveguide of our device supports fundamental TE and TM modes
(Fig. S1). The electric field profiles of the optical modes are used in the calculation of the
acousto-optic interaction.
Fig. S1. (a) Device structure for 2D numerical simulation. (b), (c) Optical electric field of
the fundamental TE and TM modes, respectively.
The simulation of the acoustic mode includes strain, electric field, and the piezoelectric
effect. Multiple acoustic modes with gigahertz resonant frequencies are found in the eigenmode
simulation. We plot only a few acoustic modes in Fig. S2. The electrical excitation of these
acoustic modes are enabled by the interdigital transducers (IDTs).
1.2. Calculation of acousto-optic interactions
The acousto-optic interactions are calculated by integrating the optical and acoustic modes with
matrices that describe moving boundary, photoelastic and electro-optic effects. Calculations here
are based on theory formulated in previous works [19,21,22,44].
In our work, the acousto-optic interactions are described by the change of optical mode index
due to the acoustic mode. The acoustic mode amplitude α, defined by the maximum displacement,
is normalized to a single phonon occupation of the acoustic resonator using Ω = 1
2 meffΩ2α2,
where Ω is the acoustic frequency. The effective mass meff of the acoustic mode is given by
meff = La
(S1)
where D defines the 2D simulation domain and the coordinate variable r ∈ D. La is the length
(perpendicular to the simulation cross-section) of the acoustic resonator, ρ is material mass
density, and Q is the displacement field.
D
D
Þ
ρ Q(r)2d r(cid:14) max
(cid:16)
Q(r)2(cid:17)
,
Lithium niobate thin filmMetal electrodesOpticalwaveguide(a)(b)(c)TE modeTM mode Fig. S2. Strain and electric field of three simulated acoustic modes with resonant frequencies
of 1.55, 2.40, and 3.24 GHz. The color map is independently normalized for each simulation.
The electric field of the optical mode is denoted as E, s refers to strain, and E refers to the
electric field of the acoustic mode. The mode index modulated by the moving boundary effect is
given by
Þ (Q · n)(cid:16)
(cid:17)
(cid:107) ∆E(cid:107) − D∗⊥∆−1D⊥
E∗
Þ E∗Edr
∆n0,MB = − n
2
dS
,
(S2)
where n is the optical mode index, n is the normal vector of the boundary facing outward, and D
is the electric displacement field of the optical mode. The subscripts (cid:107) and ⊥ indicate the parallel
and perpendicular components to the boundary. The permittivity for the optical electric field is
denoted as , while ∆ = LN − air, and ∆−1 = −1
The mode index modulated by the photoelastic effect is given by
(cid:16)
Þ dr
E∗
x E∗
y E∗
z
LN − −1
air .
(cid:17)(cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171)
Þ E∗Edr
dB1
dB6
dB5
(cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172)
(cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172)
Ex
Ey
Ez
dB6
dB2
dB4
dB5
dB4
dB3
,
(S3)
where 0 is the vacuum permittivity, and Bk (k = 1 − 6) is the optical indicatrix. The changes of
indicatrix coefficient dBk (k = 1 − 6) due to the strain sk (k = 1 − 6) is given by
(cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171)
(cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172)
(cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172)
(cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171)
s1
s2
s3
s4
s5
s6
∆n0,PE =
0n5
2
(cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171)
dB1
dB2
dB3
dB4
dB5
dB6
(cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172)
(cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171)
=
p33
p13
p13
0
0
0
p31
p31
p11
p12
p12
p11
0
0
p41 −p41
0
0
0
0
0
p66
0
p41
0
p14
−p14
0
p44
0
0
0
0
p14
0
p44
.
(S4)
where pjk are the primary elasto-optic coefficients in the condition of a constant electric field
for lithium niobate (LN), where the secondary effect via piezoelectricity and electro-optics is
excluded [25,37]. The photoelastic matrix is rotated according to the crystal orientation in our
3.24 GHz1.55 GHz2.40 GHzAcoustic resonant frequency StrainElectric field0-11-0.50.5Amplitude (arb. unit) device -- X-cut thin-film LN with acoustic wave propagating in the Z direction of the crystal. The
coordinate representations for the simulation and crystal are shown in Fig. 2(a).
The mode index modulated by the electro-optic effect ∆n0,EO is of the same form of Eq. S3,
with the changes of indicatrix coefficients [44] given by
(cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172)
0
0
0
−r22
0
−r51
0
−r22
r22
0
r51
0
(cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171)
dB1
dB2
dB3
dB4
dB5
dB6
(cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172)
(cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171)
r33
r13
r13
0
0
0
=
(cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171)
Ex
Ey
Ez
(cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172) ,
(S5)
where rjk is the primary electro-optic coefficients in the condition of constant strain in which
secondary effects via piezoelectricity and photoelasticity is excluded. The above matrix is rotated
according to the crystal orientation in our device.
The overall relative refractive index change due to a single phonon is given by,
∆n0,tot = ∆n0,MB + ∆n0,PE + ∆n0,EO.
(S6)
Table S1. Numerical simulation results of acousto-optic interactions
Optical
mode
Acoustic
mode freq.
∆n0,MB
∆n0,PE
∆n0,EO
∆n0,tot
MZI
AO cavity
GHz
1.55
2.17
2.40
3.16
3.24
1.55
2.17
2.40
3.16
3.24
TE
TE
TE
TE
TE
TM
TM
TM
TM
TM
×10−12
0.36
-0.84
-0.80
-3.07
-4.24
4.19
8.39
0.87
15.09
21.03
×10−12
1.40
1.11
-5.08
41.68
47.84
24.29
67.97
22.26
42.06
70.63
×10−12
30.5
71.43
23.06
26.38
58.91
-8.79
-24.03
-9.37
-4.28
-15.75
×10−12
32.26
71.70
17.17
64.99
102.5
19.69
52.33
13.72
52.87
75.90
Vπ L
V·cm
0.0692
0.0436
0.2009
0.0703
0.0454
0.113
0.0599
0.2505
0.0863
0.0616
Vπ
V
6.92
4.36
20.1
7.03
4.54
11.3
5.99
25.1
8.63
6.16
g0
kHz
0.5
1.1
0.27
1.0
1.6
0.3
0.8
0.2
0.8
1.2
1.3. Calculation of Vπ L
The half-wave-voltage-length product Vπ L, characterizing the modulation efficiency, defines the
voltage that is required to achieve a π phase shift for a modulation length L. Here, we derive the
Vπ L from the simulated refractive index changes with additional information on Q factors and
coupling of the acoustic resonator. While the overall refractive index change in Eq. S6 quantifies
the optical phase shift (or index change) due to a single phonon in the acoustic resonator, one
must relate the in-cavity phonon number to the applied microwave power. As discussed later in
Sec. 3, the in-cavity phonon number is given by
Npn =
4γe
γ2 Nin =
(cid:14)Ωm is the phonon input rate with the resonant frequency Ωm of the acoustic
where Nin = Pin
mode and input power Pin. The decay are and external coupling rates of the acoustic mode is γ
and γe, respectively. Given the input impedance Rin = 50Ω, the relation between input power
and peak voltage Vp is given by
Pin
Ωm
,
4γe
γ2
(S7)
The number of in-cavity phonons Npn required for a π phase shift is given by,
(S8)
(S9)
Pin =
2π
λ
∆n0,tot
1
2
V2
p
Rin
.
(cid:112)NpnL = π
(cid:115)
where λ is the optical wavelength. Taking Eqs. S7 and S8 in to Eq. S9, we derive the Vπ L of the
device:
Vπ L =
λ
2∆n0,tot
γ2ΩmRin
2γe
.
(S10)
1.4. Calculation of acousto-optic single-photon coupling strength g0
For our thin-film acoustic resonator that is coupled to an optical racetrack cavity, the acousto-optic
single-photon coupling strength g0 can be derived using the 2D simulation results using
g0 = ω0ηcav
∆n0,tot
n
,
(S11)
where ω0 is the optical resonant frequency, and ηcav is the ratio of waveguide length in the
acoustic resonator to that of the racetrack cavity.
1.5. Estimate Vπ L and g0 using the numerical simulation results
We estimate the Vπ L for the MachâĂŞZehnder interferometer (MZI) and g0 for the acousto-optic
cavity from simulation. To be consistent with the experiments, the typical measured acoustic
Q factors Qm = 2, 000 (γ = Ωm/Qm) and γe/γ = 0.15 (corresponding to a 3 dB dip in S11
measurements) are employed in the following calculation. The length of the acoustic resonator (in
direction perpendicular to the simulation cross-section) is La = 100 µm. The output impedance
of the microwave source is Rin = 50 Ω. For the acoustic-optic cavity shown in Fig. 1, the relative
length of the optical waveguide in the acoustic resonator is ηcav = 0.15. Table S1 summarizes
the interactions between optical modes and acoustic modes.
2. Derivation of Vπ from experimental measurements
2.1. Acousto-optic Mach-Zehnder interferometer
Here we relate the half-wave voltage Vπ to the measured opto-acoustic S21 for the acousto-optic
MZI. The phase modulation of one optical path is given by
Ep1(V) =
E0√
2
exp(iπV/Vπ + iφb) ,
(S12)
where E0 is the input optical field of the MZI, φb is the bias phase between two optical paths,
field is given by Ep2(V) = E0/√
and V is the applied voltage. The other optical path of MZI is not modulated, and the optical
2. The optical field at the output of the MZI is given by
E0
2
The output optical power is thus given by
Iout(V) ∝ E∗
=
Eout(V) =
Ep1(V) + Ep2(V)
√
2
(1 + exp(iπV/Vπ + iφb)) .
out Eout
E02
2
=
(1 + cos(πV/Vπ + φb)) .
(S13)
(S14)
(cid:18) πRPD Irec
(cid:19)2
Vπ
The optimum microwave to optical conversion occurs at the bias phase φb = π/2, which
corresponds to the output intensity at half maximum. Measured using a potodetector, the
opto-acoustic S21 for small input signal is given by
S21 =
,
(S15)
where Irec is the DC optical power received at the photodetector, and RPD is sensitivity of the
photodetector. Using Eq. S15, we can derive Vπ of the acousto-optic MZI by the opto-acoustic
S21 measurements.
2.2. Acousto-optic cavity
Our acousto-optic cavity operates in the sideband resolved regime, that is the frequency of the
microwave signals are greater than the decay rate of the optical mode. For a weak microwave
signal, the optical transmission is thus close to unitary at the optimum conversion wavelength,
which corresponds to that detuned from the optical resonance by the microwave frequency.
Phenomenologically, this can be understood as the light being reversibly pumped into, and out of,
the optical cavity due to the acoustic modulation. Thus, we consider the acousto-optic cavity as
an intensity modulator and the relation in Eq. S15 is also used to derive the effective Vπ.
3. Conversion between microwave, acoustic, and optical fields in acousto-optic
cavity
3.1. Dynamics of acousto-optic cavity
Here we consider an acousto-optic system with an acoustic resonator driven by a microwave
signal through the piezoelectric effect. The Heisenberg-Langevin equations of motion for an
optical cavity a coupled to an acoustic resonator b are given by
√
(cid:17)
+
κeain
(S16)
(cid:219)a = −(cid:16)
(cid:219)b = −(cid:16)
(cid:16)
b + b†(cid:17)
(cid:17)
i∆ +
κ
2
iΩm +
a − ig0a
b − ig0a†a +
γ
2
√
γebin,
(S17)
where a and b are the annihilation operators of optical and acoustic modes, respectively, g0 is
the single-photon coupling strength between the optical and acoustic resonators, ∆ = ω0 − ωp
is the optical detuning with the pump laser frequency ωp, the optical resonant frequency is ω0,
κ = κi + κe is the loss of optical mode with intrinsic loss κi and external coupling rate κe, Ωm is
the acoustic resonant frequency, γ = γi + γe is the loss of acoustic mode with intrinsic loss γi and
external coupling rate γe, and ain and bin are the optical and microwave input field, respectively.
To solve the equations of motion, we consider a single frequency microwave driving bin of the
acoustic resonator given by
bin = Bine−iΩd t,
(S18)
where Ωd is the driving frequency, Bin is the amplitude of the input field, and the input microwave
power is Pin = Ωm Bin2. In the weak optical mode limit, i.e. g0a†a (cid:28) Ωm, the optical back
action term (ig0a†a in Eq. S17) on the acoustic resonator is neglected. Taking Eq. S18 into
Eq. S17, the acoustic amplitude b is solved using
b = Be−iΩd t
√
(S19)
γe
B =
i (Ωm − Ωd) + γ
2
Bin.
For a resonant microwave drive (Ωm = Ωd), the in-resonator phonon number Npn is related to
the input microwave power by
in
=
Npn = B2
4γe
γ2 B2
4γe
γ2 Nin
4γe
Pin
Ωm
γ2
=
=
(S20)
.
Taking Eq. S19 into Eq. S16, the equation of motion for the optical mode is re-written as
a − iG 2 cos(Ωdt) a +
√
κeain,
(S21)
(cid:219)a = −(cid:16)
(cid:17)
i∆ +
κ
2
where G = g0B is the frequency shift of optical mode due to the acoustic field that is present.
T =(cid:12)(cid:12)ain − √
κea(cid:12)(cid:12)2(cid:14) ain2 .
3.2. Optical transmission with active acoustic driving
We numerically solve Eq. S21 to investigate the optical transmission spectra with various
microwave input powers. We note that the Eq. S21 assumes a weak optical input and a linear
acoustic resonator. The normalized optical transmission T under a continuous optical pump ain
is given by
(S22)
As the optical mode is being modulated by an acoustic mode at microwave frequency Ωd, the
optical transmission T is expected to associate an oscillation at the same as well as higher order
frequencies due to nonlinearity. However, in experiment, the optical transmission spectra are
captured by a low frequency (10 MHz) data acquisition card, which does not respond to gigahertz
frequencies. Numerically, we use an average to calculate the quasi-DC component of the optical
transmission using
dτT(τ),
(S23)
Þ t1+∆t
TDC =
1
∆t
t1
where time t1 is set to be greater than the initial stabilization time in numerical calculation of a,
and the average time window ∆t is chosen to be the integer periods of the driving signal, i.e. N/Ω.
The numerically-calculated optical transmission spectra (Fig. S3) exhibit sidebands in agree-
ment with the experimental measurements in Fig. 3 in the main text.
Fig. S3. Numerically calculated optical transmission spectra of acousto-optic cavity with
microwave input powers from 0 to 5 mW.
S parameter of acousto-optic cavity
3.3.
The parameter S21 is defined as the normalized microwave power generated by the receiving
photodetector, which is generated by beating the pump laser and the optical sideband at the
photodetector. To derive the power in the optical sideband, we decompose the optical amplitude
a into a series of sidebands:
a =
Aqe−iqΩd t,
(S24)
where Aq is the amplitude of optical sideband of order q. At the weak microwave input power
(i.e. G (cid:28) κ) limit and in the sideband resolved regime (i.e. Ωm (cid:29) κ), we only consider the first
order of optical sidebands, i.e. q = 0, ±1 and, for simplicity, we write the amplitude as, A0, A+,
A−. The Eq. S21 is thus decomposed into sidebands,
q
(cid:17)
(cid:17)
(cid:17)
0 = −(cid:16)
−iΩm A+ = −(cid:16)
iΩm A− = −(cid:16)
i∆ +
i∆ +
i∆ +
κ
2
κ
2
κ
2
√
A0 − iG (A+ + A−) +
A+ − iGA0
A− − iGA0,
κe Ain
(S25)
-6-4-20246012345Optical transmissionOptical detuning (GHz)Off1 mW2 mW3 mW5 mW microwave power where Ain is the input optical amplitude. The solution of Eq. S25 is given by
√
κe Ain
1
i(∆−Ωd)+κ/2 +
1
i(∆+Ωd)+κ/2
(cid:17)(cid:17)
i∆ + κ/2 + G2(cid:16)
(cid:16)
√
κe Ain
(i∆ + κ/2)
−iGA0
i (∆ − Ωd) + κ/2
−iGA0
i (∆ + Ωd) + κ/2
A0 =
(cid:39)
A+ =
A− =
For the scenario the pumping laser is blue detuned from the optical resonance by the acoustic
resonant frequency (∆ = −Ωm), and the microwave input is on resonant with the acoustic mode
(Ωd = Ωm), the in-cavity optical amplitude for the enhanced sideband A− is given by,
where acousto-optic coupling strength G = g0B = 2g0Bin
Since the pump laser is detuned from the resonant, the transmitted amplitude of the pump laser
is close to the input Ain. The output microwave voltage U from the photodetector caused by the
beating between the transmitted pump laser and the generated optical sideband given by
(S26)
(S27)
(S28)
(S29)
(S30)
A− =
√
−iG
κe Ain
(−iΩm + κ/2) κ/2,
√
U = RPD ω(cid:12)(cid:12)√
(cid:112)
= RPD
(cid:39) RPD
κ
GκeIopt
Ωmκ/2 ,
κe A− Ain
GκeIopt
m + κ2/4/2
Ω2
γe/γ.
(cid:12)(cid:12)
where optical power Iopt = ω0 A2
photodetector in the unit of V/W. The output microwave power is then given by
in, ω0 is the optical frequency, and RPD is the response of the
Pout =
=
U2
2Rload
2G2κ2
Ω2
e R2
PD I2
mκ2Rload
opt
,
(S31)
where Rload = 50 Ω is the impedance of the network analyzer. The opto-acoustic transmission
S21 is given by
S21 = Pout/Pin
8g2
e R2
PD I2
0 γe κ2
γ2Ω3
mκ2Rload
=
opt
(S32)
3.4. Estimation of acousto-optic single-photon coupling strength g0 from experimental
measurements
Using the experimental results of the acousto-optic cavity (Fig. 3 in the main text), we can extract
the single-photon coupling strength g0 using Eq. S32. Taking the insertion loss of the chip
into account, the input optical power Iopt in Eq. S32 is replaced by the power received at the
photodetector Ir ec.
We extract the acousto-optic coupling strength g0 = 1.1 kHz from the experimental results
shown in Fig. 3 and summarized in Table S2. We note this experimentally-extracted g0 is
in good agreement with the numerically-simulated value (TE mode, 2.17 GHz) in Table S1.
The discrepancy of the acoustic resonant frequency between the numerical simulation and
experimental measurement may due to the deviation in LN film thickness and etching depth in
fabrication.
3.5. Photon number conversion efficiency
The photon number conversion efficiency η relates the number of generated optical sideband
photons coupled out of the cavity
κe A− to the input microwave photons. For weak microwave
input signals, the conversion efficiency η is given by
√
(cid:12)(cid:12)(cid:12)(cid:12)√
(cid:12)(cid:12)(cid:12)(cid:12)2
η =
=
=
κe A−
Bin
16g2
0 γe κ2
ω0Ω2
4g2
0
γ κ
e Iopt
mγ2κ2
· κeIopt
mω0
Ω2
= C0 · ncav · 2γe
γ
(S33)
· 2κe
κ
· 2γe
γ
· 2κe
κ
,
0/(γκ) is the single-photon cooperativity, ncav = κeIopt/(Ω2
ω0) is the intracavity
where C0 = 4g2
photon number of the blue-detuned pump light, 2κe/κ is the external coupling efficiency of the
optical mode, and 2γe/γ is the external coupling efficiency of acoustic mode by the IDT.
m
Table S2. Estimation of acousto-optic single-photon coupling strength g0 using experimental
results
Parameter
Optical mode
ω0/ 2π
κ/ 2π
2κe/κ
γ/ 2π
2γe/γ
Ωm/ 2π
RPD
Ir ec
Rload
S21
g0/ 2π
Value
TE
200 THz
95 MHz
0.3
1.28 MHz
0.34
2.007 GHz
800 V/W
0.128 mW
50 Ω
-7.5 dB
1.1 kHz
Using the experimentally-extracted values summarized in Table S2, we estimate a single-photon
cooperativity of C0 = 4 × 10−8. At an optical power of 1 mW, where the intracavity photon
number is only about 4,400 due to the large detuning of ∆ = −Ωm from the optical resonance,
the photon number conversion efficiency is η = 0.0017%.
4. Microwave microscopy of acoustic modes
We experimentally investigate the acoustic mode profiles using transmission-mode microwave
impedance microscopy [38, 39]. The working principle is the following -- while the acoustic
resonator is driven by a microwave input on the IDT, a probe for atomic force microscopy is
scanning over the acoustic resonator and measuring any microwave electric signals. The detected
signal is mixed with the driving signal to extract the relative amplitude and phase of the acoustic
electric field. The electric amplitude profile of an acoustic mode is obtained on the top surface
and in agreement with the numerical simulation (Fig. S4).
Fig. S4. Electric and topographic profiles of an acoustic mode. The electric field amplitude
is detected by a scanning probe using transmission-mode microwave impedance microscopy,
with a driving signal at 2 GHz on the acoustic resonance. Inset: simulated electric field
amplitude profile of the acoustic mode.
1 μm-2-1012302-2-1012Position x (μm) Simulation 1Measured waveguide profile (μm)Acoustic electric amplitude (V)Measurement |
1911.01033 | 1 | 1911 | 2019-11-04T05:09:31 | Ultra-wideband Reflection-type Metasurface for Generating Integer and Fractional Orbital Angular Momentum | [
"physics.app-ph",
"physics.class-ph"
] | Vortex beams carrying orbital angular momentum are extensively studied owing to its potential to expand channel capacity of microwave and optical communication. By utilizing the Pancharatnam-Berry phase concept, an ultra-wideband single layer metasurface is proposed to realize the conversion from incident plane waves to reflected vortex beams covering a considerable bandwidth from 6.75 to 21.85 GHz . An equivalent circuit model combined with broadband phase shift network is developed to effectively design the meta-atoms in metasurface. It is the first time to design wideband metasurfaces with the phase-based characteristics. To verify the proposed model, some deformed square loop meta-atoms are proposed to construct the metasurfaces with broadband OAM characteristic. Moreover, the vortex beams with the integer (l = -3), fractional (l = -1.5), and high-order (l = -10) OAM mode are generated respectively. Based on an OAM spectral analysis, the mode purity of the generated vortex waves is discussed in detail. The experimental results achieve a good agreement with those obtained from the simulation, thus proving the effectiveness and practicability of the proposed method. | physics.app-ph | physics | This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
1
Ultra-Wideband Reflection-Type Metasurface
for Generating Integer and Fractional Orbital
Angular Momentum
Ling-Jun Yang, Student Member, IEEE, Sheng Sun, Senior Member, IEEE, and Wei E. I. Sha, Senior Member,
IEEE
Abstract -- Vortex beams carrying orbital angular momentum
are extensively studied owing to its potential to expand channel
capacity of microwave and optical communication. By utilizing
the Pancharatnam-Berry phase concept, an ultra-wideband
single-layer metasurface is proposed to realize the conversion
from incident plane waves to reflected vortex beams covering a
considerable bandwidth from 6.75 to 21.85 GHz (>105%). An
equivalent circuit model combined with broadband phase shift
network is developed to effectively design the meta-atoms in
metasurface. It is the first time to design wideband metasurfaces
with the phase-based characteristics. To verify the proposed
model, some deformed square loop meta-atoms are proposed to
construct the metasurfaces with broadband OAM characteristic.
Moreover, the vortex beams with the integer (l = −3), fractional (l
= −1.5), and high-order (l = −10) OAM mode are generated
respectively. Based on an OAM spectral analysis, the mode purity
of the generated vortex waves is discussed in detail. The
experimental results achieve a good agreement with those
obtained from the simulation, thus proving the effectiveness and
practicability of the proposed method.
Index Terms -- Orbital angular momentum (OAM), fractional
OAM mode, ultra-wideband, metasurface, Pancharatnam-Berry
phase (PB).
I. INTRODUCTION
S
INCE orbital angular momentum (OAM) was known in
the optics field in 1992 [1], the vortex waves carrying OAM
have gradually become a hot field of research. Like spin
angular momentum (SAM), OAM is one of the fundamental
physical quantities of electromagnetic (EM) waves. SAM
associated with left and right circularly polarized EM waves
only offers limited channels. However, OAM can theoretically
achieve larger channel capacity by using the orthogonality of
different OAM modes [2], [3]. Therefore, different OAM
vortex beams have been applied in radio [4]-[6], optical [7], [8],
fiber [9], [10], and quantum communications [11] to achieve
Manuscript received xx xx, 2019; revised xx xx, 2019; accepted October xx,
2019. Date of publication December xx, 2019; date of current version xx xx,
2019. This work was supported in part by the National Natural Science
Foundation of China under Grant 61622106, 61721001, 61971115, and in part
by Sichuan Science and Technology Program under Grant 2018RZ0142.
(Corresponding author: Sheng Sun.)
L.-J. Yang, and S. Sun are with the School of Electronic Science and
Engineering, University of Electronic Science and Technology of China,
Chengdu 611731, China (e-mail: [email protected]).
Wei E. I. Sha is with the College of Information Science & Electronic
Engineering, Zhejiang University, Hangzhou, Zhejiang China 310027.
high spectral efficiency and communication capacity. However,
the issues of misalignment, doughnut-shaped pattern, and mode
crosstalk in OAM systems still deserve a further study. Some
excellent analytical methods have been proposed to understand
the limitations of those radiating systems [12]-[14]. Moreover,
some useful applications of OAM were also reported including
super-resolution imaging [15]-[17], structure field formation
[18], [19], optical tweezers [20], and astronomy [21].
It is well known that OAM vortex beams, characterized by
doughnut-shaped field, have helical phase front with the
(where l is the topological charge,
azimuthal phase term of
and φ is the azimuthal angle around the propagation axis).
Therefore, it is a basic principle to generate vortex beams with
the OAM mode l by introducing a constant electric current with
a consecutive phase of lφ along a circle. One common way is to
introduce the desired phase retardation by using spiral phase
plates [22], [23] and holographic plates [24]. Obviously, this
phase retardation relies on the light propagation over distances
much larger than the wavelength to shape the wave fronts,
which is difficult, if not totally impossible, to construct
ultrathin components. The circular antenna array [25]-[27] is
another approximated way to produce OAM vortex waves with
discrete currents based on Nyquist's theory. However, it
requires a complex feeding network. By exploiting the
Pancharatnam-Berry phase concept [28], one can produce
abrupt changes in phase, amplitude or polarization of EM
waves based on the ultrathin components, which is so-called
"metasurface". Resonant metasurfaces are composed of
resonant scattering units (V-shaped [29], [30], square loop
shaped [31]) with varied geometric parameters and can
generate
linearly polarized OAM waves. Unfortunately,
controlling abrupt phase through geometric parameters is hard
to achieve broadband and high performance. Several kinds of
metasurfaces have been designed to deal with the problem
based on the spin-to-orbital conversion (Pancharatnam-Berry
phase) theory [32]-[41]. In microwave regime, a PEC (perfect
electric conductor)-PMC
(perfect magnetic conductor)
metasurface was demonstrated to achieve nearly 100%
conversion efficiency within a narrow bandwidth [40]. A
four-layer metasurface was also proposed to achieve 33%
bandwidth and approximate 60% efficiency [36]. However,
those previous works mentioned still suffer from narrow
bandwidth, low efficiency, and bulky structure. Recently, the
ultra-thin reflection-type metasurfaces were proposed such as
the dual-layer metasurfaces with spatially rotated parallel
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
jle This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
2
dipoles [42] and the single-layer metasurfaces with spatially
rotated double arrow-shaped meta-atom [43]. Although they
can generate the vortex beams with a considerable wide
bandwidth, there still lack a reliable model to design these
meta-atom structures.
In this paper, an equivalent circuit model combined with
broadband phase shift network is proposed to design those
meta-atom structures. On one hand, the equivalent circuit
model based on lumped-element provides a physical insight on
the broadband EM behavior of those meta-atom structures. On
the other hand, this constructed equivalent circuit model can be
utilized to choose appropriate structures and parameters for
meta-atoms rather than time-consuming and aimless numerical
simulation. In particular, some effective equivalent circuit
models for deformed square loops are demonstrated. The
corresponding reflection-type metasurfaces composed of the
rotated deformed square loop meta-atoms are further designed
to convert the circularly polarized EM beams with SAM into
EM vortex waves with both SAM and OAM within an
ultra-wide frequency range. The generated integer, fractional,
and high-order vortex beams are decomposed and discussed in
detail. The simulated and experimental results verify the
proposed metasurfaces.
II. DESIGN THEORY OF PROPOSED METHOD
A. Spin-to-orbital Conversion
It is convenient to adopt a Jones formalism to analyze the
incident and scattered fields for an anisotropic meta-atom in
metasurface [44], [45]. The reflected and incident fields can be
connected by the reflection coefficients in reflected Jones
matrix, and the SAM-to-OAM process can be demonstrated as
[40]:
(1a)
(1b)
(1c)
(1d)
(
where rxx, ryy, rll, and rrr are the copolarized reflection
coefficients under x-, y-, left circularly, and right circularly
polarized normal incidence. And rxy, ryx, rlr, and rrl are the
corresponding cross-polarized reflection coefficients. From (1a)
and (1d), an abrupt phase change
) could be
introduced by a meta-atom with a rotating angle of kφ. A helical
phase wave-front could also be obtained by a metasurface
composed of rotated meta-atoms. Therefore, the first task is to
design a meta-atom with high copolarized
reflection
coefficients rrr and rll. If a meta-atom is mirror-symmetric with
respect to the y -- z plane or x -- z plane as shown in Fig. 1(b), the
corresponding rxy and ryx in Jones matrix satisfy rxy = ryx = 0 [45].
To generate a high SAM-to-OAM conversion within a wide
bandwidth, the rxx() and ryy() of meta-atom should satisfy
following conditions [46]:
(2)
Fig. 1. (a) Photography of vortex beams generating device. (b) The top view
of the meta-atom. (c) The side view of the meta-atom.
(3)
B. Equivalent Circuit Analysis
As shown in Fig. 1, a single-layer dielectric substrate (F4B,
εr = 2.65, h = 3 mm) is considered in this work to maintain
simple and ultra-thin features, which is the same as that
employed in [43]. Fig. 2(a) illustrates the corresponding
equivalent circuit model, where Z0 = 377 Ω is the wave
impedance of air. Zd = Z0/(εr)0.5 = 231.6 Ω, the wave impedance
of equivalent transmission line, is related to the dielectric
substrate. The length (h = 3 mm) of the equivalent transmission
line corresponds to the thickness of the dielectric layer. The
metal ground is equivalent to a short circuit. Zi (i{1,2}), the
equivalent impedances in x and y polarization incident fields
respectively, depend on the specific periodic metal structure
printed on the dielectric. Some valid equivalent lumped circuit
models of Zi have already been built for some common periodic
meta-atom structures in literature [47]-[50], such as the mesh of
metal strips (equivalent inductance), the array of metal patch
(equivalent capacitance) and the array of square loop (a series
LC circuit). According
the wideband model of
reflection-type phase-shifter [51], a series-connected LC
network is important to makes its reflection coefficient (rxx, ryy)
easier to satisfy (3) within a wide frequency range. Therefore,
the array of square loops is a promising candidate to generate
wideband vortex beams. Without the dielectric and metal loss,
their modulus of reflection coefficients rxx and ryy in Fig. 2 are
invariably equal to one and a high reflectance in (2) can be
achieved within the considered frequency range. In order to
produce the 180 phase difference in (3), a deformed square
loop meta-atom structure is proposed in Fig. 1(b) and Fig. 1(c),
and its equivalent electric circuit is shown in Fig. 2(b).
to
Reflection coefficients rxx and ryy of corresponding circuits
can be written as
(4a)
(4b)
where
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
20.5[()()]jkllxxyyxyyxrrrjrre−=−++0.5[()()]lrxxyyyxxyrrrjrr=++−0.5[()()]rlxxyyyxxyrrrjrr=+−−20.5[()()]jkrrxxyyxyyxrrrjrre=−−+2jke−2jke()()1xxyyrrarg(())arg(())xxyyrr−00,(1)xinjixxxxiniZZrreiZZ−===+00,(2)yinjiyyyyiniZZrreiZZ−===+ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
3
Fig. 2. (a) The equivalent transmission-line model of the single-layer
metasurface meta-atom. (b) The specific lumped parameter equivalent
electrical circuit models of proposed deformed square loop metasurface.
(5)
In (4) and (5), Zin
i=2 are the impedances of the equivalent
circuits under corresponding x and y polarized incident. β is the
i=1 and Zin
propagation constant in the dielectric slab. x and y are phase
of the x and y polarized reflection coefficients respectively. The
phase difference is defined as follows:
(6)
To obtain a wideband phase difference , has to satisfy
the following condition for the frequencies ranging from 1 to
2,
(7a)
(7b)
Substituting (4)-(6) into (7), one can derive a set of effective
capacitance and inductance values (L1 = 4.51 nH, L2 = 5.09 nH,
C1 = 0.014 pF, C2 = 0.040 pF) for a frequency range from 9
GHz to 20 GHz. The corresponding results of the equivalent
circuit model are shown in Fig. 3. Within the operating
bandwidth, the phase difference remains to be 180 40 and
the d/df remains to be 15/GHz, which verifies that such a
model can achieve a broadband 180 phase difference between
x and y polarization reflection coefficients. Since no obvious
resonance occurs in the operating bandwidth, the reflection
phase x and y maintain a small change, which is also the key
point for the proposed meta-atom to achieve the wideband and
high efficiency characteristics.
this mainly affects
Combined with the empirical formula and the resonating
nature of the square loop [48] [50], the desired values of
capacitance and inductance could be obtained by adjusting the
geometric parameters of the deformed square loop. As the
value of b1 changes,
the coupling
capacitance C2 between adjacent cells along the y-direction,
which eventually affects the y-direction reflection phase y as
shown in Fig. 3(a). Similarly, the changes of the value of s2
mainly affects the inductance L2, which eventually influences
on the y-direction reflection phase y as shown in Fig. 3(b).
Correspondingly, the reflection phase x can be controlled by
changing the values of s1 and b2. By adjusting the geometric
parameters of the deformed square loop, the reflection phase of
(a)
(b)
Fig. 3. Simulation (Sim) and equivalent circuits (ECs) models are
calculated. Simulation results of x and y as a function of b2 (a) and s2
(b). Other parameters are (a): b2 = 3 mm, s1 = 3.5 mm, s2 = 0.5 mm and p
= 10 mm (b):b1 = 8 mm, b2 = 3 mm, s1 = 3.5 mm and p = 10 mm.
meta-atom obtained from the simulation are consistent with
those from the proposed equivalent circuit model.
Moreover, the geometric parameter of b1 mainly affects the
low-frequency part of reflection phase y as shown in Fig. 3(a),
which determines the low-frequency cutoff frequency (1) and
low-frequency passband performance of the meta-atom. The
geometric parameter of s2 mainly affects the high-frequency
part of reflection phase y as shown in Fig. 3(b), which affects
the high-frequency passband performance of the meta-atom.
Correspondingly, the geometric parameter of b2 also affects the
low-frequency passband performance of the meta-atom. The
geometric parameter of s1 determines the high-frequency cutoff
frequency (2) and high-frequency passband performance of
the meta-atom. These guidelines will be utilized in the design of
the proposed meta-atoms.
III. SIMULATION AND EXPERIMENTAL RESULTS
A. Simulation Results of Unit
According to
the equivalent model of
the proposed
meta-atom and the optimization guidelines in Section II-B, two
meta-atoms are designed to achieve the expected bandwidth by
adjusting the parameters b1 and s2 of the meta-atom in Fig. 3.
One
is for
high-performance
of
ultra-wideband meta-atom are p = 10 mm, b1 = 9.2 mm, s1 = 3.5
mm, b2 = 3 mm, s2 = 1.1 mm, h = 3 mm. The design parameters
of high-performance broadband meta-atom are p = 10 mm, b1 =
is for ultra-wideband (r) and
(r'). The
the other
design
parameters
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
1tan()*()1tan()diiniidiijZhjLjCZjZhjLjC+=++xy=−()120, for all ,dd()()12== This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
4
(a)
(b)
Fig. 4. The reflection spectrum for an ultra-wideband (105%)
meta-atom(r) under LP (a) and CP (b) excitations. In (b), the
corresponding efficiency are also included.
8 mm, s1 = 3.5 mm, b2 = 3 mm, s2 = 0.3 mm, h = 3 mm. Their
numerical simulation results can be obtained by a commercial
software HFSS. As shown in Fig. 4(a), both the reflection
coefficients at x-polarization and y-polarization are close to one
with a nearly phase difference within a wide frequency
bandwidth, which is crucial for an ultra-wideband OAM beam
generation. The working efficiency (efficiency = 2(rxx −
ryy)/22/[rxx2 + rxy2 + ryx2+ryy2] ) is calculated to measure the
conversion performance as illustrated in Fig. 4(b). The
proposed ultra-wideband meta-atom keeps co-polarization
reflection rll higher than 0.9 and efficiency higher than 0.81
within a wide bandwidth ranging from 6.75 to 21.85 GHz. The
achieved 105.6% fractional bandwidth is significantly higher
than the existing 82% fractional bandwidth of the multimode
metasurface [42] under the same efficiency condition. In order
to avoid the obvious resonance frequency around 21 GHz
(which also exists in the mentioned multimode metasurface),
another high-performance broadband meta-atom is designed by
adjusting the parameters b1 and s2. A major phase change of ryy
is introduced which leads to the phase difference between the
x -- polarization (r'xx) and the y -- polarization (r'yy) reflection
coefficients more close to π within 8.55-19.95 GHz as shown in
>=
Fig. 5(a). As a result, a higher co-polarization reflection (r'll
0.96) and higher efficiency (efficiency >= 0.9) can be obtained
as depicted in Fig. 5(b). It is important to observe from Fig. 5(c)
(a)
(b)
(c)
Fig. 5. The reflection spectrum for a high-performance broadband
meta-atom (r') under LP (a) and CP (b) wave excitations. In (a), the
results of ultra-wideband meta-atom(r) are included. In (b), the
corresponding efficiency are also included. (c) The reflection spectrum
for the high-performance broadband meta-atom with different rotation
angles under CP wave excitations.
that the phase responses are paralleled as expected and that the
co-polarization reflection coefficients are all higher than 0.95
within the expected frequency band for different rotation angles,
both of which are essential for the constructed metasurface with
high purity OAM characteristic. In addition, the proposed
broadband meta-atom achieves a fractional bandwidth of 80%,
which is significantly higher than the existing fractional
bandwidth of 40% (12-18 GHz) in high-performance double
arrow-shaped metasurface [43].
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
5
Fig. 6. Layouts of the proposed metasurface with the OAM mode of l =
1, 1.5, 2 and 3.
B. Numerical Near-field Results of Metasurface
As the spin-to-orbital conversion concept discussed above, a
metasurface with a desired vortex phase profile
can be
constructed by arranging the mentioned meta-atoms at different
φ with a certain rotated angle kφ. Here φ is the azimuthal angle
around the vortex beam, and l = 2k is the OAM mode
(topological charge) of the generated vortex beams. As shown
in Fig. 6, the specific topologies of metasurfaces with l = ±1,
±1.5, ±2 and ±3 are composed of a 16∗16 array of the rotated
meta-atoms. To illustrate the wideband behavior of the
proposed metasurfaces, both the near-field and far-field EM
performance has been demonstrated to verify its vortex
property. Under the excitation of a left-handed circular
polarization plane wave, the vortex beams with the OAM mode
of l = −1.5 and l = −3 are generated separately by the proposed
metasurfaces in this paper. At different frequencies, both the
left-handed (LH) and right-handed (RH) components of
reflected electric field are sampled and decomposed in Fig. 7
and Fig. 8. The sampling plane size is 180 mm 180 mm and
the distance between the metasurface and sampling plane is 100
mm. A donut-like amplitude and vortex phase of electric field
can be found from the left-handed component in the sampling
field, which are consistent with the OAM wave characteristics.
For a quantitative analysis of the purity of the OAM modes, the
Fourier transform analysis is implemented to decompose the
individual OAM modes. The corresponding equations are
given as follows [52]
(8)
(9)
Where ψ(φ) is a function of the sampled field along the
circumference of z-axis where the LH component electric field
Fig. 7. Near-field observation and corresponding spectral analyses of
vortex beam with the OAM mode l = −3.
peaks in the sampling plane. Here, the OAM modes from l = −5
to l = 5 are considered and the energy weight of the OAM mode
l is defined as follows.
(10)
The Fourier analysis results in Fig. 7 show that the expected
OAM mode l = −3 is the main part in the LH (co-polarized)
component of the reflected electric field and other OAM modes
in the LH component are small enough to be ignored except for
some possible interference OAM modes l = −3 ± 4n (n Z)
which are introduced by the directional anisotropy of square
lattice. For the RH (cross-polarized) component of the reflected
electric field, the OAM mode l = 0 occupies the dominant part.
The energy weight of RH component primarily depends on the
efficiency of the spin-to-orbital conversion of the proposed
meta-atom. As shown in Fig. 5(b), since a higher conversion
efficiency can be obtained in frequency range from 9 GHz to 14
GHz than frequency range from 14 GHz to 18 GHz, the
generated l = −3 vortex beams possess a higher energy weight
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
jle201()2jllAed−=()jlllAe=55energy weightlllAA=−= This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
6
(a)
(b)
(c)
Fig. 9. Spectral analyses of the generated far-field vortex beams with OAM
mode (a) l = −1.5, (b) l = −3. (c) both far-field directivity and energy weight
of the generated vortex beam with the OAM mode l = −3 at different
sampling polar angle θ.
and φ are the polar and azimuthal angles in the spherical
coordinates. The sampling polar angle is taken at the
maximum point in electric field pattern. For the far-field
spectral analysis of the generated l = −1.5 vortex beam in Fig.
9(a), although the mode l = −1.5 still occupies the most energy
weight, whose energy weight at far field is obviously attenuated
in comparison with the near-field results. Note that the energy
of l = −1.5 mode at far field is converted into adjacent fractional
modes. The notched doughnut patterns become unobvious than
that of the near-field results, which implies that the fractional
order modes are not suitable for long-distance transmission. For
the generated far-field beams with the OAM mode l = −3, a
uniform amplitude and continuous phase of far-field electric
field are obtained at a polar angle = 17 ° in Fig. 9(b).
Consequently, a high-purity vortex spectrum with the energy
Fig. 8. Near-field observation and corresponding spectral analyses of
vortex beam with the OAM mode l = −1.5.
at 11 GHz and 13 GHz (around 70%) than that at 15 GHz and
17 GHz (around 50%).
Fig. 8 shows a Fourier spectrum analysis of the generated l =
−1.5 vortex wave, where l in (8) satisfies l = −1.5 ± n (n Z) to
maintain the orthogonality of the fractional OAM mode for LH
component. Similar results can be obtained in the same way as
the generated l = −3 vortex wave. One significant difference
between the fractional modes and the integer ones is that the
amplitude zeros (phase singularities) of the fractional modes
occur not only in the center but also spread out in one direction
(+y direction in this work). It results in a notched donut shaped
radiation pattern as shown in Fig. 8. In addition, such a
phase-singularity structure also causes a fact that the purity of
the fractional mode (0.35~0.45 energy weight) is lower than the
integer mode (0.48~0.75 energy weight).
C. Numerical and Experimental Far-field Results of
Metasurface
The corresponding Fourier spectral analysis is further
implemented for the generated far-field OAM vortex beams.
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
7
(a)
(b)
(c)
Fig. 10. Far-field spectral analyses of vortex beams generated by
metasurfaces with (a) l = 3, 1616 array, (b) l = 10, 1616 array, and (c)
l = 10, 2020 array. The corresponding left-handed (LH) and
right-handed (RH) components of far-field electric field are also included.
weight of 0.65 can be obtained by (10). It means that the
proposed metasurface can produce high-purity integer-order
vortex waves at both the near and far fields. By observing the
directivity of generated vortex beams in Fig. 9(c), the main lobe
of the pattern become narrower and the number of side lobes
increases as the frequency increases, which have the same
characteristics as the vortex beams generated by the antenna
array [2]. As shown in the corresponding energy weight in Fig.
9(c), the mode purity near the main lobe reaches its highest
value, which is crucial for the reception of high-performance
vortex waves.
A high-order vortex beam with the OAM mode l = -10 is
generated by the metasurface with a 1616 array of the
proposed meta-atoms. Its corresponding far-field spectral
analyses from l = −10 to l = 10 are also shown in Fig. 10(b).
Compared with the results generated by l = −3 metasurface in
Fig. 10(a), the higher order vortex beam suffer more from its
(a)
(b)
(c)
Fig. 11. Comparisons of simulated and measured normalized patterns in
xoz-plane under the excitation of a LHCP spiral antenna. Co-polarized
LHCP (a) and cross-polarized RHCP (b) patterns of the proposed
metasurface with l = −3. (c) The LHCP patterns of the proposed
metasurface with l = −1.5.
divergence angle and uneven amplitude of the main lobe.
Therefore, a wider spectral crosstalk and a lower energy weight
0.53 could be achieved. In principle, it can be understood as the
Nyquist sampling theorem that the metasurface also needs a
certain number of meta-atoms to approximately produce vortex
waves with high-order OAM. Based on the results in Figs. 10(a)
and (b), it is reasonable to observe that the vortex waves with
the high-order modes are not as robust (high-performance) as
when l is small under the same size and structure of metasurface.
Fig. 10(c) shows a larger l = 10 metasurface with a 2020
array of meta-atoms and corresponding spectral analyses. The
amplitude of generated vortex beam shows a more uniform
distribution along φ than the case of metasurface with a 1616
array of meta-atoms. The corresponding spectral analyses
prove that increasing the effective area of metasurface (R2) or
the effective number of meta-atoms can improve the energy
weight of the generated vortex wave (from 0.53 to 0.61 in this
case), which is the key point to produce the high-order and
high-purity vortex waves by the metasurface. Moreover, the
divergence angle of the main lobe is also reduced from = 36
to = 30 by increases the area (R2) of metasurface, as shown in
Fig. 10(b) and Fig. 10(c).
For the experimental verification, both the metasurfaces with
l = 3 and l = 1.5 are fabricated and measured. A wideband
LH Archimedes spiral antenna with VSWR ≤2 and an axial
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
8
generated vortex beam winds by 6 around the z-axis, which
are the typical vortex wave characteristics with the OAM mode
l = −3. The OAM spectra of sampling results are shown in Fig.
12(b). The energy weight of the expected OAM mode l = −3
occupies the dominant part (nearly 39~49% energy weight)
which verifies the proposed metasurface with the OAM mode l
= 3. Notice that the achieved performance is also determined
by the excitation antenna and the alignment of the sampling
plane.
IV. CONCLUSION
In this work, a single-layer broadband meta-atom has been
proposed base on the deformed square loop structure. With
such a meta-atom, a simple and high-performance metasurface
has been synthesized to generate OAM vortex beams within a
wideband range from 8.55 GHz to 19.95 GHz. Moreover, the
corresponding equivalent circuit model has been established
and the proposed broadband meta-atom can be analyzed and
designed effectively. The proposed meta-atoms can be further
applied to other phase-based device due to its high-efficiency
and broadband features, and this broadband model can also be
extended to the transmission structures. Moreover, the vortex
beams with the fractional and high-order OAM modes have
also been generated and analyzed, which may find further
applications in target detection and communication systems.
The OAM spectral analyses of both near-field and far-field
have been performed, which illustrates the high-purity vortex
characteristics of the proposed metasurface more intuitively.
REFERENCES
[1] L. Allen, M. W. Beijersbergen, R. J. Spreeuw, and J. P. Woerdman,
"Orbital angular momentum of light and the transformation of Laguerre
Gaussian laser modes," Phys. Rev. A , vol. 45, no. 11, pp. 8185-8189, 1992.
[2] S. M. Mohammadi, L. K. S. Daldorff, J. E. S. Bergman, R. L. Karlsson, B
Thidé, K. Forozesh, T. D. carozzi, and B. Isham, "Orbital angular
momentum in radio a system study," IEEE Trans. Antennas Propag., vol.
58, no. 2, pp. 565-572, 2010.
[3] J. Wang, J. Y. Yang, I. M. Fazal, N. Ahmed, Y. Yan, H. Huang, Y. X. Ren,
S. Dolinar, M. Tur, "Terabit free-space data transmission employing
orbital angular momentum multiplexing," Nat. Photonics, vol. 6, no. 7, pp.
488496, 2012.
[4] Y. Yan, G. Xie, M. P. Lavery, H. Huang, N. Ahmed, C. Bao, Y. Ren, Y.
Cao, L. Li, Z. Zhao, A. F. Molisch, M. Tur, M. J. Padgett, and A. E.
Willner, "High-capacity millimetre-wave communications with orbital
angular momentum multiplexing," Nat. Commun., vol. 5, p. 4876, 2014.
[5] F. Tamburini, E. Mari, A. Sponselli, B. Thidé, A. Bianchini, and F.
Romanato, "Encoding many channels in the same frequency through radio
vorticity: first experimental test," New J. Phys., vol. 14, no. 11, pp.
78001-78004, 2012.
[6] X. Hui, S. Zheng, Y. Chen, Y. Hu, X. Jin, H. Chi, and X. Zhang,
"Multiplexed millimeter wave communication with dual orbital angular
momentum (OAM) mode antennas," Sci. Rep., vol. 5, p. 10148, May. 19,
2015.
[7] J. Wang, J. Y. Yang, I. M. Fazal, N. Ahmed, Y. Yan, H. Huang, Y. X. Ren,
Y. Yue, S. Dolinar, M. Tur, and A. E. Willner,, "Terabit free-space data
transmission employing orbital angular momentum multiplexing," Nat.
Photonics, vol. 6, no. 7, pp. 488496, 2012.
[8] G. Gibson, J. Courtial, and M. J. Padgett, "Free-space information transfer
using light beams carrying orbital angular momentum,", Opt. Express, vol.
12, no. 22, pp. 5448-5456, Nov. 1, 2004.
(a)
(b)
Fig. 12. The near-field sampling results under the excitation of a LHCP
spiral antenna. (a) The amplitude and phase distributions of reflection
vortex beam with OAM mode l = −3. (b) The spectral analyses of sampling
results. The photograph of corresponding experimental system is also
included.
ratio less than 4 dB within 7-18 GHz is used for the excitation.
Moreover, the spiral antenna can benefit from its small
cross-sectional radius of 12 mm to ensure negligible influence
on the reflected vortex beams. The transmitting antenna and
metasurface are fixed on a foam box at a distance of 100 mm as
shown in Fig. 1(a). Both the measured and simulated patterns
are normalized to the co-polarized (LHCP) pattern. As shown
in Fig. 11, the measured patterns of the modes l = −3 and l =
−1.5 are consistent with the simulated patterns varied from 9
GHz to 18 GHz. The cross-polarization (RHCP) patterns of l =
−3 are significantly lower than the co-polarization (LHCP)
patterns, except for the high-frequency patterns at 16 GHz and
18 GHz, where the conversion efficiency of metasurface is
small. The hollow patterns suggest that both the l = −3 and l =
−1.5 reflection beams carry the OAM characteristics. One
difference is that the patterns of the l = −1.5 vortex wave
becomes less of symmetric in comparison with the l = −3 case.
D. Numerical and Experimental Near-field Results of
Metasurface
The sampling plane size is 400 mm 400 mm and the
distance between the metasurface and the sampling plane is 220
mm. The spiral antenna is placed between the metasurface and
the sampling plane to excite the circularly-polarized plane
waves for the metasurface with OAM l = 3. Hollow vortex
beams can effectively avoid the excitation antenna placed on
the z-axis. The results of the low frequency parts are shown in
Fig. 12(a). The amplitude and phase distributions of
measurement reflection vortex beam are good agreement with
the simulation results. The expected null appears in the central
region of the electric field magnitude and the phase front of the
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
9
[9] F. Feng, X. Jin, D. O'Brien, F. P. Payne and T. D. Wilkinson, "Mode-group
multiplexed transmission using OAM modes over 1 km ring-core fiber
without MIMO processing," 2017 Optical Fiber Communications
Conference and Exhibition (OFC), Los Angeles, CA, 2017.
[30] P. Genevet, N. F. Yu, F. Aieta, J. Lin, M. A. Kats, R. Blanchard, M. O.
Scully, Z. Gaburro, and F. Capasso, "Ultra-thin plasmonic optical vortex
plate based on phase discontinuities," Appl. Phys. Lett., vol. 100, no. 1, p.
169, Jan. 2, 2012.
[10] L. Zhu, J. Li, G. Zhu, L. Wang, C. Cai, A. Wang, S. Li, M. Tang, Z. He, and
S. Yu, "First demonstration of orbital angular momentum (OAM)
distributed Raman amplifier over 18-km OAM fiber with data-carrying
OAM multiplexing and wavelength-division multiplexing," 2018 Optical
Fiber Communications Conference and Exposition (OFC), San Diego, CA,
2018.
[11] A. Mair, A. Vaziri, G. Weihs, and A. Zeilinger, "Entanglement of the
orbital angular momentum states of photons," Nature, vol. 412, no. 6844,
pp. 313-316, 2001.
[12] A. F. Morabito, L. Di Donato, and T. Isernia, "Orbital angular momentum
antennas: understanding actual possibilities through the aperture antennas
theory," IEEE Antennas Propagat. Mag., vol. 60, n. 2, pp. 59-67, 2018.
[13] O. Edfors and A. J. Johansson, "Is orbital angular momentum (OAM)
based radio communication an unexploited area?," IEEE Trans. Antennas
Propag., vol. 60, no. 2, pp. 1126-1131, Feb. 2012.
[14] M. Barbuto, M.-A. Miri, A. Alu, F. Bilotti, and A. Toscano, "Exploiting
the topological robustness of composite vortices in radiation systems,"
Prog. Electromagn. Res., vol. 162, pp. 39-50, 2018.
[15] M. Lin, Y. Gao, P. Liu, and J. Liu, "Super-resolution orbital angular
momentum based radar targets detection," Electron. Lett., vol. 52, no. 13,
pp. 1168-1170, Jun. 23 2016.
[16] K. Liu, Y. Cheng, Y. Gao, X. Li, Y. Qin, and H. Wang, "Super-resolution
radar imaging based on experimental OAM beams," Appl. Phys. Lett., vol.
110, no. 16, 2017.
[17] L. Chen, J. Lei, and J. Romero, "Quantum digital spiral imaging," Light Sci.
Appl., vol. 3, no. 3, p. e153, 2014.
[18] M. Barbuto, F. Bilotti, and A. Toscano, "Patch antenna generating
structured fields with a Möbius polarization state," IEEE Antennas Wirel.
Propag. Lett., vol. 16, pp.1345-1348, 2017.
[19] C. Zhang and D. Chen, "Large-Scale orbital angular momentum radar
pulse generation with rotational antenna," IEEE Antennas Wirel. Propag.
Lett., vol. 16, pp. 2316-2319, 2017.
[20] D. G. Grier and Y. Han, "Erratum: Vortex rings in a constant electric field,"
Nature, vol. 424, 2003.
[21] N. Uribe-Patarroyo, A. Alvarez-Herrero, A. L. Ariste, A. A. Ramos, T.
Belenguer, R. M. Sainz, C. Lemen, and B. Gelly, "Detecting photons with
orbital angular momentum in extended astronomical objects: application to
solar observations," Astron. Astrophys., vol. 526, no. 2, pp. 202-207, 2010.
[22] X. N. Hui, S. L. Zheng, Y. P. Hu, C. Xu, X. F. Jin, H. Chi, and X. M. Zhang,
"Ultralow reflectivity spiral phase plate for generation of millimeter-wave
OAM beam," IEEE Antennas Wirel. Propag. Lett., vol. 14, pp. 966-969,
2015.
[23] M. Uchida and A. Tonomura, "Generation of electron beams carrying
orbital angular momentum," Nature, vol. 464, no. 7289, pp. 737-739, Apr.
1, 2010.
[24] P. Genevet, J. Lin, M. A. Kats, and F. Capasso, "Holographic detection of
the orbital angular momentum of light with plasmonic photodiodes," Nat.
Commun., vol. 3, no. 4, p. 1278, 2012.
[25] S. M. Mohammadi, L. K. S. Daldorff, K. Forozesh, B. Thide, J. E. S.
Bergman, B. Isham, R. Karlsson, and T. D. Carozzi, "Orbital angular
momentum in radio: Measurement methods," Radio Sci., vol. 45, no. 4, Jul.
16, 2010.
[26] B. Thidé, H. Then, J. Sjoholm, K. Palmer, J. Bergman, T. D. Carozzi, Y. N.
Istomin, N. H. Ibragimov, and R. Khamitova, "Utilization of photon orbital
angular momentum in the low-frequency radio domain," Phys. Rev. Lett.,
vol. 99, no. 8, Aug. 24, 2007.
[27] K. Liu, H. Liu, Y. Qin, Y. Cheng, S. Wang, L. Xiang, and H. Q. Wang,
"Generation of OAM beams using phased array in the microwave band,"
IEEE Trans. Antennas Propag., vol. 64, no. 9, pp. 3850-3857, Sept. 2016.
[28] M. V. Berry, "The Adiabatic phase and Pancharatnam's phase for
polarized light," Opt. Acta Intern. J. Opt., vol. 34, no. 11, pp. 14011407,
1987.
[29] N. Yu, P. Genevet, M. A. Kats, F. Aieta, J. P. Tetienne, F. Capasso, and Z.
Gaburro, "Light propagation with phase discontinuities: generalized laws
of reflection and refraction," Science, vol. 334, no. 6054, pp. 333337, Oct.
21, 2011.
[31] S. X. Yu, L. Li, and N. Kou, "Generation, reception and separation of
mixed-state orbital angular momentum vortex beams using metasurfaces,"
Opt. Mater Express, vol. 7, no. 9, pp. 3312-3321, Sep. 1, 2017.
[32] F. Bouchard, I. De Leon, S. A. Schulz, J. Upham, E. Karimi, and R. W.
Boyd, "Optical spin-to-orbital angular momentum conversion in ultra-thin
metasurfaces with arbitrary topological charges," Appl. Phys. Lett., vol.
105, no. 10, pp. 36-262, 2014.
[33] E. Karimi, S. A. Schulz, I. De Leon, H. Qassim, J. Upham, and R. W. Boyd,
"Generating optical orbital angular momentum at visible wavelengths
using a plasmonic metasurface," Light-Sci. Appl., vol. 3, no. 5, p. e167,
May. 2014.
[34] L. Marrucci, C. Manzo, and D. Paparo, "Optical spin-to-orbital angular
momentum conversion in inhomogeneous anisotropic media," Phys. Rev.
Lett., vol. 96, no. 16, p. 163905, 2006.
[35] N. Yu and F. Capasso, "Flat optics with designer metasurfaces," Nat.
Mater., vol. 13, no. 2, pp. 139-50, 2014.
[36] S. Jiang, C. Chen, H. Zhang, and W. Chen, "Achromatic electromagnetic
metasurface for generating a vortex wave with orbital angular momentum
(OAM)," Opt. Express, vol. 26, no. 5, pp. 6466-6477, Mar. 5, 2018.
[37] B. Ratni, D. Zhang, K. Zhang, K. Tang, M. Lu, Q. Wu, S. N. Burokur, X.
Ding, and Y. Yuan, "Phase-engineered metalenses to generate converging
and non-diffractive vortex beam carrying orbital angular momentum in
microwave region," Opt. Express, vol. 26, no. 2, p. 1351, 2018.
[38] M. Pu, Z. Zhao, Y. Wang, X. Li, X. Ma, C. Hu, C. Wang, C. Huang, and X.
Luo, "Spatially and spectrally engineered spin-orbit interaction for
achromatic virtual shaping," Sci. Rep., vol. 5, p. 9822, 2014.
[39] K. Zhang, X. Ding, L. Zhang, and Q. Wu, "Anomalous threedimensional
refraction in the microwave region by ultra-thin high efficiency metalens
with phase discontinuities in orthogonal directions," New J. Phys., vol. 16,
no. 10, p. 103020, 2014.
[40] M. L. N. Chen, L. J. Jiang, and W. E. I. Sha, "Artificial perfect electric
conductor-perfect magnetic conductor anisotropic metasurface for
generating orbital angular momentum of microwave with nearly perfect
conversion efficiency," J. Appl. Phys., vol. 119, no. 6, 2016.
[41] G. M. Wang, H. X. Xu, J. Xiao, T. Cai, and Y. Q. Zhuang, "Tunable
PancharatnamBerry metasurface for dynamical and high-efficiency
anomalous reflection" Opt. Express, vol. 24, no. 24, pp. 27836-27848,
2016.
[42] H. X. Xu, H. Liu, X. Ling, Y. Sun, and F. Yuan, "Broadband vortex beam
generation using multimode Pancharatnam-Berry metasurface," IEEE
Trans. Antennas Propag., vol. 65, no. 12, pp. 7378-7382, Dec. 2017.
[43] Y. Ran, J. Liang, T. Cai, and H. Li, "High-performance broadband vortex
beam generator using reflective Pancharatnam-Berry metasurface," Opt.
Commun., vol. 427, pp. 101-106, 2018.
[44] Y. Yao, X. Liang, M. Zhu, W. Zhu, J. Geng, and R. Jin, "Analysis and
experiments on reflection and refraction of orbital angular momentum
waves," IEEE Trans. Antennas Propag., vol. 67, no. 4, pp. 2085-2094,
April 2019.
[45] C. Menzel, C. Rockstuhl, and F. Lederer, "An advanced Jones calculus for
the classification of periodic metamaterials," Phys. Rev. A, vol. 82, no. 5,
pp. 3464-3467, 2010.
[46] W. J. Luo, S. L. Sun, H. X. Xu, Q. He, and L. Zhou, "Transmissive
ultrathin Pancharatnam-Berry metasurfaces with nearly 100% efficiency,"
Phys. Rev. Appl., vol. 7, no. 4, Apr. 28, 2017.
[47] O. Luukkonen, C. Simovski, G. Granet, G. Goussetis, D. Lioubtchenko, A.
V. Raisanen, and S. A. Tretyakov, "Simple and accurate analytical model
of planar grids and high-impedance surfaces comprising metal strips or
patches," IEEE Trans. Antennas Propag., vol. 56, no. 6, pp. 1624-1632,
2008.
[48] M. A. R. Barrera and W. P. Carpes, "Bandwidth for the equivalent circuit
model in square-loop frequency selective surfaces" IEEE Trans. Antennas
Propag., vol. 65, no. 11, pp. 5932-5939, Nov. 2017.
[49] D. Ferreira, R. F. S. Caldeirinha, I. Cuias, and T. R. Fernandes, "Square
loop and slot frequency selective surfaces study for equivalent circuit
model optimization," IEEE Trans Antennas Propag., vol. 63, no. 9, pp.
3947-3955, 2015.
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2948711, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
10
[50] R. J. Langley and E. A. Parker, "Equivalent circuit model for arrays of
square loops," Electron. Lett., vol. 18, no. 7, pp. 294-296, 1982.
[51] K. Miyaguchi, M. Hieda, K. Nakahara, H. Kurusu, M. Nii, M. Kasahara, T.
Takagi, S. Urasaki, "An ultra-broad-band reflection-type phase-shifter
MMIC with series and parallel LC circuits" , IEEE Trans. Microw. Theory
Techn., vol. 49, no. 12, pp. 2446-2452, Dec. 2001.
[52] B. Jack, M. J. Padgett, and S. Franke-Arnold, "Angular diffraction" New J.
Phys., vol. 10, no. 10, pp. 6456-6460, 2008.
Ling-Jun Yang was born in Guilin,
Guangxi, China. He received the B.Eng.
degree in electrical engineering from
Xidian University, Xi'an, China, in 2017.
He is currently pursuing the Ph.D degree
in electromagnetic
radio
technology
the University of
Electronic Science and Technology of
China, Chengdu, China.
field and
from
His current research interests include electromagnetic vortex
beam with orbital angular momentum, spiral antennas, and
metasurface.
in
from Xi'an
Sheng Sun (S'02 -- M'07 -- SM'12) received
the B.Eng. degree
information
engineering
Jiaotong
University, Xi'an, China, in 2001, and the
Ph.D. degree in electrical and electronic
engineering from Nanyang Technological
University (NTU), Singapore, in 2006.
From 2005 to 2006, he was with the
Institute of Microelectronics, Singapore.
From 2006 to 2008, he was a Post-Doctoral Research Fellow
with NTU. From 2008 to 2010, he was a Humboldt Research
Fellow with the Institute of Microwave Techniques, University
of Ulm, Ulm, Germany. From 2010 to 2015, he was a Research
Assistant Professor with The University of Hong Kong, Hong
Kong. Since 2015, he has been a Full Professor with the
University of Electronic Science and Technology of China,
Chengdu, China. He has authored or co-authored 1 book and 2
journal and conference
book chapters, and over 160
publications. His
include
electromagnetic
computational mathematics,
multiphysics, numerical modeling of planar circuits and
antennas, microwave passive and active devices, and the
microwave- and millimeter-wave communication systems.
interests
theory,
current
research
an Associate Editor for IEEE MICROWAVE AND WIRELESS
COMPONENTS LETTERS.
Wei E. I. Sha (M'09-SM'17) Wei E.I.
Sha received the B.S. and Ph.D. degrees
in Electronic Engineering at Anhui
University, Hefei, China, in 2003 and
2008, respectively. From Jul. 2008 to
Jul. 2017, he was a Postdoctoral
Research Fellow and then a Research
Assistant Professor in the Department of
Electrical and Electronic Engineering at
the University of Hong Kong, Hong Kong. From Oct. 2017, he
joined the College of Information Science & Electronic
Engineering at Zhejiang University, Hangzhou, China, where
he is currently a tenure-tracked Assistant Professor. From Mar.
2018 to Mar. 2019, he worked at University College London as
a Marie Skłodowska-Curie Individual Fellow. His research
interests include theoretical and computational research in
electromagnetics and optics, focusing on the multiphysics and
interdisciplinary research. His research involves fundamental
and applied aspects
in multiphysical electromagnetics,
topological electromagnetics, nonlinear electromagnetics,
quantum
computational
electromagnetics. Dr. Sha has authored or coauthored 105
refereed journal papers, 106 conference publications (including
27 invited talks), four book chapters, and two books. His
Google Scholar citation is 4346 with h-index of 29. He is a
senior member of IEEE and a member of OSA. He served as
Reviewers for 50 technical journals and Technical Program
Committee Members of 9 IEEE conferences. He also served as
an Editorial Board Member of Progress In Electromagnetics
Research and Guest Editors of IEEE Journal on Multiscale and
Multiphysics Computational Techniques and The Applied
Computational Electromagnetics Society Journal. In 2015, he
was awarded Second Prize of Science and Technology from
Anhui Province Government, China. In 2007, he was awarded
the Thousand Talents Program for Distinguished Young
Scholars of China. Dr. Sha also received four Best Student
Paper Prizes and one Young Scientist Award with his students.
electromagnetics,
and
in 2004,
Dr. Sun was a recipient of the ISAP Young Scientist Travel
Grant, Japan,
the Hildegard Maier Research
Fellowship of the Alexander Von Humboldt Foundation,
Germany, in 2008, the Outstanding Reviewer Award of the
IEEE MICROWAVE AND WIRELESS COMPONENTS LETTERS in
2010, and the General Assembly Young Scientists Award from
the International Union of Radio Science in 2014. He was a
co-recipient of the several Best Student Paper Awards of
international conferences. He was an Associate Editor of the
IEICE Transactions on Electronics from 2010 to 2014 and a
the Applied Computational
Guest Associate Editor of
Electromagnetics Society Journal in 2017. Dr. Sun is currently
a member of the Editor Board of the International Journal of
RF and Microwave Computer Aided Engineering and serves as
0018-926X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
|
1908.01283 | 2 | 1908 | 2019-10-01T02:50:22 | Sub-millimetric ultra-low-field MRI detected in situ by a dressed atomic magnetometer | [
"physics.app-ph",
"physics.atom-ph",
"physics.ins-det"
] | Magnetic Resonance Imaging (MRI) is universally acknowledged as an excellent tool to extract detailed spatial information with minimally invasive measurements. Efforts toward ultra-low-field (ULF) MRI are made to simplify the scanners and to reduce artefacts and incompatibilities. Optical Atomic Magnetometers (OAMs) are among the sensitive magnetic detectors eligible for ULF operation, however they are not compatible with the strong field gradients used in MRI. We show that a magnetic-dressing technique restores the OAMs operability despite the gradient, and we demonstrate sub-millimetric resolution MRI with a compact experimental setup based on an in situ detection. The proof-of-concept experiment produces unidimensional imaging of remotely magnetized samples with a dual sensor, but the approach is suited to be adapted for 3-D imaging of samples magnetized in loco. An extension to multi-sensor architectures is also possible. | physics.app-ph | physics | Sub-millimetric ultra-low-field MRI
detected in situ by a dressed atomic magnetometer
Giuseppe Bevilacqua,1 Valerio Biancalana,1, a) Yordanka Dancheva,2, b) and Antonio Vigilante2
1)DIISM - University of Siena
Via Roma 56 Siena Italy
2)DSFTA - University of Siena
Via Roma 56 Siena Italy
(Dated: 2 October 2019)
Magnetic Resonance Imaging (MRI) is universally acknowledged as an excellent tool to extract detailed spatial infor-
mation with minimally invasive measurements. Efforts toward ultra-low-field (ULF) MRI are made to simplify the
scanners and to reduce artefacts and incompatibilities. Optical Atomic Magnetometers (OAMs) are among the sen-
sitive magnetic detectors eligible for ULF operation, however they are not compatible with the strong field gradients
used in MRI. We show that a magnetic-dressing technique restores the OAMs operability despite the gradient, and
we demonstrate sub-millimetric resolution MRI with a compact experimental setup based on an in situ detection. The
proof-of-concept experiment produces unidimensional imaging of remotely magnetized samples with a dual sensor,
but the approach is suited to be adapted for 3-D imaging of samples magnetized in loco. An extension to multi-sensor
architectures is also possible.
Isidor Rabi was awarded the Nobel Prize in Physics in
1944 for his seminal research, which, in 1938, demonstrated
the phenomenon of nuclear magnetic resonance (NMR) in
a molecular beam1. Felix Bloch and Edward Purcell were
awarded the Noble Prize for their independent contributions to
NMR (dated 1946)2,3 in 1952, the same year in which Robert
Gabillard, in his PhD thesis, studied the NMR in the presence
of magnetic field gradients (a crucial step in view of encoding
spatial information in the precessing nuclei). It took two ad-
ditional decades to realize that the potential of NMR to record
spatial distribution of precessing nuclei could be exploited as a
medical diagnostic tool, as proposed by Raymond Damadian4.
The applicability of that idea was then demonstrated by Peter
Mansfield and Paul Lauterbur5, which thirty years after were
awarded the Nobel Prize.
The attractiveness of MRI in medicine relies on its accu-
rateness and on its substantially non-invasive nature. The
latter feature is shared with ultrasonography6, whose de-
velopment occurred almost simultaneously with MRI. Both
methodologies are spreadly used and constitute favourite
choices with respect to more invasive imaging techniques
based on ionizing radiation.
The great impulse impressed to the development of the
MRI technology, led to fast and impressive progresses in the
methodologies used to generate, detect, and analyze MRI sig-
nals. These advances were facilitated by the parallel pro-
gresses achieved in some related technologies, such as elec-
tronics, computer science and cryogenics.
Most of MRI and, more generally, NMR advances followed
the straightforward direction of enhancing strength and homo-
geneity of the magnetic field (or accurate control of its gradi-
ents) as well as increasing the signal-to-noise ratio (SNR) of
a)Electronic mail: [email protected]
b)currently at: Aerospazio Tecnologie S.r.l., Via Provinciale Nord, 42a Rap-
olano, Siena (Italy)
the detection stage. Cryogenics and superconductor technolo-
gies constituted an obvious opportunity for this evolution.
At the same time, cryogenics -- allowing the development of
innovative detectors (superconducting quantum interference
devices, SQUIDs) with unrivaled sensitivity -- made available
alternative (non inductive) sensors, and opened the perspec-
tive of performing NMR and MRI at much lower precession
frequencies, that is at low and ultra-low field strengths7.
The ULF-NMR dates about three decades, similarly to
ULF-MRI8, for which intense progresses started, however,
less than two decades ago9,10.
MRI in the ULF regime comes with several valuable
advantages7. The ultimate spatial resolution of MRI is de-
termined by the NMR spectral resolution, that depends on the
absolute field inhomogeneity. At ULF, a modest relative field
homogeneity turns out to be excellent on an absolute scale:
very narrow NMR lines with a high SNR can be recorded
in ULF regime, using relatively simple and unexpensive field
generators11 -- 13. The encoding gradients necessary for MRI
can be generated by simple and low-power coil systems,
as well9,14. Further important advantages brougth to MRI
by ULF regime include the minimization of susceptibility15
and conductance12,16 artefacts. Other delicate instrumenta-
tion (not compatible with strong and/or fast-varying mag-
netic fields) can be used in conjunction with ULF-MRI in
more complex setups. In addition, the non-conventional mag-
netic detectors used in ULF-MRI can be used to record low-
frequency magnetic signals originating not only from the
precessing nuclei, but also from other (e.g. biomagnetic)
sources. Hybrid instrumentation enabling multimodal MRI
and magneto-encephalo-graphic measurements in medical ap-
plications has been demonstrated17,18 (see also Chap.5 in
Ref.7).
While in conventional NMR and MRI, the premagnetiza-
tion and precession are typically induced by one field, the two
functions are often distinguished in ULF apparatuses. Here,
the precession field can be extremely weak, and the homo-
geneity of the (strong) premagnetization field is not a critical
9
1
0
2
t
c
O
1
]
h
p
-
p
p
a
.
s
c
i
s
y
h
p
[
2
v
3
8
2
1
0
.
8
0
9
1
:
v
i
X
r
a
parameter. It is worth mentioning that schemes of no-magnet
NMR, with zero precession field and alternative premagneti-
zation methods have been proposed19.
Direct coupling of SQUID sensors with strong premagneti-
zation fields in not feasible. Successful attempts to overcome
this problem are based on using flux dams20 or finely designed
SQUID coils, with gradiometric sensitivity and very high re-
jection of the common mode term18, while another possible
approach is based on using OAMs as alternative highly sen-
sitive non-inductive detectors21. Beside robustness, OAMs
bring the advantage of not requiring cryogenics, so to be a
favourite choice whenever ULF systems are designed in view
of building up simpler and low-cost apparatuses. Despite their
simplicity, OAMs -- in some implementations -- may compete
with SQUIDs in terms of sensitivity. In facts, the literature
reports successful ULF-MRI experiments using both SQUIDs
and OAMs as highly sensitive, non-inductive sensors22,23.
OAMs operate on the basis of paramagnetic atoms in which
an atomic magnetic resonance (AMR) is induced using reso-
nant light as a polarization tool (modern laser spectroscopy
methodologies provide very effective instrumentation to this
end)24. The sensitivity of OAMs relies on the narrow spec-
tral width of the AMR. The important field gradients neces-
sary for MRI applications would broaden severely the AMR.
As matter of fact, the ULF-MRI experiments reported so far
with OAM detection are based on ex-situ measurements: the
magnetic signal produced by the precessing nuclei is coupled
to the sensor via flux transformers25, eventually resulting in
remote-detection techniques26.
In this letter we demonstrate that an approach based on an
inhomogeneous magnetic dressing of the precessing atoms27
can be used to record in situ MRI signals by means of OAM,
achieving sub-millimetric resolution. The described proof-of-
concept experiment makes use of a dual sensor, but paves the
way to multi-sensor detection, with the potential of improving
the spatial resolution, enhancing the allowed sample size, and
speeding up the acquisition.
The described setup performs MRI of samples premagne-
tized in strong field and subsequently transferred to the detec-
tion region, however the robustness of OAMs to strong mag-
netic fields would enable MRI of samples premagnetized in
the same position where the NMR signal is detected, so to
obtain a fully static operation.
A synthetic description of the apparatus is reported here be-
low, referring to pertinent bibliography for its specific parts.
The core of the setup (see Fig.1) is a dual channel OAM work-
ing in a Bell and Bloom configuration28.
The strength and the first-order gradient of the field in
which the OAM operates are numerically controlled and
optimized. Namely, eight numerically controlled current
sources29 supply the field and field-gradient coils. Automated
procedures enable the nulling of the gradient terms and guar-
antee the field alignment along a fixed direction.
The dual sensor detector produces two magnetometric sig-
nals which contain the measurement of field variations due to
both far-located and close-located sources. The first contribu-
tion is dominant and appears with the same sign on the two
sensors. The second term appears with opposite signs, pro-
2
FIG. 1. The laser beams (in red) propagate along the x direction and
cross parallelly two atomic cells (in blue). A static magnetic field is
oriented along z and varies with the position x to the purpose of per-
forming frequency encoding. Electromagnets (in yellow) produce a
strong dressing field oscillating along x much faster than the atomic
Larmor frequency. The dressing field has a position-dependent in-
tensity, such to restore the AMR width. The NMR sample cartridge
(in dark and light grey) is represented sectioned to show the internal
structure. Both the cartridge and the cells are merged in the inho-
mogeneous field required for the MRI frequency encoding (typical
values are Bz = 4µT and ∂ Bz/∂ x = 40 nT/cm), and the magnetome-
ter sensitivity is restored by the inhomogeneous dressing.
vided that the close-located source is opportunely displaced
with respect to the sensors.
The two signals are recombined to extract their common
mode (CM) term and difference-mode (DM) term: far-located
sources (which in our measurement constitute a disturbance)
contribute to the CM only, while the MRI signal appears in
the DM one.
Compensating the CM term has a twofold advantage: (i) the
disturbances and drifts of the field which would affect the nu-
clear precession are removed; (ii) unavoidable imperfections
of the differential system (a limited CM-rejection ratio) let the
CM appear residually with the DM term, thus a preliminary
reduction of the CM term improves the DM signal-to-noise
ratio.
To this aim, while extracting the MRI signal from the DM
term, the CM term is used to feed a self-optimized closed loop
system30 to actively compensate the external disturbances,
which are strongly present due to the unshielded nature of the
set-up.
As represented in Fig.1, both the Cs atoms and the sam-
ple protons precess around a static (stabilized) field Bz ori-
ented transversely with respect to the beam propagation axis
(x). The field Bz is made dependent on the position x to the
purpose of performing MRI frequency-encoding.
Its gradi-
ent G = ∂ Bz/∂ x is set by permanent magnets arranged in a
quadrupolar configuration: Bz = B0 +Gx, where B0 is the field
at the center of the cell. The proton and Cs Larmor frequen-
cies set by Bz are thus position dependent along the optical
axis x. Typical values of G amount to tens of nT/cm, which
would broaden the AMR width from few tens up to several
hundred Hz, degrading and eventually destroying the OAM
operativity.
Based on the IDEA method described in Ref.27, an inho-
mogeneous magnetic dressing technique is applied to recover
the OAM operativity. The presence of a dressing field, makes
the time evolution of the atomic magnetization (cid:126)M more com-
plicated than a simple precession around the static field Bz. In
particular, a strong BD field oscillating along x at a frequency
f much larger than γCsBz (let γCs = 2π3.5 Hz/nT be the gy-
romagnetic factor of Cs ground state ) makes the My and Mz
components follow a deeply modified trajectory on the Bloch
sphere.
In contrast, the Mx component -- the polarimetrically mea-
sured quantity that provides the magnetometer output -- keeps
oscillating harmonically31. The effect of BD on the Mx evo-
lution is just a reduction of its oscillation frequency. The re-
duced frequency νD depends on the strength of the dressing
field BD and on f , according to
νD = J0
ν0,
(1)
where Ji is the ith order Bessel function of the first kind
and ν0 = γCsBz/2π is the precession frequency in absence of
dressing field32.
An inhomogeneous dressing can compensate for the detri-
mental effects of the field gradient used for MRI frequency
encoding. As described in Ref.27, if the strength of BD has an
appropriate dependence on x, the dressing can compensate the
position-dependent AMR frequency shift caused by the gradi-
ent of Bz and the AMR width can be restored.
To this end, each sensor is coupled to an electromagnetic
BD source (a coil wound on a hollow-cylinder ferrite, with the
laser beams passing across the hole).
In a dipole approximation, at a distance x from the center
of the cell, the dressing field BD is
BD(x,t) =
µ0
2π
m(t)
(x0 + x)3 = BD0(x)cos(2π f t),
(2)
where µ0 is the vacuum permittivity, m(t) = m0 cos(2π f t) is
the oscillating dipole momentum, and x0 is the distance of the
dipole from the cell center.
Taking into account the dependence on x of both the static
and the dressing fields, Mx oscillates harmonically at a fre-
quency
(cid:18)γCsBD0(x)
(cid:19)
2π f
(cid:18)γCsBD
(cid:19)
2π f
,
(3)
(cid:21)
(cid:19)
νD(x) =
γCs
2π (B0 + Gx)J0
in a first-order Taylor approximation,
D(0)x + O(x2)
νD(x) = νD(0) + ν(cid:48)
(cid:18)
(cid:20)3B0αJ1(α)
≈ γCs
2π
B0J0(α) +
where α = (µ0/4π2)(γCsm0)/( f x3
compensating the effect of the gradient G reduces to
+ GJ0(α)
x
,
x0
0), so that the condition for
− 3
B0
x0
αJ1(α)
J0(α)
= G.
(4)
3
Under this condition, the OAM performance is recovered,
so to guarantee the sensitivity necessary to detect MRI sig-
nals. It is worth noting that the dressing effect is negligible
for protons, because of their much smaller gyromagnetic fac-
tor: γH (cid:28) γCs, which makes J0(γHBD/2π f ) ≈ 1.
Differing from the case studied in Ref.27, here two dis-
tinct Cs cells are used, each of them equipped with a dress-
ing dipole: an arrangement that increases the NMR signal im-
proving the MRI performance. With this feature, the present
results demonstrate that a multiple-sensor arrangement can be
built, with one oscillating IDEA dipole for each sensor. It is
possible to maintain cross-talking between dressed sensors at
a negligible level.
FIG. 2. The panel (a) shows a photograph of the open cartridge with
its internal structure. In the panel (b) an uni-dimensional image is
shown as reconstructed from an average trace corresponding to tight
(0.08 mm interval) positioning of the sample. Both (a) and (b) are
in the same scale, blue lines associate MRI peaks to their origin in
cartridge volume.
The NMR sample is made of water protons contained in
a polymeric cartridge having the structured shape shown in
Figs.1 and 2(a): it is a cylinder -- 19 mm in diameter, 32 mm
in length -- that contains three disks -- 2 mm in thickness -- sep-
arated by 5 mm from each other, the water (in hydrogel) is
confined in the four complementary disks. Care is taken to
avoid ferromagnetic contamination of the container33.
The setup includes an Halbach permanent-magnet assem-
bly to premagnetize the sample at 1 T and a pneumatic shuttle
system34 to move it cyclically to the measurement region (see
Refs.35,36 for additional details).
The cartridge position along x slightly changes shot-by-
shot. At each measurement, a camera monitors such sample
positioning with respect to the sensors. An automated image
analysis provides a localization xC (with respect to a fixed ori-
4
The histogram (b) shows the distribution of xC over a large set of measurements. Five subsets of measurements corresponding
FIG. 3.
to narrow intervals of xC have been selected to reconstruct the MRI traces shown in the panel (a). To this end, the selected time traces are
averaged and analysed by FFT: the real part of the Fourier transform is shown. The inclined blue line guides the eyes to localize the leftmost
peak position of each image as it varies along the histogram (3.5 mm from the upper to the lower plot). The maxima displacements match the
corresponding camera estimations xC with submillimetric precision.
gin of the camera abscissa) with an uncertainty of 0.08 mm.
The xC data are registered together with the corresponding
NMR traces to be used in post processing.
Correspondingly to the cartridge internal structure shown
in Fig.2(a), a 1-D MRI profile is reported in Fig.2(b). That
profile is obtained by averaging over 300 traces. The latter
are selected on the basis of the measured xC values, which
in this case fall within a 0.08 mm interval. The 1-D image
clearly shows the four peaks corresponding to the hydrogel
disks. Such plot is directly obtained as real part of the average-
trace Fourier transform: just minor additional data manipula-
tion is performed. The spatial resolution is set by the NMR in-
trinsic and instrumental linewidth, which -- in absence of field
gradient -- amounts to 1.5 Hz: its effect is only partially com-
pensated in the spectral analyses used in this paper.
For an alternative evaluation of the MRI spatial resolution,
we use a set of traces collected in more than 2000 shots.
Subsets of traces corresponding to narrow xC intervals in the
statistic distribution of xC are averaged to produce MRI pro-
files. The profile displacements are then compared with the
xC variations.
In Fig.3, the panel (b) shows the histogram of xC positions
and five subsets extracted to produce the average time-traces
that finally result in the MRI profiles shown in (a). The four-
peaks profile shifts progressively in accordance with the xC,
with submillimetric accuracy.
In conclusion,
this work demonstrates that
the IDEA
method enables the construction of submillimetric ULF-MRI
setups with a dual-sensor NMR detection, based on Bell and
Bloom OAMs, even operating in an unshielded environment.
Despite their lower SNR compared to SQUIDs, the OAMs
bring the advantage of an intrinsic robustness with respect to
the temporary application of strong magnetic fields, such as
those necessary for spin premagnetization and manipulation.
The described setup does not include ferromagnetic shield-
ing, which would facilitate a rearrangement devoted to apply a
strong premagnetization field in the proximity of the sensors,
so to build a fully static, in-situ experiment. The achieved
sensitivity and the residual magnetic noise make our sensors
about one order of magnitude less performing compared to
typical SQUID and SERF-OAM devices: an aspect which
would require progresses in view of practical applications.
Further steps of our research foresee further improvements
of the active field compensation system and the implemen-
tation of a in-loco premagnetization coil. In the perspective of
real-world applications, it is worth mentioning that our two-
sensor arrangement demonstrates the possibility of extending
the method to multi-sensor setups: more complex sensor ar-
rays can exploit the IDEA concept, and have the potential of
improving the sensitivity, accelerating the measurement and
enabling ULF-MRI of larger-size samples.
SUPPLEMENTARY MATERIAL
See supplementary for experimental details and for a con-
cise description of the various elements of the setup and their
principle of operation
5
21M. C. D. Tayler, T. Theis, T. F. Sjolander, J. W. Blanchard, A. Kent-
ner, S. Pustelny, A. Pines,
and D. Budker, "Invited review article:
Instrumentation for nuclear magnetic resonance in zero and ultralow
magnetic field," Review of Scientific Instruments 88, 091101 (2017),
https://aip.scitation.org/doi/pdf/10.1063/1.5003347.
22V. S. Zotev, A. N. Matlashov, P. L. Volegov, A. V. Urbaitis, M. A. Espy,
and R. H. K. Jr, "SQUID-based instrumentation for ultralow-field MRI,"
Superconductor Science and Technology 20, S367 (2007).
23I. Savukov and T. Karaulanov, "Anatomical MRI with an atomic magne-
tometer," Journal of Magnetic Resonance 231, 39 -- 45 (2013).
24D. Budker and M. Romalis, "Optical magnetometry," Nature Physics , 227 --
234 (2007).
25I. Savukov and T. Karaulanov, "Magnetic-resonance imaging of the human
brain with an atomic magnetometer," Applied Physics Letters 103, 043703
(2013).
26S. Xu, V. V. Yashchuk, M. H. Donaldson, S. M. Rochester, D. Budker, and
A. Pines, "Magnetic resonance imaging with an optical atomic magnetome-
ter," Proceedings of the National Academy of Sciences 103, 12668 -- 12671
(2006), http://www.pnas.org/content/103/34/12668.full.pdf.
27G. Bevilacqua, V. Biancalana, Y. Dancheva, and A. Vigilante, "Restor-
ing narrow linewidth to a gradient-broadened magnetic resonance by
inhomogeneous dressing," Phys. Rev. Applied 11, 024049 (2019),
arXiv:1807.01274.
28G. Bevilacqua, V. Biancalana, P. Chessa, and Y. Dancheva, "Multichannel
optical atomic magnetometer operating in unshielded environment," Ap-
plied Physics B 122, 103 (2016), arXiv:1601.06938.
29V. Biancalana, G. Bevilacqua, P. Chessa, Y. Dancheva, R. Cecchi,
and L. Stiaccini, "A low noise modular current source for stable mag-
netic field control," Review of Scientific Instruments 88, 035107 (2017),
http://dx.doi.org/10.1063/1.4977931.
30G. Bevilacqua, V. Biancalana, Y. Dancheva,
and A. Vigilante, "Self-
adaptive loop for external-disturbance reduction in a differential measure-
ment setup," Phys. Rev. Applied 11, 014029 (2019), arXiv:1803.03212.
31G. Bevilacqua, V. Biancalana, Y. Dancheva, and L. Moi, "Larmor fre-
quency dressing by a nonharmonic transverse magnetic field," Phys. Rev.
A 85, 042510 (2012), arXiv:1112.1309 [physics.atom-ph].
32S. Haroche, C. Cohen-Tannoudji, C. Audoin, and J. P. Schermann, "Modi-
fied Zeeman hyperfine spectra observed in 1H and 87Rb ground states inter-
acting with a nonresonant RF field," Phys. Rev. Lett. 24, 861 -- 864 (1970).
33G. Bevilacqua, V. Biancalana, Y. Dancheva, L. Stiaccini, and A. Vigilante,
"Spurious ferromagnetic remanence detected by hybrid magnetometer,"
Review of Scientific Instruments 90, 046106 (2019), arXiv:1904.01901.
34V. Biancalana, Y. Dancheva, and L. Stiaccini, "Note: A fast pneumatic
sample-shuttle with attenuated shocks," Review of Scientific Instruments
85, 036104 (2014), arXiv:1401.6454.
35G. Bevilacqua, V. Biancalana, A. Ben Amar Baranga, Y. Dancheva, and
C. Rossi, "Micro-Tesla NMR J-coupling spectroscopy with an unshielded
atomic magnetometer," Journal of Magnetic Resonance 263, 65 -- 70 (2016),
arXiv:1510.06250.
36G. Bevilacqua, V. Biancalana, Y. Dancheva, A. Vigilante, A. Donati, and
C. Rossi, "Simultaneous detection of H and D NMR signals in a micro-
Tesla field," The Journal of Physical Chemistry Letters 8, 6176 -- 6179
(2017), pMID: 29211488, arXiv:1510.06250.
37W. E. Bell and A. L. Bloom, "Optically driven spin precession," Phys. Rev.
Lett. 6, 280 -- 281 (1961).
38T. Scholtes, V. Schultze, R. IJsselsteijn, S. Woetzel, and H.-G. Meyer,
"Light-narrowed optically pumped Mx magnetometer with a miniaturized
Cs cell," Phys. Rev. A 84, 043416 (2011).
39G. Bevilacqua, V. Biancalana, and Y. Dancheva, "Atomic orientation driven
by broadly-frequency-modulated radiation: Theory and experiment," Phys.
Rev. A 94, 012501 (2016).
nuclear
Science
"Tumor
Damadian,
resonance,"
mag-
(1971),
detection
171,
by
1151 -- 1153
1I. I. Rabi, J. R. Zacharias, S. Millman, and P. Kusch, "A new method of
measuring nuclear magnetic moment," Phys. Rev. 53, 318 -- 318 (1938).
2F. Bloch, W. W. Hansen, and M. Packard, "Nuclear induction," Phys. Rev.
69, 127 -- 127 (1946).
3E. M. Purcell, H. C. Torrey, and R. V. Pound, "Resonance absorption by
nuclear magnetic moments in a solid," Phys. Rev. 69, 37 -- 38 (1946).
4R.
netic
https://science.sciencemag.org/content/171/3976/1151.
5P. C. Lauterbur, "Image formation by induced local interactions: Ex-
amples employing nuclear magnetic resonance," Nature 242 (1973),
10.1038/242190a0.
6C.
R. Merritt,
sues?"
Radiology
https://doi.org/10.1148/radiology.173.2.2678243.
7R. Kraus, M. Espy, P. Magnelind, and P. Volegov, Ultra-Low Field Nuclear
Magnetic Resonance A New MRI Regime (Oxford University Press, Oxford,
UK, 2014).
8R. E. Sepponen, J. T. Sipponen, and A. Sivula, "Low field (0.02 T) nuclear
magnetic resonance imaging of the brain," Journal of Computer Assisted
Tomography (19859).
9R. McDermott, S. Lee, B. t. Haken, A. H. Trabesinger, A. Pines, and
J. Clarke, "Micro-Tesla MRI with a superconducting quantum interference
device," Proceedings of the National Academy of Sciences 101, 7857 -- 7861
(2004), http://www.pnas.org/content/101/21/7857.full.pdf.
"Ultrasound
173,
the
is-
2678243,
are
pMID:
304 -- 306
safety:
what
(1989),
10M. Mössle, W. R. Myers, Seung-Kyun Lee, N. Kelso, M. Hatridge,
A. Pines, and J. Clarke, "SQUID-detected in vivo MRI at microtesla mag-
netic fields," IEEE Transactions on Applied Superconductivity 15, 757 -- 760
(2005).
11R. McDermott, A. H. Trabesinger, M. Mück, E. L. Hahn, A. Pines, and
J. Clarke, "Liquid-state NMR and scalar couplings in micro-Tesla magnetic
fields," Science 295, 2247 -- 2249 (2002).
12A. N. Matlachov, P. L. Volegov, M. A. Espy, J. S. George, and R. H. Kraus,
"SQUID detected NMR in micro-Tesla magnetic fields," Journal of Mag-
netic Resonance 170, 1 -- 7 (2004).
13M. Burghoff, S. Hartwig, L. Trahms, and J. Bernarding, "Nuclear magnetic
resonance in the nano-Tesla range," Applied Physics Letters 87, 054103
(2005), https://doi.org/10.1063/1.2006981.
14V. S. Zotev, A. N. Matlachov, P. L. Volegov, H. J. Sandin, M. A. Espy, J. C.
Mosher, A. V. Urbaitis, S. G. Newman, and R. H. Kraus, Jr, "Multi-channel
SQUID system for MEG and ultra-low-field MRI," Proceedings, Applied
Superconductivity Conference (ASC 2006): Seattle, Washington, USA, Au-
gust 27-September 1, 2006, IEEE Trans. Appl. Supercond. 17, 839 -- 842
(2007), physics/0611290.
15S. K. Lee, M. Mössle, W. Myers, N. Kelso, A. H. Trabesinger, A. Pines,
and J. Clarke, "SQUID-detected MRI at 132 µT with T1-weighted contrast
established at 10 µT-300 mT," Magnetic Resonance in Medicine 53, 9 -- 14
(2004).
16M. Mössle, S.-I. Han, W. R. Myers, S.-K. Lee, N. Kelso, M. Hatridge,
A. Pines, and J. Clarke, "SQUID-detected micro-Tesla MRI in the pres-
ence of metal," Journal of Magnetic Resonance 179, 146 -- 151 (2006).
17P. T. Vesanen, J. O. Nieminen, K. C. J. Zevenhoven, J. Dabek, L. T. Parkko-
nen, A. V. Zhdanov, J. Luomahaara, J. Hassel, J. Penttilä, J. Simola, A. I.
Ahonen, J. P. Mäkelä, and R. J. Ilmoniemi, "Hybrid ultra-low-field MRI
and magnetoencephalography system based on a commercial whole-head
neuromagnetometer," Magnetic Resonance in Medicine 69, 1795 -- 1804
(2012).
18A. N. Matlachov, P. L. Volegov, M. A. Espy, R. Stolz, L. Fritzsch, V. Za-
kosarenko, H. . Meyer, and R. H. Kraus, "Instrumentation for simultaneous
detection of low field nmr and biomagnetic signals," IEEE Transactions on
Applied Superconductivity 15, 676 -- 679 (2005).
19T. Theis, P. Ganssle, G. Kervern, S. Knappe, J. Kitching, M. P. Ledbet-
ter, D. Budker, and A. Pines, "Parahydrogen-enhanced zero-field nuclear
magnetic resonance," Nature Physics 7, 571 -- 575 (2011), arXiv:1102.5378
[physics.chem-ph].
20R. H. Koch, J. Z. Sun, V. Foglietti, and W. J. Gallagher, "Flux dam, a
method to reduce extra low frequency noise when a superconducting mag-
netometer is exposed to a magnetic field," Applied Physics Letters 67, 709 --
711 (1995), https://doi.org/10.1063/1.115282.
I. SUPPLEMENTAL MATERIAL
A. Setup overview
The setup is made up of two independent parts, an MRI
subsystem and a magnetometric sensor recording MRI signal.
The sensor is a Bell and Bloom37 OAM, which operates in an
unshielded environment28, and has a dual channel head allow-
ing for differential measurements.
Due to the presence of static bias field (along z) the sen-
sors, being scalar in nature, have first-order response only to
Bz variations:
δ B = (cid:126)B0 − δ (cid:126)B ≈ (cid:126)B0 · δ (cid:126)B
B0
= δ B(cid:107).
(5)
The common mode variations of this field component is ac-
tively stabilized.
Additional coils, belonging to MRI subsystem, permit the
application of resonant or non resonant (non-adiabatically
varying) field pulses,
to manipulate (tipping) the nuclear
spins35,36. The magnetometric operation is destroyed by these
pulses, and recovers completely in about 100 ms.
In the present implementation, a remote magnetization
approach is applied. To perform MRI, a pneumatic sys-
tem moves cyclically the sample from a 1 T premagnetiz-
ing assembly, configured as an Halbach magnet array to the
measurement region and backwards34. The sample transfer,
tipping pulse, and recovery time amount globally to about
250ms:
this interval constitutes the dead-time between the
premagnetization and the measurement.
A constant field gradient is applied to the measurement re-
gion with the aim of performing frequency encoding neces-
sary to the MRI. A good spatial resolution requires a rather
strong gradient (30-100 nT/cm). The AMR resonance would
be completely destroyed by this gradient, with a width in-
crease from about 25 Hz up to 200-700 Hz. A technique based
on inhomogeneous magnetic dressing (IDEA method27) per-
mits to maintain the AOM operability despite the gradient, so
to register successfully MRI profiles.
B. Magnetometer
The OAM uses two sealed cells (cylindrical shape, 2 cm in
length, 2.2 cm in diameter) with Cs vapour and buffer gas, and
two fiber-coupled laser sources whose radiation is mixed, split
and delivered to both the cells (see Fig.1). The first laser radi-
ation is at milli-Watt level and optically pumps the Cs vapor
into a stretched state, which then evolves due to the presence
of transverse magnetic field. This pump radiation is tuned
to the proximity the Cs D1 line (894 nm), and its optical fre-
quency is periodically made resonant with the Fg = 3→ Fe = 4
transition of the D1 line set. This condition produces both
light-narrowing38,39 -- due to strong hyperfine pumping to the
Fg = 4 ground state -- and Zeeman pumping, due to the weak
excitation of far detuned resonances starting from the Fg = 4
level, and to the light polarization, which is made circular.
6
The second laser radiation is attenuated down to microwatt
level, and serves as a probe. It co-propagates with the pump,
and is linearly polarized. The rotation of the probe polariza-
tion plane is a measure of the atomic vapour magnetization
along the propagation direction (let it be x). The probe beam
is tuned to the proximity of the Fg = 4 manyfold in the D2
line, at 852 nm.
The wavelength difference of the two radiations facilitates
two tasks: it is possible to use a multiorder waveplate which
acts as a λ /4 plate on the probe and as a full-λ on the pump
radiation, to obtain the aforementioned polarizations; it is pos-
sible to filter out the pump radiation after the interaction with
Cs, to perform polarimetry of the probe radiation only.
The pump radiation is made resonant periodically, and
sinchronously with the atomic precession around a transverse
(z oriented) static magnetic field.
Let ωM and ωL be the angular frequency of the laser modu-
lation signal and of the atomic precession, respectively. Scan-
ning of ωM around ωL enables the registration of the AMR
profile. Linewidths as narrow as 25 Hz are recorded in oper-
ative conditions. This width is mainly due to spin-exchange
relaxation and to probe perturbation: a resonance narrowing
down to 5 Hz is observed when operating at lower cell tem-
perature and weaker (or more detuned) probe radiation.
To the end of maintaining a high sensitivity level, it is im-
portant to preserve a narrow resonance width. The presence of
buffer gas (3 kPa N2) prevents time-of-flight line broadening
and radiation trapping phenomena. Spurious field gradients
are accurately compensated, and the IDEA method is used to
counteract the effect of the strong field gradient deliberately
applied in MRI to perform the frequency encoding .
During the measurements, ωM is made resonant (ωM = ωL)
and kept constant. In addition, the modulation signal at ωM
is superimposed to an adjustable offset, by which the optical
frequency of the pump laser is kept at its optimal value28.
C.
IDEA (principle of operation)
In the proximity of each sensor, displaced about 10 cm
along the x direction, a hollow ferrite nucleus (with the laser
beam passing through the hole) wound by a solenoidal coil
produces a dipolar field used to dress magnetically the Cs
atoms.
Such dipolar oscillating field BD is oriented along x -- thus
transversely to the static field Bz -- and has an important inho-
mogeneity along the x direction.
The static field Bz is made not-homogeneous along x for the
MRI-frequency encoding purpose. Its gradient G = ∂ Bz/∂ x is
set by permanent magnets (small -- 4 cm3 -- Nd magnets, well
displaced -- about 1 m -- from the measurement region) and ar-
ranged in a quadrupolar configuration: both the sample nuclei
and the sensor atoms precess in a field whose static z compo-
nent is well described by Bz = B0 +Gx, where x is the distance
from the sensor center, and B0 is the field at the center of the
cell. Thus the Cs and proton Larmor frequencies set by Bz are
position dependent along the optical axis x.
The dressing field BD (Eq.2) is produced by an alternat-
ing current supplying a coil (100 turns wound on a hollow
cylinder ferrite, Richco RRH-285-138-286: 28.5 mm external
diameter, 13.77 internal diameter, 28.57 mm length). This de-
vice produces a dipolar field with adjustable intensity (several
µT at 40 kHz). The inductance is measured (1.35 mH) and a
resonant LC circuit is built to enhance BD.
The presence of the dressing field, makes the time evolution
of the atomic magnetization more complicated than a simple
precession around the static field Bz. In particular, the y and
z components of the atomic magnetization follow a deeply
modified trajectory on the Bloch sphere. On the contrary, its
x component keeps oscillating harmonically. The only effect
of BD on the evolution of the x component is slowing down of
its oscillation frequency.
In conclusion, the x component of the atomic magnetization
-- the observed quantity revealed by the Faraday rotation of the
probe beam polarization -- oscillates harmonically at a dressed
(reduced) angular frequency with respect to its unperturbed
precession around the static field31 (see Eq.1).
D. Magnetic Field sources
To adjust the static field and its inhomogeneties, the system
includes three large size (1.8 m side) square Helmholtz coils
to compensate the environmental field, two anti-Helmholtz
coils (same size) to compensate the diagonal elements of the
field Jacobian, and three pairs of dipoles arranged to pro-
duce quadrupolar fields in the measurement region, which
compensate the off-diagonal elements. The latter are made
of permanent magnets for coarse compensation and numeri-
cally controlled electromagnets for the fine adjustments. Eight
numerically controlled current sources designed with a hy-
brid (switching plus linear architecture)29 supply the field and
field-gradient coils. Automated procedures are applied to null
the gradient terms and the transverse field, as well as to set the
amplitude of the residual field along the z direction. The latter
is furtherly stabilized as discussed in section I E.
For nuclear spin manipulation, a smaller (50 cm side)
Helmholtz pair is used to apply appropriate pulses (π/2 tip-
ping pulses). The permanent magnets necessary to generate
the quadrupole field with the frequency-encoding gradient G
are inserted after the completion of the above described field
optimization procedure.
7
It is useful to recombine the two signals to extract their
common mode (CM) term and difference-mode (DM) term:
far-located sources (which in our measurement constitute a
disturbance) contribute to the CM only, while the MRI signal
appears in the DM one.
The polarimetric signal of one channel is modeled as
s(t) = Aei[ωMt+ϕ(t)].
(6)
is acquired,
The measured polarimetric signal
real-
time numerically demodulated and elaborated by a field-
programmable-gate-array (FPGA). After the demodulation,
ϕ(t) is converted into an error signal by means of an IIR filter
characterized by a set of forward and backwards coefficients.
Afterwards, the error signal drives the current generator sup-
plying compensation coils.
As discussed in the appendix of Ref.28, a small variation
δ B(t) of the field component along the static field Bz acts on
ϕ(t) according to the equation
ϕ = −γCsδ B(t)− Γϕ,
(7)
where Γ is the AMR linewidth.
IIR forward and backwards coefficients are initially chosen
in order to combine the phase with its derivative, calculated
with a finite-difference extimation. Thus, filtering the phase
ϕ(t) by means of IIR filter, ensures that the resulting error-
signal is proportional to magnetic field variation as in Eq.7.
Thereafter IIR coefficients are automatically optimized using
the Nelder-Mead algorithm; in such way the evaluation of er-
ror signal considers firstly the ideal magnetometer response,
but is secondly adapted to compensate delays occurring dur-
ing elaboration process and specific behaviour of the chain
of elements that constitute the loop. The optimization pro-
cess is based on the analysis of the polarimetric signal power
spectrum. More precisely, the IIR optimization is targeted to
maximize the energy associated to the carrier peak at ωM with
respect to other spectral components due to disturbances.
The error signal at the output of the IIR filter feeds a
voltage-to-current converter, which supplies a pair of com-
pensation Helmholtz coils. The magnetic field generated by
means of the coils is nominally oppostite to the CM of the
magnetic disturbances detected by the sensors, in such way
those disturbances can be successfully reduced to a large
extent30.
E. Magnetic Field stabilization
As said in the main text, the dual sensor detector produces
two magnetometric signals which contain the measurement
of field variations due to both far-located and close-located
sources. The first contribution is dominant and appears with
the same sign on the two sensors. The second term appears
with opposite signs, provided that the close-located source
(that is the sample) is opportunely located with respect to the
sensors.
F. Polarimetric signal elaboration
After the interaction with each atomic sample, the pump ra-
diation is stopped by a narrowband interference filter and the
probe beam polarization is analyzed by a balanced polarime-
ter (a 45 degree Wollaston prism followed by a couple of Si
photodetectors, whose photocurrent unbalance is preamplified
by a ultra-low noise transimpendance amplifier).
The two polarimetric signals (one for each cell) are nearly
harmonic (Eq.6), at the oscillation frequency of the x compo-
nent of the atomic magnetization.
These signals are digitized by a 1MS/s 16 bit card for the
duration of the measurement: 6 sec after the application of
the tipping pulse, with a start delay of 150 msec, necessary
to recover the magnetometer's steady state. This signal dig-
itization is performed parallelly and independently from the
FPGA real-time acquisition discussed in Sec.I E.
Provided that the pump modulation is performed at a fixed
frequency matching the AMR center, the measured signal os-
cillates at such forcing frequency, while the field information
is contained in its phase ϕ, as mentioned above (Eq.7).
A discrete Hilbert transform is used to infer the imaginary
part of the digitized polarimetric signal. The phase ϕ is then
extracted from the complex signal after having demodulated
the frequency of the forcing term, its time derivative is fi-
nally evaluated so to estimate the field variation at the time
as δ B(t) = (Γϕ − ϕ)/γCs.
The mean value of the two δ B's measured by the two sen-
sors constitutes the CM term (driven by disturbances from far
locates field sources), while their difference (DM term) con-
tains information about close located sources and is further
elaborated to extract the MRI traces.
G. MRI detection and signal elaboration
As shown in Fig.1 the MRI sample is located at the same x,
intermediate z and different y with respect to the sensors. Con-
sequently, when the nuclear magnetization precesses around
the static field Bz, it produces periodic variations of the mea-
sured field component (the measured δ B is the variation of
the z component, Eq.5). These variations occur with opposite
signs in the two sensors, so to appear in the DM term.
The reconstructed images reported in this work are obtained
from the real part of the Fourier transform of the DM term
S(t): S(Ω) = F (S(t)), then scaling from Ω to x via the gradi-
ent γHG.
Only minor additional signal manipulations are applied.
The first one is a time-domain windowing having the dual
scope of apodizing the signal, reducing the noise contribution
from the final (signal-less) part of the time trace, and of par-
tially compensating the exponential decay of the NMR signal.
To this aim we apply a window in the form
W (t) = exp(t/T ) [1− tanh (β (t −tcut))]2 ,
8
(8)
where T is kept lower than the measured proton decay time,
and tcut and β are parameters adjusting the position and the
slope of the transition to zero of the term in square brackets.
The second operation is to take into account the delay t0
between the application of the tipping pulse and the measure-
ment: the transformed signal S(Ω) is scaled by exp(iΩt0) be-
fore extracting the real part.
The latter operation (that is a translation in the time domain)
comes with some artefacts, making a third manipulation op-
portune. Due to spurious deterministic disturbances at 50 Hz
and harmonics as well as at ωM, additional terms appear su-
perimposed to the MRI profile. These terms have a periodicity
t0 in the frequency domain, which facilitates their identifica-
tion and subtraction.
H. MRI Sample
The sample container has the external shape of a cylinder
(19 mm in diameter diam, 32 mm in length). It is made of
polyether ether ketone (PEEK, Victrex), a non-conductive ma-
terial selected for its excellent mechanical properties in matter
of machinability and robustness (the cartridge must withstand
thousands of shuttling stresses).
For MRI test purposes, the cap of the cartridge includes
an extension (see Figs.1 and 2), which makes the measured
material (water in hydrogels, to prevent diffusion processes)
occupy four possible regions having the shape of hollow disks
(5mm thick, 15mm external diameter, 5mm internal diame-
ter).
As previously pointed out33, despite the negligible intrinsic
magnetic remanence of the PEEK, the machining process may
introduce detectable surface contamination, making the con-
tainer spuriously and weakly ferromagnetic. It is known that a
paramagnetic response produces MRI artefacts that are impor-
tant at high field, but negligible in ULF regime. On the other
hand, the ferromagnetic response produce effects that persist
in ULF, up to become a dominant artefact source. Lathing the
cartridge with non magnetic tools (glass and Ti blades in our
case), turned out to be necessary to prevent this problem.
|
1802.10081 | 2 | 1802 | 2018-03-23T02:05:52 | Observation of the topological edge state in X-ray band | [
"physics.app-ph",
"physics.acc-ph",
"physics.optics"
] | The possibility of obtaining robust edge state of light by mimicking the topological properties of solid state system, have brought a profound impact on optical sciences. With the advent of high-brilliance, accelerator-driven light sources such as storage rings or X-ray lasers, it has become attractive to extend the concept of optical topological manipulation to the X-ray regime. In this paper, we theoretically proposed and experimentally demonstrated the topological edge state at the interface of two photonic crystals having different band-gap topological characteristics for X-ray. Remarkably, this topologically protected edge state is immune to the weak disorder in form of the thickness disorder and strong disorder in form of the positional disorder of layers in the structure, as long as the zero-average-effective-mass condition is satisfied. Our investigation therefore brings the topological characteristics to the X-ray regime, provides new theoretical tools to study X-ray optics and may pave way to exploit some important potential applications, such as the high efficiency band filter in X-ray band. | physics.app-ph | physics | Observation of the topological edge state in X-ray band
Qiushi Huang*, Zhiwei Guo*, Jiangtao Feng, Changyang Yu, Haitao Jiang†, Zhong Zhang,
Zhanshan Wang†, and Hong Chen
Key Laboratory of Advanced Micro-structure Materials, MOE, School of Physics Science and Engineering, Tongji
University, Shanghai 200092, China
Abstract
The possibility of obtaining robust edge state of light by mimicking the topological properties of solid
state system, have brought a profound impact on optical sciences. With the advent of high-brilliance,
accelerator-driven light sources such as storage rings or X-ray lasers, it has become attractive to extend
the concept of optical topological manipulation to the X-ray regime. In this paper, we theoretically
proposed and experimentally demonstrated the topological edge state at the interface of two photonic
crystals having different band-gap topological characteristics for X-ray. Remarkably, this topologically
protected edge state is immune to the weak disorder in form of the thickness disorder and strong
disorder
in form of
the positional disorder of
layers
in
the structure, as
long as
the
zero-average-effective-mass condition is satisfied. Our investigation therefore brings the topological
characteristics to the X-ray regime, provides new theoretical tools to study X-ray optics and may pave
way to exploit some important potential applications, such as the high efficiency band filter in X-ray
band.
Key words: Photonic crystal; Topological edge mode; X-ray band
* These authors contributed equally to this work
† [email protected]
† [email protected]
1
I. INTRODUCTION
The explorations of topological properties and robust chiral transport have been the subject of
intensive research in many braches of physics, from electronics [1, 2] to photonics [3-25], acoustic
and phononics [26-29] as well as mechanics [30, 31]. Photonic topological insulators that can
manipulate light in a manner similar to the topological insulators controlling the electrons are
rapidly developing in the past few years. The early studies are focused on photonic systems with
relatively large physical size in the low-frequency micro-wave band, including gyromagnetic
materials [5, 6], bi-anisotropic metamaterials [7, 22], and coupled resonator waveguides [9, 10,
19]. Recently, based on dielectric photonic crystals (PCs), photonic topological insulators have
been realized in visible band and used for light splitting, unidirectional propagation, etc. [8, 25].
The most remarkable property of photonic topological insulators is that there would be some kinds
of topological protection for the light propagation. As a result, photonic topological insulators
facilitate the design of novel photonic devices that are robust against the disorder and fluctuations
[7, 17, 24]. On the other hand, the short wavelength optical science, particularly in the X-ray
region, is undergoing a tremendous development owing to its unique advantages of high resolution
in microscopy, spectroscopy and ultrafast dynamics studies [32-35]. With the advent of a new
generation of high-brilliance, high coherent X-ray sources, such as the lab-based plasma source,
high harmonic generation, and the accelerator-based large facilities, the advantages of X-ray
science and technology are further promoted. To materialize the powerful X-ray tool, novel optics
with flexible functionalities are demanded to reflect, focus and shape the X-ray beam in an
exquisite way [36, 37]. Beside the application need, modern thin film deposition and
nanofabrication
technology
is also advancing rapidly and one-dimensional (1-D) and
two-dimensional (2-D) nanoscale structures with large area have been frequently realized. For
2
example, ultra short period multilayer mirrors with periodic bilayer thickness of 1.6~2.5 nm are
developed for soft X-ray water window microscopes [38] or hard X-ray monochromators for
synchrotron beam lines [39], in order to select the desired waveband. In addition, aperiodic
multilayer mirrors are frequently produced to achieve broadband angular or spectral response [40].
Two dimensional 15 nm particle arrays have also been fabricated in centimeters area using
photolithography [41]. For X-ray devices working in the very short wavelength, the fabricating
difficulty is much higher than those working in the visible wavelength. The fabricating accuracy
for X-ray devices should be very high and even small perturbations would strongly degrade the
performance of the devices. Therefore, given to the wide applications and significant progress of
technology for X rays, a question naturally arises: can the topological properties be applied in the
X-ray region and enable new phenomena and devices that are robust against perturbations? In this
paper, we use a 1-D symmetric PC system to demonstrate the topological band-gaps and the edge
states in X-ray band for the first time. Very interestingly, these topologically protected edge states
are robust against a variety of perturbations, such as thickness disorder (weak disorder) and
positional disorder (strong disorder), which can facilitate the design of novel X-ray filters/mirrors
immune to some kinds of disorders.
In general, the edge state can be formed at the boundary separating two PCs having different
band-gap topological invariant, such as geometric phase in 1-D systems [14, 27]. As a geometric
phase initially acquired from the electron Bloch state, the Zak phase is considered to the a 𝑍2
topological invariant (π or 0) for classifying bands of 1-D electronic systems with inversion
symmetry [42]. By analogy, for the 1-D photonic system with inherent mirror symmetry, the
corresponding Zak phase is a quantized value of 0 or π [27, 42]. In particular, the Zak phase of
3
the bands can be directly determined by calculating the reflection phase of the band-gap [43].
Moreover, by mapping Maxwell's equations to the Dirac equation, the band-gaps of 1-D PCs with
different topological orders can be mapped to the band-gaps with effective negative permittivity or
negative permeability [4, 13]. Here, we will verify the topological nontrivial insulators in X-ray
band from two kinds of reflection phases and the effective parameters. Considering the interest of
fundamental science, we firstly designed and validated the topological edge state in X-ray band.
Actually, almost all the natural material can hardly reflect the X wave and it cannot be used as a
reflector for X wave. As a good solution, the X-ray reflector is constructed by utilizing band-gap
of multilayer structures to reflect the X-rays. To achieve designed property, periodic structures
with the unit-cell size comparable with or even smaller than the wavelength are needed. As a
result, the performance of the multilayer structures is often degraded by the fluctuation of the
layers and the difficulty of designing X-ray band-gap is how to deposit the nano-meter scale thick
multi-layer with layer thicknesses typically oscillating in the range of 1.0-5.0 nm [44, 45]. Here
we introduce the concept of topology into X-ray band and realize the X-ray topological edge state.
At this topological edge state, we experimentally demonstrate that the transmission property does
not undergo noticeable degradation with moderate degree of disorders, which may relieve the
strict tolerance in the fabrication of X-ray filters.
The paper is organized as follows. In Sec. II, by calculating the reflection phases and the
effective parameters, we reveal that two photonic insulators with different topological properties
for X-ray band can be realized in 1-D symmetry PCs. Moreover, we demonstrated that the
topological edge state forms at the boundary separating two PCs having different band-gap
topological characteristics for X-ray. In Sec. III, we demonstrated experimentally that the
4
topologically protected edge state is robust against a variety of perturbations, such as thickness
disorder and positional disorder. Finally, we conclude in Sec. V.
II. THE DESIGN OF THE TOPOLOGICAL EDGE STATE FOR X-RAYS
We begin with the design of the two 1-D symmetry PCs for X-ray band with different
topological properties. This is because the topological property of the symmetric structure is easier
to characterize than the asymmetric structure [4, 14, 27]. Two different PCs are marked by 𝑃𝐶𝐴
and 𝑃𝐶𝐵, respectively, as is shown in Fig. 1. The unit cells of the PCs are shown in the inset of Fig.
1. It is generally known that the refractive indices of materials are close to 1 in X-ray regime. This
feature is very unfavorable to design band-gap for PC and to obtain enough reflectance it must be
grazing incident to the PC. Considering the factors of high refractive index difference, low loss,
thin film thickness and the experimental processing technology, the layer 𝐴1 and 𝐴2 in 𝑃𝐶𝐴 are
carbon (C) and tungsten (W) with 𝑑𝑐 = 1.5𝑛𝑚, 𝑑𝑤 = 1.5𝑛𝑚. These materials have relatively
large refractive indices contrast in the frequency range of interest. When the frequency is fixed to
1.94 × 1018 𝐻𝑧, the refractive indices are [46]:
𝑛𝐶 ≈ 0.99999 + 1.11473 × 10−8
𝑛𝑊 ≈ 0.99995 + 3.94227 × 10−6 . (1)
The structure 𝑃𝐶𝐴 is denoted by (𝐶𝑊𝐶)10 with the thickness of the unit cell and the period of the
unit cell are 𝑑𝐴 = 4.5𝑛𝑚 and 𝑁𝐴 = 10, respectively. In general, the band-gap realized by PC can be
tuned to be closed and reopened by changing the materials or the thickness of the different layers.
Moreover, there is a topological transition in this process [14, 27]. So the 𝑃𝐶𝐵 with distinct
topological properties to 𝑃𝐶𝐴 can be easily realized by changing the layers of 𝑃𝐶𝐴. In order to
simplify the experimental operation, the layer 𝐵1 and 𝐵2 in 𝑃𝐶𝐵 are tungsten (W) and carbon (C)
5
with 𝑑𝑊 = 1.5𝑛𝑚, 𝑑𝐶 = 1.5𝑛𝑚, respectively. The structure 𝑃𝐶𝐵 is denoted by (𝑊𝐶𝑊)10 with the
thickness of the unit cell and the period of the unit cell are 𝑑𝐵 = 4.5𝑛𝑚 and 𝑁𝐵 = 10, respectively.
Take the first structure (be marked as 𝑃𝐶𝐴) for example, which consists of alternative layers
of 𝐴1(𝑛𝐴1), 𝐴2(𝑛𝐴2) and 𝐴1(𝑛𝐴1) with the thickness of 𝑑𝐴1, 𝑑𝐴2 and 𝑑𝐴1 , respectively. 𝑛𝐴1
and 𝑛𝐴2 are the refractive indices of 𝐴1 and 𝐴2 layer, respectively. The thickness of the unit cell
is 𝑑𝐴 = 𝑑𝐴1 + 𝑑𝐴2 + 𝑑𝐴1. The period number of unit cell is 𝑁𝐴. To demonstrate the topological
differences of the 1-D PCs in X-ray band, we have calculated the reflection phase and effective
parameters for the designed PCs. Two different PCs are working at the frequency 1.94 × 1018 𝐻𝑧
(Cu-Kα line E = 8.04 kev) in the grazing incidence range of 𝛼 ∈ (0.8𝑜, 1.3𝑜).
Fig. 1. The scheme of the topological edge mode formed at the interface separating two PCs
having different bandgap topological characteristics when the X-ray impinges on the sample at
oblique incidence. 𝑃𝐶𝐴 is on the left-hand of the interface and 𝑃𝐶𝐵 is on the right-hand of the
interface. The corresponding unit cells of two different PCs are shown in the inset.
Next, we analyze the topological properties of 𝑃𝐶𝐴 and 𝑃𝐶𝐵. Particularly, the physical
mechanism of the formation of the topological edge state and the corresponding electric field
distribution are verified. Firstly, the effective parameters of 𝑃𝐶𝐴 and 𝑃𝐶𝐵 can be get by the
optimal effective medium theory [4, 47], which are shown in Fig. 2(a) and Fig. 2(b), respectively.
6
By comparing the Dirac equation of electronic system and the Maxwell equation of photonic
system, the Maxwell equations can be written as the form of Dirac equation [13]
[−𝑖𝜎𝑥𝜕𝑥 + 𝑚(𝑥)𝜎𝑧 + 𝑉(𝑥)] (
√𝜀0𝐸𝑧
√𝜇0𝐻𝑦
) = 𝐸 (
√𝜀0𝐸𝑧
√𝜇0𝐻𝑦
), (2)
where 𝑚(𝑥) = (
𝜔
2𝑐
) [𝜀𝑟(𝑥) − 𝜇𝑟(𝑥)] denotes the effective mass. 𝜔 is the angle frequency. 𝑉(𝑥)
and 𝐸 are the effective potential and energy eigenvalue, respectively [13]. Considering the X-ray
band-gap of 𝑃𝐶𝐴 , which can be effective to the epsilon-negative (ENG) metamaterial with
negative permittivity and positive permeability, the effective mass is a negative value. However,
for the X-ray band-gap of the structure 𝑃𝐶𝐵, which can be effective to the mu-negative (MNG)
metamaterial with negative permeability and positive permittivity, the effective mass is a positive
value. According to the Dirac equation, the edge state exists in the hetero-structure composed of
positive and negative masses when 𝑚 = ∫ 𝑚1(𝑥)𝑑𝑥 + ∫ 𝑚2(𝑥)𝑑𝑥
= 0, where 𝑚1(𝑥) and
0
−𝐿1
𝐿2
0
𝑚2(𝑥) are effective masses of the length 𝐿1 and 𝐿2, respectively [48]. Moreover, the sign of
effective mass corresponds to different topological order [48]. For the Cu-Kα line (λ=0.154 nm)
1.94 × 1018 𝐻𝑧, there is an edge mode exist in the angle spectrum when the incident grazing
angle is α = 1.07𝑜. In this case the effective masses of 𝑃𝐶𝐴 and 𝑃𝐶𝐵 are 𝑚𝐴 ≈ −2.03 × 108
and 𝑚𝐵 ≈ +2.03 × 108, respectively. It has shown that edge state in 1-D system is robust against
effective mass distribution so long as 𝑚 = 𝑚𝐴 + 𝑚𝐵 = 0 [4, 13]. According to the relationship
between effective parameters and the topological invariant [4, 13], we can conclude the band-gaps
of 𝑃𝐶𝐴 and 𝑃𝐶𝐵 can be seen as two X-ray insulators with different topological properties and the
edge state will appear at the interface of two topological distinct structures.
In order to the further verify the topological difference between 𝑃𝐶𝐴 and 𝑃𝐶𝐵, we have also
calculated the reflection phase, which is another quantity that is considered to be able to determine
7
the topological property of the band-gap [27]. The reflectance phases of 𝑃𝐶𝐴 and 𝑃𝐶𝐵 belong to
(−π, 0) and (0, π), respectively, which are shown in Fig. 2(c). So, we can conclude that these
two PCs are topologically distinct. Take into account the losses of C and W layers in 𝑃𝐶𝐴 and
𝑃𝐶𝐵, the incident X-ray will be partially reflected on the incident plane rather than the perfect
tunnelling. But in any way, the topological edge state can be localized at the interface of two
different topologically distinct structures, as is shown in Fig. 2(d). Moreover, the topological edge
states are protected by the topological phase transition across the interface. Thus they are robust
against the defects, the loss, and the disorder, which do not change the topological phase of the
structure [17, 24].
Fig.2 The separate structructures 𝑃𝐶𝐴 and 𝑃𝐶𝐵 are corresponding to the ENG (a) and MNG
metamaterials (b), respectively. (c) The reflection phases of 𝑃𝐶𝐴 (marked by the green line) and
𝑃𝐶𝐵 (marked by the purple line), respectively. (d) The electric field distribution in the composite
structure 𝑃𝐶𝐴 − 𝑃𝐶𝐵. The edge mode is formed at the boundary between two topologically distinct
structures 𝑃𝐶𝐴 and 𝑃𝐶𝐵. The reflection at the entrance face of the structure is mainly caused by
the losses of C and W layers.
The reflectance of grazing incidence X-ray is performed using Bede Refs software and IMD
software [48]. When the X-ray is incident to the 𝑃𝐶𝐴, there is a band-gap (painted in green ) in the
angular spectrum, as is shown in Fig. 3(a). Similarly, there is also a band-gap in the angular
8
spectrum (painted in pink ) when the X-ray is incident to 𝑃𝐶𝐵, see Fig. 3(b). It can be found that
when the X-ray is incident to the hetero-structure composed of 𝑃𝐶𝐴 and 𝑃𝐶𝐵, the reflection dip
will appear in the middle of the band-gap, as is shown in Fig. 3(c). In other words, the separate
𝑃𝐶𝐴 and 𝑃𝐶𝐵 are the insulators of X-ray. However, when the X-ray impinges on the structure
combined of 𝑃𝐶𝐴 and 𝑃𝐶𝐵, the structure become nearly transparent because the topological edge
state will form at the boundary separating two PCs, which corresponds to the reflection dip α =
1.07𝑜 in Fig. 3(c). It should be noted that this topological edge mode is not limited at the fixed
frequency 1.94 × 1018 𝐻𝑧 and it can be tuned in a certain frequency scope within the bandgap
The reflectance of the edge state in Fig. 3(c) is not zero, which is caused by the losses of C and W
layers in 𝑃𝐶𝐴 and 𝑃𝐶𝐵. Considering the three structures mentioned above, we have calculated the
reflectance for different frequency and the grazing incident angle, which is shown in Fig. 3(d)-(f).
The reflection angle spectra of Fig. 3(a)-(c) correspond to the special case (1.94 × 1018 𝐻𝑧) of
Fig. 3(d)-(f), which are marked by the white dashed lines.
Fig.3 (a), (b), (c) The reflection angle spectra of the separate structructures 𝑃𝐶𝐴, 𝑃𝐶𝐵 and the
composite structure 𝑃𝐶𝐴 − 𝑃𝐶𝐵 , respectively.(d), (e), (f) The reflectance for the different
frequency and the grazing incident angle. The white dashed lines marked in the picture indicate
the position of 1.94 × 1018 𝐻𝑧.
9
III. EXPERIMENTAL VERIFICATION OF THE TOPOLOGICALLY PROTECTED
EDGE STATE
In this section, we experimentally demonstrate our theoretical design of topological edge
state in X-ray by fabricating and measuring the corresponding 1-D PC structures. The 1-D PCs
with and without disorders are fabricated by direct current magnetron sputtering technique [49,
50], owing to the extremely small layer thickness. High purity tungsten (99.95%) and carbon
(99.999%) are used as sputtering targets. The base pressure before deposition was ~2×10-4 Pa and
the sputtering gas is argon with the pressure of 0.13 Pa. The deposition rate of W and C are
controlled to be very small, 0.07 nm/s and 0.03 nm/s, in order to grow the structure accurately.
Super-polished silicon wafer is used as the substrates with the root-mean-squared surface
roughness of 0.2 nm. After deposition, the standard 𝑃𝐶𝐴 − 𝑃𝐶𝐵 sample is measured with
transmission electron microscope (TEM) to check the structure of the composite layered system
and the layer growth quality. The X-ray reflectance of all samples are measured by a commercial
X-ray diffractometer (Bede D1) at the Cu-Kα emission line (1.94 × 1018 𝐻𝑧). The measurements
are performed by the standard θ-2θ scan, while θ is the grazing incident angle (angular position of
the sample) and 2θ is the angular position of detector. Here, θ represents the same angle as α
defined above. The angular divergence of the incident beam is 0.007° and the scanning angular
step is Δθ=0.005°.
The TEM image of the fabricated 𝑃𝐶𝐴 − 𝑃𝐶𝐵 sample is shown in Fig. 4(a). The bright layers
are C and the dark layers are W. The transition area between the two PCs is measured and
magnified displayed in Fig. 4(b). The nano-layers are grown with smooth and sharp interfaces
which indicate a good quality of the multilayers. For the PCA structure ((CWC)10), the carbon
10
layers from neighboring unit cells grow continuously forming a thicker carbon layer as seen in the
top part of the image. For the PCB structure ((WCW)10), the tungsten layer looks thicker for the
same reason. To further determine the layer thicknesses, line profiles are drawn on the
cross-section as shown in Fig. 4(c). According to the variation of the gray-scale values, the W
layers in each unit cell are estimated to be ~1.7 nm and C layers are ~ 1.4 nm. The thicknesses are
very close to the designed structure. A small inter-diffusion region of 0.4-0.5 nm width is found at
the interfaces between W and C. Figure 4(d) shows the measured reflectance profiles of the 𝑃𝐶𝐴,
𝑃𝐶𝐵, and the composite 𝑃𝐶𝐴 − 𝑃𝐶𝐵 samples, respectively. We can clearly see that separate 𝑃𝐶𝐴 and
𝑃𝐶𝐵 exhibit a typical Bragg peak of the mirrors. However, once the topologically distinct 𝑃𝐶𝐴
and 𝑃𝐶𝐵 are combined together, there is a reflection dip in the X-ray band-gap, which is
predicted by the topological edge state. The experimental results in Fig. 4(d) well coincide with
the theoretical simulation in Fig. 2(c) except that the angular position of the edge state shifts a
little to the small angle side, owing to the slightly larger thickness of the deposited layers. The
agreement between the test and the theory shows that the topological edge state model is
applicable to the X-ray band.
According to the theory described in the previous part II, the topological interface state will
exist as long as the zero-average-effective-mass (𝑚 = 0) condition is satisfied. In the next, keeping
the structure of 𝑃𝐶𝐴 unchanged and adding some kinds of disorders into 𝑃𝐶𝐵, we have revealed
that the X-ray edge state in the hetero-structure is robust against a variety of disorders. At first, for
the thickness disorder of the layers, we have demonstrated that the topological edge is maintained
′
for the weak disorder when 𝑚 = 0 is satisfied. The first structure of 𝑃𝐶𝐵
with weak disordered
is marked as (𝐵1𝐵2𝐵1)1(𝐵1𝐵3𝐵1)1(𝐵1𝐵4𝐵1)2(𝐵1𝐵3𝐵1)1(𝐵1𝐵2𝐵1)2(𝐵1𝐵3𝐵1)1(𝐵1𝐵2𝐵1)1(𝐵1𝐵4𝐵1)1,
11
where the layer 𝐵3 and 𝐵4 are the carbon layer with thickness 2 nm and 1 nm, respectively.
Subscripts represent the number of the unit cell. In this case, the edge mode in the hetero-structure
𝑃𝐶𝐴 − 𝑃𝐶𝐵
′ is maintained as the case without disorders because the effective-mass of 𝑃𝐶𝐵
′ is still
Fig. 4(a). TEM image of the sample. Enlarged picture is shown in the inset picture, here the
interface of 𝑃𝐶𝐴 and 𝑃𝐶𝐵 is marked by the green line. (b). Zoomed-in photo of the dashed
region in (a). (c). Line profile of 𝑃𝐶𝐴 − 𝑃𝐶𝐵 along the red dashed line marked in (b). (d). The
reflection angle spectrums of the separate structructures 𝑃𝐶𝐴, 𝑃𝐶𝐵 and the composite structure
𝑃𝐶𝐴 − 𝑃𝐶𝐵 measured by experiment. The combine of two kinds PCs is marked by the shaded area.
𝑚𝐵 ≈ +2.03 × 108 and 𝑚 = 0, as is shown in Fig. 5(a). The weak disordered samples are
fabricated while part of the carbon layers are intentionally deposited to 2 nm and 1 nm according
12
to the design. The experimental measured reflectance is shown in Fig. 5(b) and the topological
edge mode of the 𝑃𝐶𝐴 − 𝑃𝐶𝐵
′ with 𝑚 = 0 remains almost the same as the standard structure. The
experiment results are in good agreement with the theoretical results. Furthermore, the reflection
peaks near the topological edge state also are maintained for the weak disordered case. This
feature is very useful in the application of X-ray filter. Secondly, in the case of 𝑚 ≠ 0, the edge
mode is significantly distorted as predicted by the theory. Considering the second weak disordered
case
that
the
′′
𝑃𝐶𝐵
is
marked
as
(𝐵1𝐵4𝐵1)2(𝐵1𝐵2𝐵1)1(𝐵1𝐵4𝐵1)2(𝐵1𝐵3𝐵1)1(𝐵1𝐵2𝐵1)2(𝐵1𝐵3𝐵1)1(𝐵1𝐵2𝐵1)1, the X-ray edge state in
the structure 𝑃𝐶𝐴 − 𝑃𝐶𝐵 will be influenced obviously, because the effective-mass of 𝑃𝐶𝐵
′′ is not
equal to 2.03 × 108 and 𝑚 ≠ 0, as are shown in Fig. 5(c) and 5(d). So, we demonstrate
experimentally that the topological edge state is protected by the 𝑚 = 0 condition.
Fig. 5(a), (b) For the weak disorder (thickness disorder) with 𝑚 = 0, the X-ray edge mode is
hardly affected. (c), (d) For the weak disorder (thickness disorder) with 𝑚 ≠ 0, the X-ray edge
mode is greatly influenced.
At last, different from the type of weak disorder added into the structure, in this part, by
keeping the thickness of the layers invariant, the robustness of the topological edge state is further
13
examined by adding strong disorder (positional disorder) in the hetero-structure 𝑃𝐶𝐴 − 𝑃𝐶𝐵
′′′ as
(𝐴1𝐴2𝐴1)2(𝐵1𝐵2𝐵1)2(𝐴1𝐴2𝐴1)3(𝐵1𝐵2𝐵1)2(𝐴1𝐴2𝐴1)1(𝐵1𝐵2𝐵1)4(𝐴1𝐴2𝐴1)1(𝐵1𝐵2𝐵1)2(𝐴1𝐴2𝐴1)3
.
Although this situation corresponds to a strong disorder case, the topological edge is robust
against the perturbation because the condition of 𝑚 = 0 is satisfied, as is shown in Fig. 6.
Different from the weak disordered structure introduced in Figs. 5(a) and 5(b), although the
topological edge state is still maintained in the case of strong disordered structure in Fig. 6, the
reflection peaks near the edge state are obviously suppressed. The whole reflectance curve is very
close to the designed result of the corresponding structure which implies an accurate fabrication
process. The edge mode remains almost at the same position as the standard structure and the
small shift is caused by the little thickness deviation of the composing unit cells between the two
samples. It is demonstrated again that, even for strong disorders, the 𝑚 = 0 condition can still
protect the edge mode against randomness, which is very useful in the applications of X-rays.
Fig. 6 The robustness of X-ray edge mode against the strong disorder with
has been
demonstrated by simulation (a) and experiment (b), respectively.
14
_0m
In discussion, we firstly extend the concept of topology to the X-ray band. The distinct
topological properties of two different PCs have been revealed from the effective parameters and
the reflection phases. The predicted topologically protected edge states at the interface of two PCs
with different topological properties also have been demonstrated. When a variety of disorders are
introduced into the system, the topological edge state can always be maintained as long as the
m = 0 condition is satisfied. All the results have been verified by the experiments that are in good
agreement with the theory. Our results will provide new theoretical tools to study X-ray optics and
may provide a possibility for the development of new X-ray devices.
IV. CONCLUTION
In summary, topological edge state in X-ray band has been theoretically proposed and
experimentally demonstrated for the first time based on two 1-D PCs with different topological
properties. This edge state is robust against a variety of perturbations, such as thickness disorder
(weak disorder) and positional disorder (strong disorder), as long as the
condition is
satisfied. Our system provides a photonic platform for exploring X-ray propagations and
developing possible topological devices. Our results may promote the development of X-ray
optics by introducing the new topological freedom. Although our results are based on 1-D system,
the concept of topological manipulation might be consulted for higher dimensional system and
related X-ray devices. Therefore, our investigation not only extends the range of applicability of
the photonic topological theory, but also provides insightful guidance to exploring the exciting
applications associated with topological transport to higher spectral range.
15
_0m
APPENDIX
For comparison, different geometrical structures with and without disorders are given in
Table 1. The materials and thicknesses of different layers are listed in Table 2. At last, Table 3
shows the relationship between the
condition and the topologically protected edge state
immune to the disorders.
Table 1. Different structures
Table 2. Materials and the thicknesses of different layers
Table 3. Relationship between the
condition and the topologically protected edge
state
16
_0m_0m
ACKNOWLEDGMENT
This work is supported by the National Key R&D Program (No. 2016YFA0301101,
2016YFA0401304, 2017YFA0403302); by the National Nature Science Foundation of China
(NSFC) (Grant Nos. 11774261, 11474220, 61621001, 11505129, and U1732268); the National
Natural Science Foundation of China Academy of Engi-neering Physics (NSAF) (No. U1430131),
Science Foundation of Shanghai (No. 17ZR1443800); We thank Kun Yu, Feng Wu, and Yang
Yang for helpful discussions.
References
1. C. L. Kane and E. J. Mele, Quantum spin Hall effect in graphene, Phys. Rev. Lett. 95,
226801 (2005).
2. B. A. Bernevig and S. C. Zhang, Quantum spin Hall effect, Phys. Rev. Lett. 96, 106802
(2006).
3. F. D. M. Haldane and S. Raghu, Possible realization of directional optical waveguides in
photonic crystals with broken time-reversal symmetry, Phys. Rev. Lett. 100, 013904 (2008).
4. J. Y. Guo, Y. Sun, Y. W. Zhang, H. Q. Li, H. T. Jiang, and H. Chen, Experimental
investigation of interface states in photonic crystal heterostructures, Phys. Rev. E 78, 026607.
(2008).
5. Z. Wang, Y. D. Chong, J. D. Joannopoulos, and M. Soljačić, Observation of unidirectional
backscattering-immune topological electromagnetic states, Nature 461, 772-775 (2009).
6. Y. Poo, R. X. Wu, Z. F. Lin, Y. Yang, and C. T. Chan, Experimental realization of
self-guiding unidirectional electromagnetic edge states, Phys. Rev. Lett. 106, 093903 (2011).
7. A. B. Khanikaev, S. H. Mousavi, W. K. Tse, M. Kargarian, A. H. MacDonald, and G.
Shvets, Photonic topological insulators, Nat. Mater. 12, 233-239 (2013).
8. M. C. Rechtsman, J. M. Zeuner, Y. Rlotnik, Y. Lumer, D. Podolsky, F. Dreisow, S. Nolte,
M. Segev, and A. Szameit, Photonic Floquet topological insulators, Nature 496, 196-200
(2013).
9. M. Hafezi, S. Mittal, J. Fan, A. Migdall, and J. M. Taylor, Imaging topological edge states
in silicon photonics, Nat. Photon. 7, 1001 (2013).
10. G. Q. Liang and Y. D. Chong, Optical Resonator Analog of a two-dimensional
topological insulator, Phys. Rev. Lett. 110, 203904 (2013).
11. L. L. Zhang, X. Z. Chen, B. F. Bai, Q. F. Tan, G. F. Jin, T. Zentgraf, and S. Zhang,
Helicity dependent directional surface plasmon polariton excitation using a metasurface with
interfacial phase discontinuity, Light-Sci. Appl. 2, e70 (2013).
12. L. D. Tzuang, K. Fang, P. Nussenzveig, S. Fan, and M. Lipson, Non-reciprocal phase
shift induced by an effective magnetic flux for light, Nature Photon. 8, 701 (2014).
17
13. W. Tan, Y. Sun, H. Chen, and S. Q. Shen, Photonic simulation of topological excitations
in metamaterials, Sci. Rep. 4, 3842 (2014).
14. M. Xiao, Z. Q. Zhang, and C. T. Chan, Surface impedance and bulk band geometric
phases in one-dimensional systems, Phys. Rev. X 4, 021017 (2014).
15. L. Lu, J. D. Joannopoulos, and M. Soljačić, Topological photonics, Nat. Photon. 8,
821-829 (2014).
16. L. H. Wu and X. Hu, Scheme for achieving a topological photonic crystal by using
dielectric material, Phys. Rev. Lett. 114, 223901 (2015).
17. C. Poli, M. Bellec, U. Kuhl, F. Mortessagne, and H. Schomerus, Selective enhancement
of topologically induced interface states in a dielectric resonator chain, Nat. commun. 6, 6710
(2015).
18. W. L. Gao, F. Z. Fang, Y. M. Liu, and S. Zhang, Chiral surface waves supported by
biaxial hyperbolic metamaterials, Light-Sci. Appl. 4, e328 (2015).
19. F. Gao, Z. Gao, X. Shi, Z. Yang, X. Lin, J. D. Joannopoulos, M. Soljačić, H. Chen, L. Lu,
Y. Chong, and B. Zhang, Probing the limits of topological protection in a designer surface
plasmon structure, Nat. Commun. 7, 11619 (2016).
20. S. Weimann, M. Kremer, Y. Plotnik, Y. Lumer, S. Nolte, K. G. Makris, M. Segev, M. C.
Rechtsman, and A. Szameit, Topologically protected bound states
in photonic
parity–time-symmetric crystals, Nat. Mater. 16, 433-438 (2017).
21. Z. W. Guo, H. T. Jiang, Y. Long, K. Yu, J. Ren, C. H. Xue, and H. Chen, Photonic spin
Hall efect in waveguides composed of two types of single-negative metamaterials, Sci. Rep 7,
7742 (2017).
22. J. W. Dong, X. D. Chen, H. Zhu, Y. Wang, and X. Zhang, Valley photonic crystals for
control of spin and topology, Nat. Mater. 16, 298-302 (2017).
23. A. Dareau, E. Levy, M. B. Aguilera, R. Bouganne, E. Akkermans, F. Gerbier, and J.
Beugnon, Revealing the topology of quasicrystals with a diffraction experiment, Phys. Rev.
Lett. 119, 215304 (2017).
24. C. X. Liu, W. L. Gao, B. Yang, and S. Zhang, Disorder-induced topological state
transition in photonic metamaterials, Phys. Rev. Lett. 119, 183901 (2017).
25. B. Bahari, A. Ndao, F. Vallini, A. E. Amili, Y. Fainman, and B. Kanté, Nonreciprocal
lasing in topological cavities of arbitrary geometries, Science 358, 636–640 (2017).
26. A. B. Khanikaev, R. Fleury, S. H. Mousavi, and A. Alu, Topologically robust sound
propagation in an angular-momentum-biased graphene-like resonator lattice, Nat. commun. 6,
8260 (2015).
27. M. Xiao, G. C. Ma, Z. Y. Yang, P. Sheng, Z. Q. Zhang, and C. T. Chan, Geometric phase
and band inversion in periodic acoustic systems, Nat. Phys. 11, 240-244 (2015).
28. C. He, X. Ni, H. Ge, X. C. Sun, Y. B. Chen, M. H. Lu, X. P. Liu, and Y. F. Chen, Acoustic
topological insulator and robust one-way sound transport, Nat. Phys. 12, 1124-1129 (2016).
29. J. Y. Lu, C. Y. Qiu, L. P. Ye, X. Y. Fan, M. Z. Ke, F. Zhang, and Z. Y. Liu, Observation of
topological valley transport of sound in sonic crystals, Nat. Phys. 13, 369-374 (2017).
30. L. M. Nash, D. Kleckner, A. Read, V. Vitelli, A. M. Turner, and T. M. Irvine, Topological
mechanics of gyroscopic metamaterials, Proc. Natl Acad. Sci. USA 112, 14495–14500 (2015).
31. R. Süsstrunk and S. D. Huber, Observation of phononic helical edge states in a
mechanical topological insulator, Science 349, 47–50 (2015).
18
32. R. Röhlsberger, H. C. Wille, K. Schlage, and B. Sahoo, Electromagnetically induced
transparency with resonant nuclei in a cavity, Nature, 482, 199-203 (2012).
33. X. S. Liu, W. L. Yang, and Z. Liu, Recent progress on synchrotron-based In situ soft
X-ray spectroscopy for energy materials, Adv. Mater. 26, 7710 (2014).
34. F. Calegari, D. Ayuso, A. Trabattoni, L. Belshaw, S. De Camillis, S. Anumula, F. Frassetto,
L. Poletto, A. Palacios, P. Decleva, J. B. Greenwood, F. Martín, M. Nisoli, Ultrafast electron
dynamics in phenylalanine initiated by attosecond pulses, Science, 346, 336-339 (2014).
35. F. Pfeiffer, X-ray ptychography, Nature Photon. 12, 9–17 (2018).
36. Q. S. Huang, V. Medvedev, R. V. D. Kruijs, A. Yakshin, E. Louis, and F. Bijkerk, Spectral
tailoring of nanoscale EUV and soft X-ray multilayer optics, Appl. Phys. Rev. 4, 011104
(2017).
37. H. Mimura, H. Yumoto, S. Matsuyama, T. Koyama, K. Tono, Y. Inubushi, T. Togashi, T.
Sato, J. Kim, R. Fukui, Y. Sano, M. Yabashi, H. Ohashi, T. Ishikawa, and K. Yamauchi,
Generation of 1020 W cm−2 hard X-ray laser pulses with two-stage reflective focusing system,
Nat. commun. 5, 3539 (2014).
38. Q. S. Huang, Q. Yi, Z. D. Cao, R. Qi, R. A. Loch, P. Jonnard, M. Wu, A. Giglia, W. B. Li,
E. Louis, F. Bijkerk, Z. Zhang, and Z. S. Wang, High reflectance nanoscale V/Sc multilayer
for soft X-ray water window region, Sci. Rep. 7, 12929 (2017).
39. A. Rack, T. Weitkamp, M. Riotte, D. Grigoriev, T. Rack, L. Helfen, T. Baumbach, R.
Dietsch, T. Holz, M. Krämer, F. Siewert, M. Meduňa, P. Cloetens, and E. Ziegler,
Comparative study of multilayers used in monochromators for synchrotron-based coherent
hard X-ray imaging, J. Synchrotron. Rad. 17, 496–510 (2010).
40. A. E. Yakshin, I. V. Kozhevnikov, E. Zoethout, E. Louis, and F. Bijkerk, Properties of
broadband depth-graded multilayer mirrors for EUV optical systems, Opt. Express. 18, 6957
(2010).
41. W. Karim, S. A. Tschupp, M. Oezaslan, T. J. Schmidt, J. Gobrecht, J. A. V. Bokhoven,
and Y. Ekinci，High-resolution and large-area nanoparticle arrays using EUV interference
lithography, Nanoscale 7, 7386 (2015)
42. J. Zak, Berry's phase for energy bands in solids, Phys. Rev. Lett. 62, 2747 (1989).
43. M. Atala, M. Aidelsburger, J. T. Barreiro, D. Abanin, T. Kitagawa, E. Demler, and I.
Bloch, Direct measurement of the Zak phase in topological Bloch bands, Nat. Phys. 9,
795-800 (2013).
44. Z. S. Wang, X. B. Cheng, J. T. Zhu, Q. S. Huang, Z. Zhong, and L. Y. Chen, Investigation
of aperiodic W/C multi-layer mirror for X-ray optics, Thin Solid Films 519, 6712-6715
(2011).
45. X. B. Cheng, J. L. Zhang, T. Ding, Z. Y. Wei, H. Q. Li, and Z. S. Wang, The effect of an
electric field on the thermomechanical damage of nodular defects in dielectric multilayer
coatings irradiated by nanosecond laser pulses, Light-Sci. Appl. 2, e80 (2013).
46. D. L. Windt, IMD-software for modeling the optical properties of multilayer films,
Comput. Phys. 12 360-370 (1998).
47. D. R. Smith, W. J. Padilla, D. C. Vier, S. C. N. Nasser, and S. Schultz, Composite
medium with simultaneously negative permeability and permittivity, Phys. Rev. Lett. 84, 4184
(2000).
48. S. Q. Shen, Topological insulators: Dirac equation in condensed matters (Springer-Verlag,
19
Berlin, 2012).
49. E. Louis, A. E. Yakshin, T. Tsarfati, and F. Bijkerk, Nanometer interface and materials
control for multilayer EUV-optical applications, Progress in Surface Science 86 255–294
(2011).
50. Q. S. Huang, J. Fei, Y. Liu, P. Li, M. W. Wen, C. Xie, P. Jonnard, A. Giglia, Z. Zhang,
Kun Wang, and Z. S. Wang, High reflectance Cr/V multilayer with B4C barrier layer for water
window wavelength region, Opt. Lett. 41, 701 (2016).
20
|
1806.01978 | 1 | 1806 | 2018-06-06T02:11:20 | Control of magnetization dynamics by spin Nernst torque | [
"physics.app-ph",
"cond-mat.mes-hall"
] | Control of magnetization dynamics is one of the primary goals in spintronics. It has been demonstrated using spin Hall effect i.e charge current to spin current conversion in non-magnetic metal which has large spin-orbit coupling such as Pt, W etc. Recently different groups have shown generation of spin current in Pt, W while thermal gradient is created by virtue of spin Nernst effect. In this work we show the evidence of magnetization control by spin Nernst torque in Pt/Py bi-layer. We compared relative strength of spin Nernst Torque and spin Hall torque by measuring the systematic variation of magnetic linewidth on application of constant heat or charge current. Spin-torque ferromagnetic resonance (ST-FMR) technique is adopted to excite the magnet and to measure line-width precisely from the symmetric and anti-symmetric voltage component. Control of magnetization dynamics by spin Nernst torque will emerge as an alternative way to manipulate nano-magnets. | physics.app-ph | physics | Control of magnetization dynamics by spin Nernst torque
Arnab Bose, Ambika Shankar Shukla, Sutapa Dutta, Swapnil Bhuktare, Hanuman Singh, Ashwin A.
Tulapurkar
Department of Electrical Engineering, Indian Institute of Technology-Bombay, Powai, Mumbai, India 400076
Relativistic interaction between electron's spin and orbital angular momentum has provided efficient
mechanism to control magnetization of nano-magnets. Extensive research has been done to
understand and improve spin-orbit interaction driven torques generated by non-magnets while
applying electric current. In this work, we show that heat current in non-magnet can also couple to
its spin-orbit interaction to produce torque on adjacent ferromagnet. Hence, this work provides a
platform to study spin-orbito-caloritronic effects in heavy metal/ferromagnet bi-layers.
Since last few years considerable attention has been drawn to control the magnetization
dynamics by pure spin current generated by spin Hall effect (SHE)1,2,3,4 and interfacial magnetic
fields5,6 by Rashba effect. SHE and Rashba effects are relativistic phenomena which couple electron's
spin and orbital motion and can be used to exert spin-orbit torques7,8. On the other hand, thermal
gradient in ferromagnet can also create pure spin current9,10,11,12,13 which can further produce thermal
spin torques14,15,16,17,18 and domain wall motion19,20. Conversion of heat current into spin current in a
non-magnet has been shown recently via the spin Nernst effect (SNE)21,22,23,24. But an important
question remains unanswered whether thermal gradient in non-magnet can generate spin toque owing
to its spin-orbit coupling, which in turn could be used for manipulating magnetization. In this letter
we demonstrate that, interplay of heat current and spin-orbit coupling in non-magnetic Platinum (Pt)
can generate thermally driven spin-orbit torque (equivalent to spin Nernst torque (SNT)). SNT has
been predicted recently25,26 but it is lacking the experimental evidence. Here, we show that effective
magnetic damping can be controlled by SNT while creating thermal gradient in Pt/Ni81Fe19 bilayer.
This can open a new avenue to manipulate spins in magnetic nanostructures for technological
applications27,28,29.
Fig. 1. Schematic representation of the experiment. (a) Schematic diagram of Pt/Py
bilayer with direction of spin separation due to SHE or SNE. (b,c) Change of linewidth (or
effective damping) on application of dc spin current (generated by SHE or SNE). (d)
Direction of anti-damping and damping like torques. (e) Coloured SEM image of the
fabricated device with current and voltage terminals.
1
While thermal gradient is established in heavy metal (Pt in this work), spins of opposite
polarity separate out in a direction orthogonal to the direction of heat flow due to SNE as shown in
Fig. 1(a). This situation is thermal analogous to SHE. Thus Pt converts heat current into pure spin
current which is then injected into neighbouring ferromagnet (FM). If the injected spin current
density is enough, spin torque is expected on the FM causing enhanced or reduced damping
depending upon the direction of spin vectors absorbed by FM (Fig 1(b-c)). We compare the change
in resonance linewidth of FM due to the spin torque generated by SNE and SHE by performing spin-
torque ferromagnetic resonance (ST-FMR) experiment2,4,6,30,31,32. Basic working principle of ST-
FMR is the following. Radio frequency (rf) current is applied along X axis and dc voltage is measured
in the same terminals using a bias-T (AMR based detection of ST-FMR2,32) while external magnetic
field is swept at angle θ with respect to X-axis (Fig. 1(e)). Pt converts rf charge current into rf spin
current which is injected to ferromagnet (Ni81Fe19: Py here after). Radio frequency spin current and
current induced rf fields excite the magnet to undergo small oscillation around its equilibrium
position. Due to the AMR effect, resistance of the magnet (hence FM/HM stack) also oscillates.
Homodyne mixture of RF applied current and RF resistance of the sample produces dc voltage.
Advantage of ST-FMR is that at resonance large dc voltage can be obtained. This dc voltage is
typically combination of symmetric Lorentzian (VS) and anti-symmetric Lorentzian (VA).
and
where C1 and C2 are the amplitude of VS
and VA respectively, H is externally applied field, H0 is resonant field position and ∆ is the resonance
linewidth (FWHM). VS indicates the contribution of spin current induced torque and VA indicates in-
plane field induced torques. In Pt/Py bi-layers, the Oersted magnetic field is the dominant source of
in-plane field.2,32. So charge to spin current conversion efficiency (or effective spin Hall angle) in
HM can be quantified from C1/C2 ratio as following:
where
MS is saturation magnetization, tPt, tPy are thickness of Pt and Py film respectively,
is
perpendicular magnetic anisotropy field. Now, if dc current is superimposed on the RF current then
non-zero dc spin current is injected which can change the resonance linewidth of Py2,3,32. It also
provides direct quantification of effective spin Hall angle as following:
where JC is charge current density through Pt, f
is frequency of applied rf current, γ is gyromagnetic ratio. In this work, we shave superimposed a dc
heat current on rf charge current. We could modulate the resonance line width of Py which provides
a direct evidence of control of magnetization dynamics by spin Nernst torque (SNT).
We fabricated the device as shown in Fig. 1(e). The crossbar is made of Pt (15 nm) and a
rectangular shaped dot of Py (2 nm) is deposited at the centre of Pt crossbar. Top of Py was capped
with Ta (1.5 nm). On the top and bottom lead of the Pt crossbar, two heater lines are fabricated which
are electrically isolated by SiOx (30 nm) from Pt. Heater lines are prepared with Ta/Pt (60 nm).
Numbers in bracket indicate thickness of metals. Entire fabrication is done by standard electron beam
lithography, sputtering and lift-off technique. Before deposition of Py, surface of Pt was cleaned by
Ar ion without breaking the vacuum. We have earlier shown that linewidth change can be sensitively
measured in planar Hall structure doing ST-FMR [Ref. 32]. We follow the same approach in this
work to compare the strength of spin Hall torque (SHT) and spin Nernst torque (SNT). Our detection
2
()212204SVCHH=−+()()0222044AHHVCHH−=−+()1/20'121/2SPtPySHOeMttCHHC⊥=+H⊥0'21cos42SPySHcSlopeMtedHHfdJ⊥=+ method is the following. Radio frequency current is applied along X-axis; dc voltage is measured
along same direction (hence AMR based detection) but dc heat current (or charge current) is passed
along Y-axis to modulate the line width by SNT (or SHT) as shown in Fig. 1(e). Application of dc
current (heat current or charge current) perpendicular to the direction of voltage measurement reduces
noise and hence measurement sensitivity significantly increases as shown in Ref 32.
Fig. 2. Characterization of spin Hall torque. (a) DC voltage generated by ST-FMR for
different frequencies of applied rf current. (b) Fit of dc voltage by VS and VA. Inset of (b)
shows Kittel's fit. (c) ST-FMR spectrum when dc charge current is applied perpendicular to
applied rf current. (d) Modulation of linewidth on application of dc charge current.
Fig. 2 shows the characterization of spin-Hall torque by measuring ST-FMR. Fig. 2(a) shows
the typical dc voltage spectrum as external magnetic field is swept for different frequencies of applied
current. This dc voltage can be fitted to the sum of VS and VA (Fig. 2(b)). Red squares in Fig. 2(b)
show the experimental data and black curve shows fitting which is sum of VS (pink curve) and VA
(blue curve). The symmetric component confirms the spin-orbit torque generated by spin Hall effect.
Inset of Fig. 2(b) shows the Kittel's fit for resonant magnetic fields and frequencies. From this we
=8.05 kOe (which corresponds to Ms = 6.5×105 A/m.) Fig. 2(c) shows the voltage spectrum
obtain
when dc charge current is applied orthogonal to the direction of rf current flow. We can clearly see
the dominant change in the shape of the voltage signal as the linewidth significantly changes. For the
positive current linewidth is more in positive field values and it is less in negative field values (vice
versa for negative applied current). Our detection method is so sensitive that linewidth change is
clearly visible in Fig. 2(c) itself. Linewidth (∆) as a function of applied dc current is shown in top
panel of Fig. 2(d). It shows expected linear dependence as function of applied current. Bottom panel
of Fig. 2(d) shows the difference in linewidth for θ=350 and θ=2150 (∆'=∆(350)-∆(2150)) as a function
of applied current which also shows linear dependence. Bottom panel of Fig. 2(d) represents only the
contribution of spin current induced damping change eliminating the overall heating effect if any.
From this measurement of linewidth modulation, we can extract the effective spin Hall angle of Pt to
be 0.12±0.06. Extracted value of spin Hall angle from C1/C2 ratio is somewhat lower. We have further
3
H⊥ noticed that in this planar Hall geometry when dc current is applied perpendicular to rf current
modulation in VS is higher compared to the line-width modulation as a function of applied dc current.
Same behaviour was also observed in our previous work (Fig. 3(c) of ref 32 and Fig. 2(c) in this
work). However, we have verified that the line width modulation is the same irrespective of the
detection methods (AMR or PHE based detection or hybrid planar Hall detection method). So in this
work we quantify the strength of SHT and SNT in Pt by measuring linewidth change by injecting dc
spin current by SHE and SNE respectively.
Fig. 3. Characterization of SNT. (a) dc voltage spectrum on application of heat current
along +Y axis and –Y axis. (b) dc voltage spectrum when positive and negative field data
points are superimposed for dc thermal gradient along Y axis. (c,d) Difference in linewidth
between positive and negative field values (∆') for different applied heater powers and
different angles (θ) (respectively).
Now we show the evidence of spin Nernst torque by passing dc heat current instead of
applying dc charge current while doing ST-FMR as discussed above. Two different heater lines are
fabricated on Hall bar (Fig. 1(e)) to create thermal gradient along ±Y-axis. When current flows in
heaterline-1 (HL1) it becomes hot due to Joule's heating and most of the heat is carried by Pt below
the heat line. This creates thermal gradient in Pt/Py bilayers along +Y axis. Similarly, when current
is applied in heaterline-2 (HL2) thermal gradient is created along -Y axis. Fig. 3(a) shows the dc
voltage spectrum generated by ST-FMR while thermal gradient is created along ±Y axis. We can see
the difference in dc voltage spectrum for positive and negative thermal gradients. This cannot be
explained by overall heating effect since overall heating would be same for both the direction of
thermal gradients. All these measurements are performed when θ is 350. In Fig. 3(b) we further show
that when voltage signal of positive field and negative field is superimposed, there is distinct change
in the shape of voltage spectrum (hence the linewidth) which further confirms the existence of SNT.
Difference in the shape of voltage signal in Fig. 3(b) cannot be explained by overall heating effect as
it would be same for both 350 and 2150. As shown earlier (Fig. 2(d)), line-width difference for positive
and negative field (∆'=∆(θ) - ∆(θ+1800)) is measured for different applied heater powers (Fig. 3(c))
and for different angles (Fig. 3(d)). ∆' is proportional to the heater power (Fig. 3(c)) and it closely
follows cosθ dependence (Fig. 3(d)). In our geometry (Fig. 1(a)) the line width modulation due to
SNT is expected to be maximum at θ=00, but ST-FMR signal becomes zero at θ=00. We further
confirmed that polarity of heater current has negligible effect on ∆' as heater line is electrically
insulated from Pt hall bar and Py dot is quite far away (1.5 μm) from the heater line to get affected
by magnetic (Oersted) field produced by the heater current. In our control experiment we have
applied heater current in both HL1 and HL2 which would cause the same overall heating but fail to
set up well directed thermal gradient. We found negligible effect on the line width in this case. Our
observed results shown in Fig. 3 strongly supports the evidence of spin Nernst torque in Pt.
4
Fig. 4. Temperature profile. (a) Surface plot of 2D thermal gradient. Estimated temperature
(b) and temperature gradient (c) in Pt/Py interface.
We have followed the same approach to find thermal gradient in Pt as reported in Ref 24.
From the resistance value of the heater line, its overall temperature is known (on chip temperature
calibration). Once the temperature of heater line is known temperature profile of Pt/Py interface can
be calculated from COMSOL simulation (Fig. 4(b) and 4(c)). Overall temperature rise is around 25
K which is fairly small and hence we observe negligible contribution from overall heating effect. We
consider the thermal gradient at the interface to be 15 K/μm while estimating SNT. Our estimated
temperature gradient in this kind of geometry is in good agreement with previous results20.
Comparing the line-width modulation by SHT and SNT we can quantify that 15 K/ μm horizontal
thermal gradient in Pt/Py interface is equivalent to the application of 4.9×109 A/m2 amount of charge
current density in Pt. This reported value of heat current to spin current conversion efficiency by SNE
is consistent with our previous report of SNE24 and comparable to reports by other groups21,23.
In conclusion, we have demonstrated the control of magnetization dynamics by thermally
driven spin Hall torque (or spin Nernst torque). We report approximately 0.9 % line-width change
due to spin Nernst torque effect. It indicates that about 100 times more thermal gradient needs be
created to achieve switching by spin Nernst torque which could be achieved in material having large
spin Nernst angle and implementing efficient mechanism (such as Laser heating) to create large
thermal gradient at nano scale. Further by using ferromagnets with less out-of-plane anisotropy, the
thermal gradients required for switching can be reduced. These results will beneficial in energy
harvesting sector where heat current in non-magnets can be utilized to write magnetic memories for
non-volatile memory applications.
References
1 J. Sinova, S. O. Valenzuela, J. Wunderlich, C. H. Back, and T. Jungwirth, "Spin Hall effects". Rev. Mod.
Phys. 87, 1213 (2015)
2 L. Q. Liu, T. Moriyama, D. C. Ralph, and R. A. Buhrman, "Spin-Torque Ferromagnetic Resonance
Induced by the Spin Hall Effect", Phys. Rev. Lett. 106, 036601 (2011)
3 K. Ando, S. Takahashi, K. Harii, K. Sasage, J. Ieda, S. Maekawa, and E. Saitoh, "Electric Manipulation of
Spin Relaxation Using the Spin Hall Effect", Phys. Rev. Lett. 101, 036601 (2008)
4 L. Liu, Chi-Feng Pai, D. C. Ralph, and R. A. Buhrman, Magnetic Oscillations Driven by the Spin Hall
Effect in 3-Terminal Magnetic Tunnel Junction Devices, Phys. Rev. Lett. 109, 186602 (2012).
5 I. M . Miron, G. Gaudin, S. Auffret, B. Rodmacq,A. Schuhl, S. Pizzini, J. Vogel, P. Gambardella.
"Current-driven spin torque induced by the Rashba effect ina ferromagnetic metal layer", Nat. Mater. 9, 230
(2010)
6 A. Bose, H. Singh, V. K. Kushwaha, S. Bhuktare, S. Dutta, and A. A. Tulapurkar. "Sign Reversal of Field-
like Spin-Orbit Torque in an Ultrathin Cr/Ni Bilayer", Phys. Rev. Appl. 9, 014022 (2018)
5
7 Anjan Soumyanarayanan, Nicolas Reyren, Albert Fert, and Christos Panagopoulos, "Emergent phenomena
induced by spin-orbit coupling at surfaces, and interfaces", Nature (London) 539, 509 (2016)
8 A. Manchon, H. C. Koo, J. Nitta, S. M. Frolov, and R. A. Duine, "New perspectives for Rashba spin orbit
coupling", Nat. Mater. 14, 871 (2015)
9 K. Uchida, H. Adachi, T. Ota, H. Nakayama, S. Maekawa, and E. Saitoh. "Observation of longitudinal
spin-Seebeck effect in magnetic insulators", Appl. Phys. Lett. 97, 172505 (2010)
10 G. E. W. Bauer, E. Saitoh and B. J. Van Wees. "Spin Caloritronics", Nat. Mater. 11, 391 (2012)
11 M. Walter, J. Walowski, V. Zbarsky, M. M€ unzenberg, M. Sch€afers, D. Ebke, G. Reiss, A. Thomas, P.
Peretzki, M. Seibt et al., "Seebeck effect in magnetic tunnel junctions," Nat. Mater. 10, 742–746 (2011)
12 S. Jain, D. D. Lam, A. Bose, H. Sharma, V. R. Palkar, C. V. Tomy, Y.Suzuki, and A. A. Tulapurkar,
"Magneto-Seebeck effect in spin-valve with heat current in-plane thermal gradient", AIP Adv. 4, 127145
(2014)
13 M. Schmid, S. Srichandan, D. Meier, T. Kuschel, J.-M. Schmalhorst, M. Vogel, G. Reiss, C. Strunk and
C. H. Back. "Transverse Spin Seebeck Effect versus Anomalous and Planar Nernst Effects
in Permalloy Thin Films", Phys. Rev. Lett. 111, 187201 (2013)
14 H. Yu, S. Granville, D. P. Yu, and Ph. J. Ansermet, "Evidence for thermal spin-transfer torque", Phys.
Rev. Lett. 104, 146601 (2010)
15 M. G. Choi, H. C. Moon, C. B. Min, J. K. Lee, and G. D. Cahill, "Thermal spin-transfer torque driven by
the spin-dependent Seebeck effect in metallic spin-valves", Nat. Phys. 11, 576–582 (2015).
16 A. Pushp, T. Phung, C. Rettner, B. P. Hughes, S. H. Yang, and S. S. P. Parkin, "Giant thermal spin
torque-assisted magnetic tunnel junction switching," Proc. Natl. Acad. Sci. U.S.A. 112, 6585–6590 (2015).
17 A. Bose et al., "Observation of thermally driven field-like spin torque in magnetic tunnel junctions", Appl.
Phys. Lett. 109, 032406 (2016)
18 M. B. Jungfleisch, T. An, K. Ando, Y. Kajiwara, K. Uchida, V. I. Vasyuchka, A. V. Chumak, A. A. Serga,
E. Saitoh and B. Hillebrands. "Heat-induced damping modification in yttrium iron garnet/platinum hetero-
structures", Appl. Phys. Lett. 102, 062417 (2013)
19 W. Jiang, P. Upadhyaya, Y. Fan, J. Zhao, M. Wang, L. Chang, M. Lang, K. L. Wong, M. Lewis, Y.-T. Lin
et. al. "Direct Imaging of Thermally Driven Domain Wall Motion in Magnetic Insulators", Phys. Rev. Lett.
110, 177202 (2013)
20 P. Krzysteczko, X. Hu, N. Liebing, S. Sievers, and H. W. Schumacher. "Domain wall magneto-Seebeck
effect", Phys. Rev. B 92, 140405(R) (2015)
21 S. Meyer, Y.-T. Chen, S. Wimmer, M. Althammer, T. Wimmer, R. Schlitz, S. Geprägs, H. Huebl, D.
Ködderitzsch, H. Ebert, G.E.W. Bauer, R. Gross, and S.T.B. Goennenwein, "Spin Nernst effect", Nat. Mater.
16, 977 (2017)
22 P. Sheng, Y. Sakuraba, Y. Lau, S. Takahashi, S. Mitani, and M. Hayashi, "The spin Nernst effect in
tungsten", Sci. Adv. 3, e1701503 (2017)
23 D.J. Kim, C.Y. Jeon, J. G. Choi, J.W. Lee, S. Surabhi, J.R. Jeong, K.J. Lee, and B.G. Park, Nat.
"Observation of transverse spin Nernst magnetoresistance induced by thermal spin current in
ferromagnet/non-magnet bilayers", Nat. Commun. 8, 1 (2017)
24 A. Bose, S. Bhuktare, H. Singh, S. Dutta, V. Achanta, A. A. Tulapurkar, "Direct detection of spin Nernst
effect in Pt", Appl. Phys. Lett. 112, 162401 (2018)
25 G. Geranton, F. Freimuth, S. Blugel, and Y. Mokrousov. "Spin-orbit torques in L10-FePt/Pt thin films
driven by electrical and thermal currents", Phys. Rev. B 91, 014417 (2015)
26 A. A. Kovalev and V. Zyuzin. "Spin torque and Nernst effects in Dzyaloshinskii-Moriya ferromagnets",
Phys. Rev. B 93, 161106(R) (2016).
27 S. Bhuktare, H. Singh, A. Bose, and A. A. Tulapurkar, "Spintronic Oscillator Based on Spin-Current
Feedback Using the Spin Hall Effect", Phys. Rev. Appl. 7, 014022 (2017)
28 Swapnil Bhuktare, Arnab Bose, Hanuman Singh & Ashwin A. Tulapurkar, "Gyrator Based on Magneto-
elastic Coupling at a Ferromagnetic/Piezoelectric Interface", Scientific Reports 7: 840
DOI:10.1038/s41598-017-00960-9
29 H. Singh et. al. "Integer, Fractional, and Sideband Injection Locking of a Spintronic Feedback Nano-
Oscillator to a Microwave Signal", Phys. Rev. Appl. 8, 064011 (2017)
30 A. A. Tulapurkar, Y. Suzuki, A. Fukushima, H. Kubota, H. Maehara, K. Tsunekawa, D. D. Djayaprawira,
N. Watanabe, and S. Yuasa, "Spin Torque Diode effect in Magnetic Tunnel Junctions", Nature (London)
438, 339 (2005)
6
31 A. Bose, S. Dutta, S. Bhuktare, H. Singh, and A. A. Tulapurkar, "Sensitive measurement of spin orbit
torque driven ferromagnetic resonance detected by planar Hall geometry", Appl. Phys. Lett. 111, 162405
(2017)
32 A. Bose, D. D. Lam, S. Bhuktare, H. Singh, S. Miwa and A. Tulapurkar, "Observation of anomalous spin-
torque generated by a ferromagnet", arXiv:1706.07245 (2017)
7
Supplementary information
Arnab Bose, Ambika Shukla, Sutapa Dutta, Swapnil Bhuktare, Hanuman Singh, Ashwin A
Tulapurkar
Department of Electrical Engineering, Indian Institute of Technology-Bombay, Powai, Mumbai,
India 400076
S1. Temperature calibration
Thermal gradient is obtained by doing on-chip calibration. As current is passed through heater line it becomes
hot due to Joule's heating and its resistance increases. Hence resistance of heater line provides information of
average temperature in heater line. Same approach was adopted in our previous work [S1] and by other
researchers [S2,S3,S4]. Heater line was made of Ta/Pt (~50 nm). Its resistance changes from 50 Ω to 65 Ω on
application of maximum heater power (0.3 W). This corresponds to overall temperature of heater line to be
approximately 420 K which is obtained from the temperature dependent resistance measurement of the heater
line. We observed minor change in resistance in Pt Hall bar which contains rectangular Py dot at the centre. It
indicates that the temperature of horizontal Pt line does not increase much and the heat is locally centred near
the heater line. To get exact temperature profile, COMSOL simulation is done with heat transfer module with
following boundary conditions: (1) temperature of heater line is 420 K (experimentally obtained) and (2)
temperature of the bottom of Si is 293 (lab temperature).
Fig. S1. (a) Temperature profile at the interface of Pt/Py. (b) Isothermal temperature contour. (c)
Schematics of device structure. Arrow indicates direction of heat flow. (d) Temperature along Y-axis.
(e) Temperature gradient along Y-axis.
Fig. S1(a) shows the surface temperature. Bright yellow colour indicates hot region and temperature of heater
is set 420 K. Deep red indicates the colder region. Fig. S2(b) shows the isothermal contour which shows that
8
heat is also localized in very small region (approx. 10×10×10 μm3) as expected. Fig. S1(d) shows the
temperature as along Y axis which shows exponential decay of thermal gradient (Fig. S1(e)). Temperature
becomes around 300 K when we go 10 μm away from the heater line. It shows that overall temperature of Py
is slightly more than room temperature (~325 K). Maximum thermal gradient along Y axis in Pt/Py interface
is nearly 20 K/μm. Estimated temperature profile is in good agreement with previous works [S1-S6]. We have
considered below parameters in simulation. We have also checked that slight variation of these parameter does
not influence the simulated result much. Finally estimated thermal gradient will be in order tens of K/μm in
this kind of geometry.
Material
Thermal conductivity (SI unit)
Pt
85
Si
120
SiO2
1.4
References
[S1] A. Bose, S. Bhuktare, H. Singh, S. Dutta, V. Achanta, A. A. Tulapurkar. Appl. Phys. Lett. (accepted)
[S2] P. Krzysteczko, X. Hu, N. Liebing, S. Sievers, and H. W. Schumacher. Domain wall magneto-Seebeck
effect. Phys. Rev. B 92, 140405(R) (2015)
[S3] S. Meyer, Y.-T. Chen, S. Wimmer, M. Althammer, T. Wimmer, R. Schlitz, S. Geprägs, H. Huebl, D.
Ködderitzsch, H. Ebert, G.E.W. Bauer, R. Gross, and S.T.B. Goennenwein, Nat. Mater. 16, 977 (2017)
[S4] A. Pushp, T. Phung, C. Rettner, B. P. Hughes, S. H. Yang, and S. S. P. Parkin, "Giant thermal spin torque-
assisted magnetic tunnel junction switching," Proc. Natl. Acad. Sci. U.S.A. 112, 6585–6590 (2015).
[S5] A. Slachter, L. F. Bekker, J.-P. Adam, and J. B. van Wees, "Thermally driven spin injection from a
ferromagnet into a non-magnetic metal," Nat. Phys. 6, 879–882 (2010)
[S6] M. Walter, J. Walowski, V. Zbarsky, M. M€ unzenberg, M. Sch€afers, D. Ebke, G. Reiss, A. Thomas, P.
Peretzki, M. Seibt et al., "Seebeck effect in magnetic tunnel junctions," Nat. Mater. 10, 742–746 (2011)
9
|
1710.09098 | 1 | 1710 | 2017-10-25T07:47:25 | Transformable Soft Quantum Device based on Liquid Metals with Sandwiched Liquid Junctions | [
"physics.app-ph"
] | Quantum tunneling effect has been an important issue in both fundamental science and practical applications. Classical quantum tunneling devices are generally rigid in structures which may encounter technical difficulties during fabrication, functional tuning and shape adapting. Here through introducing the room-temperature liquid metals as two conductive electrodes and a soft even liquid insulating layer sandwiched between them, we proposed a conceptually innovative all-soft or liquid quantum device which would help realize a couple of unconventional quantum capabilities such as flexibility, deformability, transformability, and reconfigurability, etc. which may not be easily offered by a rigid quantum device. Representative structural configurations to make such transformable quantum devices via sandwiching various liquid metal-insulating layer-liquid metal (LM-IL-LM) components are suggested. The feasibility for making such an all-soft quantum device is interpreted through experimental evidences and theoretical evaluations. Potential future applications of the proposed devices in a group of emerging fields including intelligent quantum systems and quantum computing are prospected. | physics.app-ph | physics | Transformable Soft Quantum Device based on Liquid Metals with
Sandwiched Liquid Junctions
Xi Zhao,1,2 Jianbo Tang,3 Yongze Yu,1 and Jing Liu,1,2,3,4*
1Technical Institute of Physics and Chemistry, Chinese Academy of Sciences, Beijing 100190, China.
2School of Future Technology, University of Chinese Academy of Sciences, Beijing 100049, China.
3Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing 100084, China.
4Institute for Advanced Study on Liquid Metal, Yunnan University, Yunnan Province 650504, China
*Corresponding author. Email: [email protected].
Abstract: Quantum tunneling effect has been an important issue in both fundamental
science and practical applications. Classical quantum tunneling devices are generally
rigid in structures which may encounter technical difficulties during fabrication,
functional tuning and shape adapting. Here through introducing the room-temperature
liquid metals as two conductive electrodes and a soft even liquid insulating layer
sandwiched between them, we proposed a conceptually innovative all-soft or liquid
quantum device which would help realize a couple of unconventional quantum
capabilities such as flexibility, deformability, transformability, and reconfigurability,
etc. which may not be easily offered by a rigid quantum device. Representative
structural configurations
transformable quantum devices via
sandwiching various
(LM-IL-LM)
components are suggested. The feasibility for making such an all-soft quantum device
is interpreted through experimental evidences and theoretical evaluations. Potential
future applications of the proposed devices in a group of emerging fields including
intelligent quantum systems and quantum computing are prospected.
Keywords: Quantum tunneling effect; All-liquid quantum device (A-LQD);
liquid metal-insulating
to make such
layer-liquid metal
Liquid metal; Transformable soft quantum device.
1. Introduction
It is a well-known fact that if there exists an insulating layer between two
conductors (or semiconductors and superconductors), electrons are not able to pass
through the layer from one side to the other. The insulating layer here serves as a
barrier, or called potential junction, to electrons. However, when the thickness of such
insulating layer becomes comparable to the de Broglie wavelength, electrons can pass
through the barrier [1]. This special phenomenon is called quantum tunneling effect.
In fact, the quantum tunneling effect plays an essential role in numerous fundamental
physical phenomena such as the nuclear fusion taking place in the sun [2] and the
biological photosynthesis process [3]. Moreover, quantum tunneling theory is also
widely used in many fields [4, 5] such as scanning tunneling microscopy, optics,
semiconductor physics and superconductor physics (Fig. 1) etc.
1
Fig. 1. Typical situations where quantum tunneling effect plays important role.
So far, nearly all the quantum tunneling devices are composed of a rigid
sandwich structure with a thin insulating layer (0.1~10 nm thickness) inserted
between two electrodes on its both sides [6]. Since these components are all solid, the
device structure will not be readily adjusted. Thus the entire device cannot be
reshaped, split or reassembled either. Once fabricated, such devices would only be
able to perform certain specified functions restricted by its predefined structure. This
gives rise to a basic question that could such device be transformable in shape and
also tunable in its quantum functions? Clearly, if the building components of a
quantum tunneling device can all be replaced by specific liquid or soft materials,
all-soft even liquid quantum tunneling device could be realized, which may provide
totally different performances compared to the traditional rigid ones. Such soft feature
would bring about possibilities for quantum tunneling devices to realize novel
capabilities such as flexibility, deformability, transformability, and controllability.
As the initial trial, a class of distinguishing candidates for the conductive soft
electrode material could be room-temperature liquid metals. Among them, the
gallium-based liquid metals (LM) especially combine the virtues of liquid (fluidity,
flexibility, and deformability) and the advantages of metals (high electrical and heat
conductivity) [7]. Additionally, it has been disclosed recently that the LMs would
display a series of intriguing interfacial phenomena in different types of medium, such
as acid solution [8, 9], basic solution [10-13], surfactant solution [14], and organic
gels [15], etc. When immersed in a specific solution, a LM-IL-LM sandwich structure
forms spontaneously, which reminds us that such structure would lead to a potential
quantum tunneling device given that the insulating layer could reach the characteristic
length (LC) for the quantum tunneling effect. It is based on this basic consideration,
2
we are dedicated here to propose a new conceptual quantum device which can be
termed as all-soft or all liquid quantum device (A-LQD) to innovate the conventional
rigid ones. The highly transformable performance of such kind of quantum device is
more conducive to the popularization and application of future low-cost and
intelligent quantum technology.
2. Quantum tunneling principles
The quantum tunneling effect can be explained in terms of uncertainty relations
[16, 17] and the wave–particle duality [18, 19]. Considering the case of a quantum
tunneling junction with a rectangular barrier (Fig. 2), a particle (e.g., electron) with
energy 𝐸(> 0) is shot toward the barrier in the x-axis direction. The potential
function of the barrier is:
Fig. 2. The schematic of a quantum tunneling junction with a rectangular barrier.
𝑉(x) = {
(I)
0, 𝑥 < 0
𝑉0, 0 < 𝑥 < 𝑎 (II)
(III)
0, 𝑥 > 𝑎
(1)
Using the treatment of wave–particle duality of matter in quantum mechanics, the
Schrödinger equation for a steady state implies [5]:
ℏ2
2𝑚
d2𝛹(𝑥)
d𝑥2 = [𝑉(𝑥) − 𝐸]𝛹(𝑥) (2)
Assuming the wave is incident from the left zone I (𝑥 = −∞), a portion of the
wave will be reflected at the left side of the barrier (𝑥 = 0) while the rest enters the
barrier (zone II). Likewise, a similar process takes place at the right hand side of zone
II (𝑥 = 𝑎). As a result, a portion of wave overcomes the barrier and enters zone III.
The wave functions in the three zones can be expressed as follows:
{
𝛹I(𝑥) = 𝑒i𝑝𝑥 ℏ⁄ + 𝑅𝑒−i𝑝𝑥 ℏ⁄
𝛹II(𝑥) = 𝐴𝑒i𝑝′𝑥 ℏ⁄ + 𝐵𝑒−i𝑝′𝑥 ℏ⁄
𝛹III(𝑥) = 𝑇𝑒i𝑝𝑥 ℏ⁄
(3)
where 𝑝 = √2𝑚𝐸, 𝑝′ = √2𝑚(𝐸 − 𝑉0), 𝑅, 𝑇, 𝐴, 𝐵 are parameters determined by the
convergence condition of the wave function on both sides of the barrier (𝑥 = 0, 𝑥 =
3
𝑎). Thus the reflection coefficient and the transmission coefficient of the barrier are
𝑅2 and 𝑇2, respectively. Since 𝛹(𝑥) and 𝛹′(𝑥) should be continuous at 𝑥 = 0
and 𝑥 = 𝑎, one has:
𝑇2 ≈
16𝐸(𝑉0−𝐸)
2
𝑉0
𝑒−2𝛽𝑎 (4)
where 𝛽 =
𝑝′
ℏ
= √2𝑚(𝑉0−𝐸)
.
ℏ
It can be seen from Eq. (4) that the particle has certain possibility to pass through
the barrier, and the tunneling probability decreases exponentially with the increase of
the barrier height 𝑎. And if 𝑎 could be adjusted on demand, the tunneling possibility
of particles at the junction could also be controlled, resulting in intelligently tunable
performance of the quantum tunneling devices.
3. Configurations of quantum device
3.1 Rigid and soft quantum device
As indicated in Fig. 3(a), the sandwich structure of a traditional rigid quantum
tunneling device is typically composed of three parts: two solid electrodes and an
insulating layer. The intermediate barrier may consist of a non-superconducting metal,
a thin layer of insulating liquid or air. Since tunneling effect occurs with barriers of
thickness about 0.1~10 nm, it usually requires precise manufacturing of the electrodes.
Besides, solid surfaces suffer from wear and corrosion, thereby reducing the
reliability and durability of the devices.
Fig. 3. The schematic of (a) rigid quantum device and (b) all-soft quantum device.
As explained in the former section, all-soft quantum devices can be obtained
when replacing solid conductors with conducting liquid electrodes made of
room-temperature gallium-based LMs (Fig. 3(b)). Without losing generality, such
insulating layer can be water, oils, surfactant solutions, hydrogels and many other soft
materials. Distinct from the traditional rigid-body quantum devices which cannot
change their structures to implement different applications, the tunneling junction of
the present all-soft quantum device can be easily transformed and manipulated
4
through applying external forces or fields, thus exhibiting high flexibility and
controllability. Soft quantum tunneling devices will not be limited by the traditional
sandwich structure and they can also work on various configurations.
3.2 All-liquid quantum device (A-LQD) with an insulating liquid layer
If appropriate components are adopted, an all-liquid quantum device (A-LQD)
can be constructed. In a recent study of a LM-electrolyte system, we have found that
LM droplets could persistently float on the same LM bath without coalescence when
subject to an electric field [20]. The electrolyte here serves as a solvent to remove the
oxide layer on the LM so as to maintain a perfect smooth surface. A thin running
liquid layer preventing the droplet from direct contacting with the lower LM bath was
found to be responsible for such phenomenon (Fig. 4(a)). A stable thickness of the
liquid layer 𝑒0 was determined to be of micrometer scale and it was a result of the
balance between the lubricating force (𝐹L) and gravity (G). For a transient case, a
nanometer or even infinitely small scale gap can easily be realized over the coalesce
process between the upper metal droplet and the lower liquid metal pool. According to
the lubricating theory [21], the lubricating force of the liquid layer reads as
𝐹L ∝ 𝜇E𝑣𝑅3 𝑒0
2⁄
(5)
where 𝜇E is the dynamic viscosity of the liquid, 𝑣 is its velocity, and 𝑅 is the
droplet radius. As can be seen from Eq. (5), 𝑒0 is a highly adjustable parameter and it
can be minimized by using smaller droplets or controlling other parameters, though its
usual length scale is beyond that of LC.
Fig. 4. (a) A LM droplet floating on a LM bath separated by a liquid layer (cannot be seen from the
image) under an external electrical field; (b) Resistance evolution of the liquid layer after cutting off
the applied voltage; (c) Schematics of the coalescence of a LM droplet and the successive satellite
droplet ejecting with corresponding stages indicated in Fig. 4(b). Note that original data comes from
the experiments in [20].
5
Clearly, once the electric field was cut down, the levitating LM droplet became
unstable and merged with the LM bath. The droplet then coalesced with the LM bath
accompanied with satellite droplet ejecting. As can be seen from our experimentally
measured development of the resistance between the two LM bodies (Fig. 4(b)), this
process began with an insulating state between the two and subsequently went
through a direct conducting state, which implied the existence of a period when the
thickness of the liquid layer could reach LC and quantum tunneling condition could be
satisfied. From a contrary way, such transient quantum tunneling conditions can also
be expected from the cases of either satellite ejecting process or successive
coalescence of the satellite droplets as well (Fig. 4(c)).
While the lubricating theory requires flowing insulating medium to generate a
stable liquid layer, there exist another mechanism to prevent coalescing and to
maintain a sandwich structure. This mechanism arises from
intrinsic
electric-double-layer (EDL) structure of the liquid metal/electrolyte interface (Fig. 5).
The liquid metal immersing in an electrolyte obtains a charged screen and two liquid
metal droplets thus can be recognized as two like-charge bodies. Therefore, the
droplets will experience repulsive force when they approach each other due to
electrostatic interactions. The screening length 𝜆D of the EDL can be obtained as [22]
the
𝜆D = √𝜀0𝜀r𝑁A𝑘B𝑇
2𝐹2𝐼
∈ O(nm)
(6)
where 𝜀0 is the permittivity of free space, 𝜀r the dielectric constant the electrolyte,
𝑁A the Avogadro constant, 𝑘B the Boltzmann constant, 𝑇 the temperature, 𝐹 the
Faraday constant, and 𝐼 the ionic strength of the electrolyte. Since for most cases,
𝜆D and LC are of the same scale, the electrostatic repulsion between the EDL of
neighboring LM droplet will be significant around LC which may potentially help
sustain the LM-IL-LM sandwich structure.
Fig. 5. Schematic drawing of the EDL structure of two adjacent liquid metal droplets in electrolyte and
the repulsive force produced through electrostatic interaction of the charged surfaces.
3.3 All-soft quantum device with an insulating thin film
The insulating junction will not necessarily be liquid alone. That means, a soft
layer can also be used as such junction for making transformable quantum device. In
6
this sense, an all-soft quantum device (A-SQD) can be fabricated. In the above cases,
oxidation of the LM surface is prevented by using dissolving electrolytes. As a matter
of fact, the nonconductive oxide layer on the LM surface can serve as a naturally
formed insulating layer. Importantly, the oxide layer formed on the LM under ambient
atmosphere usually has a thickness around 1 nm [23] which is feasible for realizing
quantum tunneling effect (Fig. 6(a)). Moreover, the thickness of the oxide layer can be
adjusted through controlling temperature and oxygen content which means the
tunneling junction can be flexibly tuned. The LM oxide film can also be preserved in
liquid environment like oil, ethanol, surfactant solutions, etc. In Fig. 6(b) and (c), the
large amount of closely packed LM droplets in a surfactant solution (Sodium dodecyl
sulfate, SDS) has been experimentally demonstrated [14]. Consequently, the proposed
soft LM quantum tunneling device can be made by directly combining two oxidized
LM electrodes.
More generally, the junction can also be made of other types of thin films. For
examples, thin insulating film of adjustable LC can be coated on the LM surface using
physical or chemical vapor deposition methods. Also, self-assembled monolayers
(SAMs) are typically within the LC scale and can be an alternative as well [23].
Fig. 6. (a) Schematics of all-soft LM quantum tunneling devices made by joining two liquid metal
electrodes coated with thin insulating film; (b) Optical image of stable micro liquid metal droplets
closely packed in SDS solution; (c) Microscopic image showing multiple point contact between micro
LM droplets.
3.4 All-soft quantum device with a compressible insulating layer
Regarding the A-SQD, more technical strategies could help make such elements.
Compressible materials such as elastomers and gels represent another class medium
which can serve as the insulating layer for the proposed all-soft liquid metal quantum
7
device. In what follows, we explain an easy going route to fabricate an array of
transformable quantum tunneling devices within a LM droplet-hydrogel (Carbopol
hydrogel) framework. As shown in Fig. 7(a), micrometer LM droplets can be well
distributed in the hydrogel [15]. Micrometer distance between the outer surfaces of
adjacent droplets (d0) can be readily achieved (Fig. 7(b)) and with better control
capability, i.e. sub micrometer distances are possible for realizing required functions.
Since the hydrogel is capable of performing elastic deformation, the original distance
d0 can be reduced through compressing action. Using a liner elasticity assumption, the
relation between the final distance d and the external stress ∆𝑝 can be obtained as
follows (Fig. 7(c)):
𝑑 = 𝑑0(1 −
∆𝑝
𝐸
)
(7)
where 𝐸 is the modulus of compression of the hydrogel. Fig. 7(d) plots 𝑑 against
∆𝑝/𝐸 for different original distance d0. For each curve, the adjustable range of ∆𝑝/𝐸
is located in the filled region. It can be seen that small starting distance d0 is favorable
in order to compress the hydrogel layer to the quantum tunneling scale LC (0.1~10 nm,
filled region in Fig. 7(d)). Also, a small d0 is favorable for a wider adjustable range.
Typically, d0 should be of sub micrometer scale for practical thickness control. A
rough estimate of the adjustable range of ∆𝑝 is on the scale of 10 to 100 Pa for
hydrogels with 𝐸=100~1000 Pa.
Fig. 7. (a) Microscope image of the LM extruded into Carbopol hydrogel [15]; (b) Microscale distance
between LM droplets; (c) A schematic drawing shows exteral stress compresses the hydrogel layer; (d)
The development of d as a function of normalized stress ∆𝑝/𝐸 for different original distance d0. The
filled gray rigion indicates d enters the characteristic scale for quantum tunneling.
8
4. Discussion
In this work, strong evidence for quantum tunneling effect in the LM-IL-LM
sandwich system is presented with potential quantum device configurations proposed,
though there is no direct measurement of quantum tunneling effect in the current
system. Straightforward advantages of the all-soft quantum devices over the
traditional rigid ones are also illustrated. As summarized in Fig. 8, the insulating layer
between LM electrodes can be water, oils, the native oxide layer of LM, SAMs,
hydrogels and many other soft materials. Besides, the soft quantum tunneling devices
will not be limited by the traditional sandwich structure and they can also work on
various configurations. Since the LM can be transformed into various shapes, the LM
electrodes can contact each other in different ways. For example, two spherical LM
drops exhibit a point contact, two LM films exhibit plane contact, and two deformed
LM drops may contact at more than one point. These manifold configuration modes
can render more complex structure of all-soft quantum device.
Fig. 8. All-soft quantum device: junction types, configurations and adjusting methods [18-20]
Overall, the present transformable all-soft quantum device owns a group of
unique merits which are different from the traditional rigid-body one. And it implies
important theoretical significance for developing future generation quantum devices.
As outlined as below, for one particular case, the gallium-based room temperature
LMs are quite suitable for manufacturing A-LQD in virtue of its unique advantages.
High electric conductivity. LMs generally own the inherent merit of good
electric conductivity [24]. The electric conductivity of gallium-based LMs is on the
order of 106 S/m, which builds a basis of the LM as the electrode material of
quantum devices.
9
Highly smooth and uniform surface. As we have mentioned in section 3.1, the
quantum tunneling only occurs in the case of a barrier thickness of 0.1 − 10 nm. For
traditional solid electrodes of quantum devices, high quality of perfect surface flatness
and uniformity are required, which significantly increases the processing difficulty
and manufacturing cost. If the electrodes of quantum tunneling devices can be
replaced by a material with highly smooth and uniform surface, these problems can be
easily resolved. Studies have shown that LM has extremely smooth surface in the
liquid environment compared to the solid surface, a hydrogen generation experiment
demonstrates that bubbles are all produced from the liquid-solid interface while no
bubble would appear at the LM-liquid interface since the latter provides little
vaporization core [25]. Therefore, the LM-I-LM structure offers a naturally-smooth
interface for making quantum tunneling junctions.
Controllability and transformability. In addition to utilize the external force to
directly change the droplet space, numerous external fields, including electric field,
magnetic field, chemical field, optical field, etc. can all be possibly adopted to
manipulate the droplet shape and the structure of the LM electrodes, thus the whole
A-LQD is highly flexible, changeable and tunable. This easy manipulation character
greatly extends the diversification and compatibility of A-LQD. In addition, since the
electrolyte interface can be tuned into rather small nanometer size while its
surrounding two neighboring liquid metal electrodes can be very large. In this sense,
the current A-LQD offers a straight forward to make macroscale quantum device
which will be rather easy to handle and operate.
Flexibility and diversity. The LM-I-LM sandwich structure is not just a
changeless framework. It can in fact be an adaptive combination of a variety of
transformable structures. For example, a series of quantum tunneling devices can be
formed spontaneously when LM droplets with different sizes are immersed in a
surfactant solution or a supporting hydrogel (Fig. 7(a,b)), because the LM droplets are
separated by a thin layer of insulating liquid [14, 15]. Additionally, the droplet size,
droplet space and solution type are all adjustable. These features present the LM
quantum device an unusually high degree of freedom, so as to provide a convenient
and efficient way to the fabrication of quantum devices with complex structures. In
the coming time, through constructing specific droplet arrays with designed logistic
structures, the current A-LQD or A-SQD can be developed as intelligent quantum
device including new quantum computational systems.
Easy manipulation: A significant feature of LM still lies in that its surface
tension is pretty high, which is almost ten times that of water (~700 mN/m) [26].
Consequently, a surface tension gradient across the LM body will result in a
propelling force, which is sufficient to cause droplet deformation and even drive small
droplets. Some external fields such as electrical field [27], magnetic field [28],
chemical field [11, 29], optical field [30], electromagnetic field [31] and combined
fields [32] etc. have all been found to effectively affect the deformation and
movement of LM through intentionally altering its surface tension. And the
controllability is pretty strong. For example, in certain case, through applying an
electric field, a large LM film can quickly transform into a tiny sphere, with over one
10
thousand times variation in specific surface area [27]. Accompanied by the chemical
field, LM can perform even more complex transformations. It is found that LM
droplet changes to a flat and dull puddle when deposited on a graphite surface, which
is easier to be manipulated as various stable shapes with sharp angles [12]. With these
means to flexibly manipulate LM objects in LM-based quantum devices, the shape of
the LM electrodes and the liquid interlayer thickness can be adjusted and regulated to
obtain the desired shape and function.
5. Conclusion
The all soft or liquid quantum device is a fundamentally conceptual innovation
over traditional rigid quantum devices, metal materials and understanding of solution
characteristics. Such type of unconventional quantum device owns a group of unique
advantages such as high flexibility, large deformability, convenient transformability
and easy manufacturability, including straightforward manipulation. This is highly
conducive to the polarization of low-cost quantum device and may promote the
development of future generation quantum technology. Further studies should focus
on the way to accurately control the thickness of the liquid interlayer, regulate the
whole A-LQD or A-SQD structure, and achieve a group of fully controllable
transformable quantum devices including developing their applications in intelligent
and quantum computing systems.
Acknowledgment
This work is partially supported by the Dean's Research Funding and the
Frontier Project of the Chinese Academy of Sciences.
References
1. Dirac PAM. The Principles of Quantum Mechanics [M]. Oxford University Press,
1981.
2. Serway R, Faughn J, Vuille C. College Physics [M]. Cengage Learning, 2008.
3. Trixler F. Quantum tunneling to the origin and evolution of life [J]. Current
Organic Chemistry, 2013, 17(16): 1758-1770.
4. Taylor J R, Dubson M A, Zafiratos C D. Modern Physics for Scientists and
Engineers [M]. Prentice-Hall, 2004.
5. Razavy M. Quantum Theory of Tunneling [M]. World Scientific, 2013.
6. Lerner R G, Trigg G L, Rigden J S. Encyclopedia of Physics [M]. Wiley-VCH,
1991.
7. Zhao X, Xu S, Liu J. Surface tension of liquid metal: role, mechanism and
application [J]. Frontiers in Energy, 2017. DOI: 10.1007/s11708-017-0463-9
8. Tang J, Zhao X, Li J, et al. Liquid metal phagocytosis: Intermetallic wetting
induced particle internalization [J]. Advanced Science, 2017, 4(5): 1700024.
9. Tang J, Wang J, Liu J, et al. Jumping liquid metal droplet in electrolyte triggered
11
by solid metal particles [J]. Applied Physics Letters, 2016, 108: 223901.
10. Zhang J, Yao Y, Sheng L, et al. Self‐fueled biomimetic liquid metal mollusk [J].
Advanced Materials, 2015, 27(16): 2648-2655.
11. Zavabeti A, Daeneke T, Chrimes A F, et al. Ionic imbalance induced
self-propulsion of liquid metals. Nature Communications, 2016, 7: 12402.
12. Hu L, Wang L, Ding Y, et al. Manipulation of liquid metals on a graphite surface
[J]. Advanced Materials, 2016, 28(41): 9210-9217.
13. Yi L, Ding Y, Yuan B, et al. Breathing to harvest energy as a mechanism towards
making a liquid metal beating heart [J]. RSC Advances, 2016, 6: 94692.
14. Yu Y, Wang Q, Yi L, et al. Channelless fabrication for large-scale preparation of
room temperature liquid metal droplets [J]. Advanced Engineering Materials,
2014, 16(2): 255-262.
15. Yu Y, Liu F, Zhang R, et al. Suspension 3D printing of liquid metal into self‐
healing hydrogel [J]. Advanced Materials Technologies, 2017: 1700173.
https://doi.org/10.1002/admt.201700173
16. Sen D. The uncertainty relations in quantum mechanics [J]. Current Science,
2014, 107(2): 203-218.
17. Heisenberg W. The Physical Principles of the Quantum Theory [M]. Courier
Corporation, 1949.
18. Greiner W. Quantum Mechanics [J]. Springer Berlin, 1994, 71(1): 1–24.
19. Eisberg R, Resnick R, Brown J. Quantum physics of atoms, molecules, solids,
nuclei, and particles [J]. Physics Today, 1986, 39: 110.
20. Zhao X, Tang J, Liu J. Surfing liquid metal droplet on the same metal bath via
electrolyte interface [J]. Applied Physics Letters, 2017, 111(10): 101603.
21. Couder Y, Fort E, Gautier C H, et al. From bouncing to floating: Noncoalescence
of drops on a fluid bath. Physical Review Letters, 2005, 94(17): 177801.
22. Devanathan M A V, Tilak B. V. K. S. R. A. The structure of the electrical double
layer at the metal-solution interface. Chemical Reviews, 1965, 65(6): 635-684.
23. Thuo M M,Reus W F, Nijhuis C A, et al. Odd-even effects in charge transport
across self assembled monolayer based Metal-SAM-Metal junctions. Journal of
the American Chemical Society, 2011, 133: 2962-2975.
24. Sen P, Kim C-J C. Microscale liquid-metal switches-A review [J]. IEEE
Transactions on Industrial Electronics, 2009, 56(4): 1314-1330.
25. Yuan B, Tan S, Liu J. Dynamic hydrogen generation phenomenon of aluminum
fed liquid phase Ga–In alloy inside NaOH electrolyte [J]. International Journal of
Hydrogen Energy, 2016, 41(3): 1453-1459.
26. Zrnic D, Swatik D. On the resistivity and surface tension of the eutectic alloy of
gallium and indium [J]. Journal of the Less Common Metals, 1969, 18(1): 67-68.
27. Sheng L, Zhang J, Liu J. Diverse transformations of liquid metals between
different morphologies [J]. Advanced Materials, 2014, 26(34): 6036-6042.
28. Tan S-C, Gui H, Yuan B, et al. Magnetic trap effect to restrict motion of
self-powered tiny liquid metal motors [J]. Applied Physics Letters, 2015, 107(7):
071904.
29. Zhang J, Sheng L, Liu J. Synthetically chemical-electrical mechanism for
12
controlling large scale reversible deformation of liquid metal objects [J].
Scientific Reports, 2014, 4: 7116.
30. Tang X, Tang S-Y, Sivan V, et al. Photochemically induced motion of liquid
metal marbles [J]. Applied Physics Letters, 2013, 103(17): 174104.
31. Wang L, Liu J. Liquid metal folding patterns induced by electric capillary force
[J]. Applied Physics Letters, 2016, 108(16): 161602.
32. Hu L, Li J, Tang J, et al. Surface effects of liquid metal amoeba [J]. Science
Bulletin, 2017(10): 700-706.
13
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.